T23A-0548 1340h
Frictional Melting can Terminate Seismic Slips: Experimental Results of Stick-slips
Whether frictionally melted layers are weak or strong is a question in issue. We conducted stick-slip experiments for granite samples at 150 MPa confining pressure using a tri-axial apparatus. The pre-cut surfaces were mirror finished. In order to detect the exact time of melting, we set sensors inside the pressure vessel; two strain gauges for measurement of axial stress and fault slip, two electrodes on a pre-cut surface to measure tribo-electromotive force, and a troidal coil for monitoring the current which flows along the slip zone. From the electrode potential and the potential induced in the coil we calculate the resistance of the slip zone which is expected to decrease by several orders of magnitude once the slip zone is melted. The signals from these sensors were recorded synchronously at 2 MHz sampling rate. A moderately large stick-slip event was analyzed in detail. The fault slip, stress drop, rise time and maximum slip velocity were 0.32 mm, 230 MPa, 23 $\mu$s and 40 m/s. The sensors detected precisely the time point when the slip zone melted. This occurred only 2 $\mu$s after the slip velocity reached the maximum, and at the same time the friction coefficient reached a minimum (0.3). Immediately thereafter, it recovered promptly and remarkably, and the slip stopped eventually. Our SEM and EPMA observations ascertained the melting of the slip zone that was evidenced by a glass layer a few $\mu$m thick in the experimented sample. The early half of the slip event is assumed to have been governed by solid interface friction, because carrot-shaped grooved and blobs of scratched debris were well developed in other experimented samples which experienced small events with ca. 0.1 mm slip. Our numerical simulations for frictional melting using observed time-shear stress and/or time-slip velocity data successfully reproduced the temperature and thickness of the melt layer, validating our experimental result at least phenomenologically. Therefore, we conclude that frictional melting is of the potential to stop seismic slips. We conducted the numerical simulations also for the spring-slider model with viscous friction. They were associated with a long tail of dumped slip velocity, while the tail was obviously shortened in our experiments. Our stick-slip experiments and numerical simulations demand a mechanism that melt layers are enforced to be cooled and strengthened. This mechanism is discussed in Otsuki and Koizumi (next presentation).
T23A-0549 1340h
Melt Patches Formation during Dynamic Fault Motion; its Braking Effect and Implications for Modeling Incipient Frictional Melting
Following the first slip-weakening associated with flash heating at asperity contacts [Rice, 1999], frictional melting occurs in three stages; (1) incipient frictional melting, accompanied by marked strengthening of a fault, where melt patches form and develop into a continuous molten layer, (2) growth of molten layer resulting in the second slip-weakening, and (3) steady state with a nearly constant shear resistance where melt production is nearly balanced with melt loss. Growth of molten layer (stage 2) can be modeled as a Stefan problem with viscous shearing of molten layer as a heat source and with melting surfaces as moving boundaries [Hirose & Shimamoto, 2002]. Satomi & Shirono [2003, 2004] and Matsuzawa & Takeo [2004] solved frictional melting problem as a 1D Stefan problem and obtained results in reasonable agreement with laboratory data. In particular, the latter authors analyzed rupture propagation along a fault incorporating frictional melting. The incipient frictional melting and melt loss have to be included for complete modeling in the future. Modeling the incipient frictional melting is difficult because formation of melt patches is a 3D problem, so that the above authors started modeling from a continuous molten layer. This paper focuses on how melt patches grow to form a molten layer during the incipient frictional melting, based on frictional melting experiments on gabbro. This process causes marked strengthening of a fault and is significant to evaluate a breaking effect of the incipient frictional melting on fault motion. Hollow-cylindrical specimens of gabbro were sheared dry to 3.6-76.5 m in displacements at a slip rate of 0.85 m/s and at a normal stress of 1.3 MPa using a rotary-shear high-velocity friction apparatus in Kyoto. Frictional coefficient $\mu$ is $\sim$0.9 at the first peak-friction and it drops to transient steady-state friction ($\mu$ of $\sim$0.35) after the first weakening with the slip-weakening distance of $\sim$0.5 m. Melt patches of 7-10 $\mu$m in thickness and of 110-160 $\mu$m in width form sporadically on fault surfaces just after the first slip-weakening. Melt patches first increase in number without much change in their geometry, and then begin to coalesce, thickens to 16-20 $\mu$m and widens to 2-4 mm. This process eventually leads to the formation of a continuous molten layer of about 20 to 25 $\mu$m in thickness near the second peak-friction. Fault area occupied by melt patches increases from about 10 % at the onset of melt patches formation to 70-80 % close to the second peak-friction. Viscous shear resistance of melt patches no doubt causes marked strengthening of a fault, but measured shear resistance alone cannot separate contributions from solid friction and from melt patches. Infinitely-thin molten layer poses infinite shear resistance for a given slip rate. Nature avoids such difficulty by forming melt patches of finite thickness. What determines the geometry of melt patches and how they deform under extreme shear would be a clue to model the incipient frictional melting.
T23A-0550 1340h
Seismicity without a fault ? Structural evidence from pseudotachylites in the UHP Dora Maira Massif
The Dora Maira massif, in the Italian Alps, is a world famous example of ultra-high pressure (UHP) metamorphism of crustal rocks with peak pressure of ca 3 GPa (depth $\approx$ 100 km) at about 35 My. Exhumation began shortly after the peak of metamorphism. The UHP rocks had equilibrated at a temperature of $250\deg$C around 30 My. Pseudotachylite veins (1 to 20 mm thick) formed parallel to the syn-exhumation mylonitic foliation and were dated at 20.1 $\pm$ 0.5 Ma by Ar/Ar method. The geochronologic data suggests that the pseudotachylite veins formed at depths shallower than $\approx$ 10 km. These veins are found mainly in gently dipping medium grained phengite/coesite bearing orthogneiss. The generation veins are parallel to the foliation, constant in thickness and up to several tens of meters in length. Injection veins are less common and generally only a few cm in length. Structural mapping of UHP mylonitic gneisses and surrounding rocks reveals the dominance of high ductile strain, a puzzling lack of localized cataclastic deformation and the absence of fault associated with the pseudotachylite veins. The pseudotachylite veins are concentrated along a 50 m wide, foliation-parallel corridor. Conversely, the two major faults along which the UHP rocks were exhumed are $>$ 500 m away from the pseudotachylite-bearing rocks and do not contain pseudotachylite veins. The formation of these pseudotachylite veins is clearly related to exhumation history of UHP rocks. The formation of pseudotachylite results from frictional melting as shown by microstructural criteria including corroded quartz grains and spherulitic microstructures. The localized heat source required for melting the UHP orthogneisses along narrow bands is attributed to frictional heat produced along the foliation plane. Fault-related pseudotachylite studies and friction experiments have shown that seismic strain rates (in the order of s$^{-1}$) are required to maintain heat along the fault plane. At slower strain rates, heat would be transferred to the host rock which then would behave in a ductile manner. These new results expose a paradox in which pseudotachylites originated from seismic rupture along foliation planes (long generation veins) but these planes did not further localized deformation and did not become fault planes. We propose that exhumation deformation became localized along the two bounding faults when the UHP rocks moved across the isograd corresponding to the quartz brittle-ductile transition.
T23A-0551 1340h
Evidence of Seismic Rupture Directivity from Pseudotachylyte-bearing Fault Vein Networks (Adamello, Southern Alps, Italy)
East-West striking dextral strike-slip paleoseismic faults and branching fractures networks associated to the major Tonale fault are exposed in vast glacially polished outcrops in the Adamello massif (Italian Alps). As faults and fractures are filled by solidified friction-induced melts (pseudotachylytes), the formation of fractures was coseismic. Ambient conditions during faulting were 9-11 km in depth and $250-300\deg$C. Most of the branching of secondary fractures occurs at $30\deg$ (consistently with the inferred orientation of the principal horizontal stress) and $90\deg$ to the strike of the main faults, and, most interestingly, fractures take off preferentially on one side of the wall rocks, i.e., into the southern block. We interpret these features in terms of stress fluctuations associated to a propagating fracture. With a dynamic faulting numerical simulation we estimate the stress fluctuations at the time of fracture propagation. The numerical model is in agreement with field observations, host rock mechanical properties and other geological constraints. The preferential attitude of the observed fractures is compatible, in terms of the dynamic Coulomb stress fluctuations, with a fracture starting at the western end of the fault (where it connects to the main Tonale fault structure) and propagating toward the East. Numerical modeling together with field observation place several constraints on the friction level and the dynamic stress drop. Non-linear processes occurring in a lateral band off-fault, such as secondary faulting, or diffuse material damage such as breccia and cataclasite formation, should contribute to the amount of energy flow G, the dissipation associated to fracture propagation. In this case we find negligible amount of diffuse material damage, but rather numerous secondary fractures whose contribution to the energy balance may be estimated. Finally, sensible amounts of melt in some portions of the faults, mainly injected into the lateral secondary fractures, yield further information on the coseismic thermal regime and the possible influence of partial melting on friction and pore pressure fluctuations.
T23A-0552 1340h
Occurrence of pseudotachylytes from cataclastic-plastic transition and near-surface crustal levels
This paper introduces natural examples of pseudotachylytes formed from different crustal levels from the cataclastic-plastic transition (10-15km) to near-surface (-1km). Pseudotachylytes from the cataclastic-plastic transition regime have been investigated from two shear zones in Japan: the Hatagawa Fault Zone in NE Japan and the Asuke Shear Zone in SW Japan. All studied pseudotachylytes have the following characteristics. 1. Each pseudotachylyte-bearing shear zone is relatively narrow, and a close occurrence of pseudotachylyte, cataclasite and mylonite (especially ultramylonite) is commonly observed. 2. Mylonitized pseudotachylytes or mylonitized cataclasites occur as well as fractured mylonites. 3. Polyphase generation of pseudotachylytes (pseudotachylyte vein transecting previous pseudotachylyte or co-occurrence of non-deformed and deformed pseudo-tachylyte) is observed. The Asuke Shear Zone is a sinistral-extensional shear zone dominantly composed of small-scale cataclasite zones often associated with pseudotachylyte and ultramylonite. One of the major factors that determines whether pseudotachylyte is associated or not is the attitude of fault surface: faults parallel to the P shear sense tend to be associated with pseudotachylyte when compared to faults parallel to the R1 shear sense. This implies that transpressional shear zones, where normal stress is relatively high, has greater frictional melting compared to transtensional shear zones. On the other hand, pseudotachylytes from Langtang, Nepal occur in a near-surface (about 700 m depth) regime and are considered to have been generated during large-scale landsliding. The evidence for landslides is not only topographic features but also kinematic indicators in foliated fault breccia overlying pseudotachylytes showing normal fault sense, and the pseudotachylyte itself contains many large bubbles and is fairly glassy. Even the normal stress must have been quite low there, but frictional melting during the landslide was possible due to the high velocity of sliding. Recent Fission-Track ages of zircon in pseudotachylyte from the Asuke Shear Zone (c. 53 Ma: Murakami et al., 2003) and from the Langtang (c. 50 ka: Takagi unpublished data) are successful examples of the dating of frictional melting.
T23A-0553 1340h
Dramatic frictional-viscous slip fluctuations within an exhumed multi-strand fault; evidence of fluid- with velocity-sensitivity?
Detailed field surveying of exhumed fault strand domains on the pull-apart basin flanks of a still active transtensional fault zone reveal intriguing fault rocks that are highly informative of the relationships between aseismic creep and co-seismic accelerated slip. We examine part of the 100's km Damxung-Jiali Shear Zone (DJSZ) in eastern Tibet, a crustal-escape-tectonics lithospheric failure zone whose locally exceptional width (5-15 km) incorporates a range of protoliths. This wide, multi-lithology fault architecture is associated with extensive pseudotachylite and cataclasite development, co-located with a range fault rocks that failed with plastic bulk-rheology. Highly unusual S-C fabrics in cataclasites are identified as well as spectacular subsequent re-brecciation indicating dramatic temporal fluctuations in viscous and frictional failure processes. Their co-location across a broad (original) depth interval together with numerous pseudotachylite generations (from co-existence with quartz flowing at 300-350 deg.C to low-cohesion cataclasites) indicates repeated excursions into velocity weakening exploited instabilities as displacement has continued. Extensive fluid-involvement in both frictional and viscous processes in all the fault strand lithologies is taken as evidence of "fluid sensitivity" as well as velocity sensitivity. Pseudotachylite-coated brittle localisation surfaces that short-cut stronger domains between creeping (i.e. metavolcanics versus carbonates) indicate the interplay of stable creep versus co-seismic slip acceleration with velocity weakening amongst the strand domains. We propose that the multiple strand and multi lithology nature of the fault zone allowed unstable stress/slip values to be preferentially enhanced in key fault strands (i.e. restricted cross-fault permeability "compartments") while other compartment were reciprocally undergoing stable creep. This provides a view of the potentially very complex nature of co-seismic slip frictional-viscous interplay and evolution in a heterogeneous, large fault zone.
T23A-0554 1340h
Seismogenesis and Strain Localization Across Accretionary Prisms: the Role of Decollement and Out-of-Sequence Thrusts.
The most destructive earthquakes occur along the decollement thrusts at convergent margins. Neverthless, the seaward limit of recorded microearthquakes seems to correlate with zones where out-of-sequence thrusts (OSTs) splay from the decollement, suggesting the significant role of these thrusts in releasing seismic energy. In the Kodiak accretionary complex of Alaska, representing the ancient analogue of the Aleutian margin, well exposed examples of both subduction and OS thrusts occur. Both thrusts were primarily active between 59 and 65 Ma. In the Paleocene Ghost Rock Fm, a map-scale melange primarily comprised of turbiditic argillites, variably continuous massive sandstones and rare greenstones, two approximately 15m thick bands of highly sheared cataclasites have been mapped. The cataclasites, sub-parallel to the melange fabric, represent episodes of localized shear in the melange during subduction thrusting at about 13 Km of depth. Extreme strain localization into the cataclastic thrust zones is testified by development of ultrafine fault rocks occurring as tens of cm thick planar to irregular beds. They show ductile flow which intrude and deform the cataclasite, and can be described as ultracataclasites and/or pseudotachylites. The Uganik Thrust (UT) juxtaposes the early-mid Cretaceous Uyak Complex over the latest Cretaceous Kodiak Fm, represented by its tectonized upper boundary, known as Waterfall Bay Melange (WBM). The WBM is interpreted as formed by flattening and shearing of the coherent turbiditic Kodiak Fm during underthrusting along the decollement. The steep geometry of the UT and the overprinting relationship to the melange, allow the interpretation of the UT as an OST. UT core deforms through Riedel fractures separating domains where sandstone blocks are elongated parallel to an S-foliation into the argillite. Several narrow cataclastic zones within the footwall accommodate significant shear. These shear zones crosscut the previous melange fabric and contain evidence of fluid flow, in the form of abundant quartz precipitates along the shear fabrics. Both thrust examples include strain localization features (pseudotachylites and quartz-infused shear zones) consistent with stick-slip behavior. Thus, ancient OSTs in addition to the decollement thrust may have helped generate large earthquakes along this ancient plate boundary in a relatively constrained time frame.
T23A-0555 1340h
Decimeter Scale Ultra-Fine Fault Rocks (Possible Pseudotachylites) in an Ancient Subduction Thrust Zone
Large bodies of ultrafine fault rock (possible pseudotachylite or frictional melt) occur within cataclastic thrust zones in the Ghost Rocks Formation, Kodiak Accretionary Complex, Alaska. The Paleocene Ghost Rocks Formation includes map-scale melange belts formed by flattening and shearing of seafloor sediments and volcanic rocks at about 250 degrees C and 325 MPa (~13 km depth) during subduction between 65-60Ma. Ten to 15-meter thick cataclastite zones crosscut the melange fabric at a low angle, representing a stage of increasingly localized shear during subduction thrusting. Ultrafine fault rocks occur as thick (10-25cm) continuous planar beds along the boundaries of cataclastites, or in discontinuous accumulation bodies within cataclastite zones. The boundaries of the ultrafine fault rocks are intrusive, sharp but irregular and deform the cataclastite host fabric. Single pulse intrusions of the ultrafine fault rock range up to 0.5m in intrusive dimension and form complex morphologies resembling both upward and downward directed flame structures and dike-sill complexes, as well as sheath folds and disharmonic flow banding and folding. These field characteristics indicate fluidization and perhaps frictional melting of the ultrafine fault rocks. Ultrafine fault rock bodies can be traced laterally for meters to tens of meters at individual outcrops and occur for about 2 km along strike. Preliminary SEM analysis reveals that the primary matrix material is physically and chemically homogenous down to few-micron scale, consistent with the field identification of pseudotachylite. Thin sections show rounded remnant quartz aggregates, typical of pseudotachylytes. Although some thin sections show suggest melting others may represent ultracataclastite. Some ultrafine fault rock material is rebrecciated and cataclastized to a fine scale, indicating reactivation of previous fault rock generation surfaces. These ultrafine fault rock zones represent the most highly deformed slip surfaces in Ghost Rocks melange and are prime candidates for failure surfaces of large earthquakes.
T23A-0556 1340h
Some Mechanical Implications of the Development and Evolution of Y shears in Simulated Granite Gouge
Frictional sliding experiments were performed in a rotary shear machine at 25 MPa normal stress on 2-mm thick layers of simulated Westerly granite gouge(particle size$\leq$85$\mu$m)with the objective of studying microstructural aspects of displacement dependence of the frictional behavior of the gouge. Sliding velocity was regularly stepped between 1 and 10$\mu$m/s in all the experiments. Successful trials were terminated for examination of the samples after: (1) 60mm of sliding involving a persistent reduction in friction, (2) 144mm of sliding involving that reduction followed by a persistent increase in friction, and (3) 386mm of sliding involving several varied amplitude fluctuations in friction. The sheared gouge sections were imaged at 0.25K-64K times magnification in SEM/BSE mode. In experiment (1) a highly comminuted zone was separated from marginal gouge by a sharp particle size gradient. A single 5-10$\mu$m-wide linear Y slip surface appeared to have nucleated within the highly comminuted zone. The gouge also included a field of large survivor particles interlaced by numerous R slip surfaces. In contrast, a complex system of microstructures including a full set of Riedel shear surfaces and pervasive comminution had occurred in experiment (3). The most notable microstructure, absent in both (1) and (2), was an intermittent central zone of faulted, thrusted, and rotated sliver-shaped particles. The slivers appeared to be riding in a less competent surrounding gouge. Detailed microscopy revealed that in every case the slivers (5-25$\mu$m thick) have an internal layering defined by a mass of particles graded in size from 20nm on one side to about 300nm on the other side. At the highest magnification the sliver material had an image-estimated porosity of 10-15$%$, and consisted of a mixture of rounded and sub-rounded quartz and feldspar particles. The correlation of microstructural and mechanical data in experiments (1) and (2) is generally consistent with previously proposed models (Beeler et al. 1996) that associate Y slip surfaces with significant shear localization and reduction in friction followed by hardening of the localized layers and consequent delocalization. The data suggests that fluctuations in friction observed in experiment (3)may be related to occurrences of hardening and subsequent brecciation of the gouge during slip on multiple Y shear surfaces as well as to the discontinuous nature of the sliver breccias along the gouge zone. The internal layering of the brecciated slivers and their well-developed particle size grading indicates that once created Y slip surfaces may continue to localize shear strain until particles that line the slip surface are reduced to a critical average size or perhaps achieve a critical packing density. The mechanism for hardening of gouge and disruption into the observed slivers, although appears to be size-related, is not well understood. If subsequent studies show that it does involve an increase in packing density due to wider PSD, then the hardening could result from an increase in the contact area per volume or the contact area along potential shear surfaces.
T23A-0557 1340h
The differences in clay minerals between the northern and southern Chelungpu fault, Taiwan
In 1999, we obtained a detailed data about motion of fault from the Taiwan Chi-Chi earthquake. The motion represents the high frequency of acceleration and small slip distance in southern part, and low frequency of acceleration and large slip distance in the northern part. Those differences in the fault motion between the southern and northern parts are coincidence with occurrences of deformation textures of rocks which were sampled by drilling of shallow parts (a few hundreds meter) of the fault in 2000. In the southern core, a relatively strong deformation structure is preserved in total, and gouge containing fragments of pseudotachylytes and ultracataclasites is observed at the Chi-Chi- earthquake fault, which indicates that the main deformation mechanisms for the southern part of the fault was brittle. On the other hands, in the northern part, sand layer with much amount of water is found at the Chi-Chi- earthquake fault zone, and no breakage of sand grain is observed, which suggests that the deformation mechanism for northern part is independent particulate flow. The purpose of this study is to reveal the differences in clay minerals between the southern and northern part of the Chi-Chi earthquake fault. And then, we discuss about rock-fluid interaction and frictional heating characterized in seismogenic fault system. We analyzed clay minerals by X-ray diffract meter (XRD) after classification of rock types such as sandstone, alteration of sandstone and mudstone, breccia, and gouge. 1.33 micron meter of grains are obtained. Oriented sample was made. XRD analysis was conducted under following condition; 35kV, 15mA, 1 degree per minute of scan rate, and 0.02 degree of scan step. Range of 2 theta was from 2 degree to 35 degree. At first, air-dried condition of samples was measured. After that, ethylene glycol solvated samples were measured. The result represents that all samples contain smectite, illite, chlorite. No difference in components of clay mineral is observed between the southern and northern site. We focused on Chlorite which is contained in all analyzed samples. Total number of iron and magnesium (Y value), or asymmetry of iron between silicate layer and hydroxide layer (D value) is affected by pH of fluid when the chlorite is precipitated. In the northern site, Y value increases and D value decrease in gouge relatively to that in other rocks. On the other hand, in southern site, Y value decreases and D value increases in gouge which is located just at the Chi-Chi earthquake fault. This result suggests that pH of fluid differed at the time of fault activities. One of the possibilities of this cause is radical reaction. The differences of deformation mechanisms between the southern site and northern may be affected by whether radical reaction occurred or not at the time of chlorite precipitation. Other characteristic in clay mineral is smectite consumption. Illite% to smectite of host rock represents about 96% in average. Whereas, some gouge samples have no smectite. Smectite-illite transition is mainly affected by temperature. Treated samples are about 15cm in size. Therefore, this smectite consumption may be due to localized heating, which may be frictional heating. Temperature loggings after 1.5 year from the Chi-Chi- earthquake represent temperature anomalies both in the northern and southern site. If those temperature anomalies are due to frictional heating, localized increment in temperature is about a few tens degrees Celsius in the order. This temperature rise can not consume smectite to transit to illite from 96% to under 1%. Therefore, the smectite consumption may be integrated result of repetition of frictional heating in seismic history.
T23A-0558 1340h
Large-magnitude Dextral Slip on the Wairarapa Fault, New Zealand
Dextral slip associated with an 1855 Ms 8.0+ event on the Wairarapa fault near Wellington, New Zealand was reported to be 12+/-1 m along a rupture length of at least 148km (Grapes, 1999), one of the largest single-event strike-slip offsets documented worldwide. Initial results from a new study involving detailed neotectonic mapping and microtopographic surveys of offset landforms (including many beheaded, inactive streams) strongly suggest that dextral slip was as much as 50% greater than previously measured. 1855 surface ruptures were mapped with certainty where a linear scarp characterized by steep slopes (30-90°) and exposed alluvium cuts across active or inactive stream channels. The fifteen individual strands comprising the Wairarapa fault zone that we have mapped to date are 1200+/-700 m long and typically left-stepping. Slip in the stepover zones between these strands is distributed amongst two or more ruptures and intervening anticlines, a situation that causes along-strike variations in slip and which locally complicates the interpretation of 1855 displacement. We focused on seven of the best-preserved sites where low-discharge streams are disrupted by the fault zone, including five that had been previously attributed by Grapes (1999) to coseismic slip during the 1855 earthquake. One of these (Pigeon Bush) includes two sequentially displaced, now beheaded linear stream channels, oriented perpendicular to the fault scarp, that preserve distinct offsets with respect to a single deeply incised, originally contiguous gorge on the opposite side of the fault. To quantify the minimum fault displacements at each site, we made 1:500 scale topographic maps employing n = 2,000-10,000 points collected with GPS and laser instrumentation. Measured dextral slip values, here attributed to the 1855 earthquake, include 16.4+/-1.0m (Hinaburn), 12.9+/-2.0m (Cross Creek), 17.2+/-2.5m (Lake Meadows), 18.7+/-1.0m (Pigeon Bush), 13.0+/-1.5m (Pigeon Bush 2), 15.1+/-1.0m (Pigeon Bush 3), and 16.0+/-1.5m (Tauwharenikau). Reverse slip at these localities ranged from 0.5 to 3.8+/-0.5m. Tape measurement of two other offset streams in dense bush yielded two further dextral slip measurements of 13.5+/-0.5m and 17.5+/-1.5m. AMS radiocarbon dating was undertaken at two sites to test whether slip occurred during one rupture event in 1855, or possibly could have accrued as a result of two or more earthquake ruptures. At the classic Pigeon Bush site, the youngest, most proximal beheaded stream channel is partially infilled by fluvial conglomerate with abundant charcoal. Two samples at depths of 15cm and 154cm yield calibrated dates of AD 1364+/-63 and AD 1355+/-60 (all dates are 2 sigma intervals). The stream must have been offset and abandoned after this time, and with no historical record of any other local earthquake, the 18.7+/-1.0m offset at Pigeon Bush is inferred to have occurred entirely in 1855. At Tauwharenikau, an abandoned channel is underlain by gravel on the upthrown side, but this gravel is overlain by 30 cm of swamp deposits on the downthrown side. We interpret the swamp deposits to reflect post-rupture incursion of groundwater into the down-thrown block. Basal swamp grasses yield calibrated dates of AD 1709+/-26 (27% probability) or 1869+/-60 (71% probability) for one sample and AD 1723+/-49 (34% probability) or 1871+/-70 (64% probability) for another sample, evidence that the 16.0+/-1.5m of slip at Tauwharenikau occurred entirely in 1855. These initial results strongly support the assertion that the southern Wairarapa fault experienced the largest single-event strike-slip offset yet documented worldwide.
T23A-0559 1340h
Dramatic slip weakening of Nojima fault gouge at high-velocities and its implication for dynamic fault motion
The use of Teflon sleeve around simulated gouge between cylindrical specimens has made it possible to perform high-velocity friction experiments on fault gouge, using a rotary-shear high-velocity frictional testing machine in Kyoto [Mizoguchi and Shimamoto, 2002, 2003, 2004]. We present a summary of our experimental data on Nojima fault gouge and demonstrate its significance on seismogenic fault motion. A series of experiments were examined at constant normal stresses of 0.3 to 1.8 MPa and a constant equivalent slip rates of 1.03 m/s [1200rpm], with unconfined and dry conditions. The fault gouge used in the experiments is blue gray gouge derived from granodiorite collected at the Hirabayashi trench along the Nojima fault. The representative mechanical behavior of a simulated fault at the condition, that the normal stress is 0.6 MPa and rotation speed is 1200 rpm, is as follows. At the initiation of a run, friction coefficient rapidly increased to about 0.8 and then decreased gradually. The friction coefficient finally attained to nearly constant [-0.3]. The slip weakening displacement [Dc] to the residual friction was around 30. We also measured temperature of 4 positions in solid-cylindrical granite specimens by using CA thermocouples and then calculated temperature distribution in the specimens numerically to estimate temperature of fault gouge. The results indicate that the maximum temperature of fault gouge during high-velocity friction is around 380 _E#381;. Local temperature in the gouge zone might be higher than the estimated value due to shear localization, but the bulk of the gouge would not melt. Friction and Dc are important parameters to analysis earthquake generation process. The frictional coefficient inferred from the heat flow measurements along the San Andreas Fault was lower than 0.2 and the inconsistent with the laboratory tests confuses many seismologists for a long time [reviewed in Zoback (2000)]. However our results in laboratory suggested that frictional strength of a fault at seismic slip rates might be enough low to explain the low frictional strength of natural faults. As to Dc, there also is a wide difference between laboratory friction test at slow slip rates and seismic wave analysis. The difference is thought to be due to the scaling of fault surface topography. The Dc at low normal stress range from 20 to 40 m and are one order of magnitude larger than the values estimated from seismic wave [Ide and Takeo, 1997; Olsen et al, 1997]. However the Dc at the higher normal stress is the same order of the Dc based on seismic wave analysis. We can explain the inconsistency about the Dc by high-velocity frictional properties of fault gouge. Although weakening mechanism of gouge is unknown at present, our results as well as Goldsby and Tullis (2003) and DiTro et al. (2004) indicate that there is another slip-weakening mechanism enhancing fault instability, in addition to thermal pressurization and frictional melting.
T23A-0560 1340h
Laboratory Experiments of Silica Powder Lubrication Between Rock Faces at Coseismic Velocities
One of the unresolved problems in earthquake mechanics is the physical process controlling friction on faults during the rupture of large earthquakes. Many studies suggest that coseismic friction is low even at great depths and several mechanisms have been introduced to explain these observations. In these experiments, we attempt to investigate the physics of mechanical lubrication between rock surfaces by using dry powder. To simulate rock friction, we utilize a tribo-rheometer where two novaculite disks, with 1-inch diameter and 5-micron surface roughness, are compressed together with a thin layer of 5-micron silica powder applied in between. The tribo-rheometer is a highly sensitive instrument that measures torque and normal force when a test substance is placed between the rotating plates. The measurements can be used to directly calculate the viscosity and the friction coefficient. These experiments investigate the velocity dependence of friction by rotating the top disk through velocities from 10$^{-3}$ to 10$^{2}$ rad/sec while the normal stress is kept constant on the order of 10$^{4}$ Pa. The preliminary experiments show frictional regimes of boundary, mixed, and hydrodynamic lubrication; together known as the Stribeck curve. At high shear rates of $>$10 rad/sec, hydrodynamic lubrication occurs when fluid-like behavior of granular flow are responsible for the shear stress between the surfaces. In contrast, boundary lubrication has full asperity contact between the top and bottom surfaces during low shear rates of $<$0.01 rad/sec and shear stress arises from physical interactions. Between the two regimes above, the mixed lubrication is where there is a combination of surface asperity and powder lubricant interactions. From the data, we find the friction coefficient drops from a boundary lubrication value of $\sim$0.3 -- 0.4 to a mixed regime minimum of $\sim$0.2 -- 0.3 while transitioning to the hydrodynamic lubrication. The transition corresponds to a change from solid-friction behavior to viscous fluid-like resistance. More experiments are planned to investigate normal stress dependence.
T23A-0561 1340h
Permeability structure of a serpentinite-bearing fault zone along the Gokasho-Arashima Tectonic Line, Central Japan and its implication for thermal pressurization
Serpentinites are often trapped in large-scale fault zone, but no thorough studies have been conducted on internal and permeability structures of serpentinite-bearing fault zones. Kurosegawa tectonic zone, extending over several hundred kilometers across Shikoku Island and Kii Peninsula in Japan, is well-known for bearing numerous large serpentinite and other tectonic blocks. Gokasho-Arashima tectonic line is a fault in this zone in the eastern part of Mie Prefecture, central Japan. A large serpentinite body up to 600 m in thickness is adjacent to Cretaceous Matsuo Group, composed mostly of alternated sandstone and shale, across this tectonic line. Fault zone of this tectonic line at Matsuo outcrop consists of (1) fractured sedimentary rocks and fault breccia derived mostly from Matsuo Group ($>$ 30 m in width), (2) clayey fault gouge of about 50 mm in width, (3) foliated gouge of about 0.6 - 0.9 m in width, and (4) brecciated serpentinites ($>$ 250 m). Permeability has been measured with an intravessel deformation-fluid-flow apparatus in Kyoto on fault rocks taken from each structural division (1) to (4) above, using steady-flow method with Nitrogen gas as the pore fluids. Measurements were done during confining pressure cycling (from 5 MPa up to 120 MPa, and then back to 5 MPa, corresponding roughly to 0.3 - 7.5 km in depth. Permeability values at 100 MPa are 10-15 to 10-17 m2 for zone (1) above, around 3 x 10-18 m2 for zones (2) and (3), and greater than about 10-16 m2 for zone (4). Thus the fault zone is characterized by fault gouge (fault core) with low to intermediate permeability and by permeable damage zone. We also have measured constitutive properties of this fault zone as studied with a biaxial frictional testing machine and will report on these in our presentation. The possible effect of thermal pressurization in this fault zone was evaluated using the same method as Noda et al. (2003), a numerical analysis based on Lachenbruch (1980), fully incorporating changes in permeability, porosity and physical properties of water with build-up of pore pressure. The effect of large shearing deformation is not taken into account. Initial parameters were chosen from measured data at an effective pressure corresponding to 3 km in depth and analyses were done for deformation-zone widths of 10, 40 and 50 mm. In all cases thermal pressurization caused slip-weakening with characteristic weakening distance of a few meters. Permeability of fault gouge is not very low, so that thermal pressurization is not very effective for the present serpentinite-bearing fault zone. Key words: serpentinite-bearing fault zone, thermal pressurization, dynamic fault motion, fault rock, permeability
T23A-0562 1340h
Frictional Properties of Feldspar Gouge Under High-Pressure and High-Temperature and Their Implications of Seismogenic Process
Understanding physical and chemical processes in the deep extensions of seismogenic fault zones is very important, since most of the main rupture of intraplate large earthquake starts at the base of the seismogenic zone. Geological studies on the exhumed fault zones, geophysical measurements of natural fault materials from seismogenic zones, and laboratory experimental studies under high-pressure and high-temperature conditions are carried out. Although it is essential to understand the mechanisms by which inland earthquakes occur and to establish physics-based forecast methods, the mechanism by which inland earthquakes occur is poorly understood in contrast to interpolate earthquakes. In order to understand the earthquake generation process, we need to understand the frictional and rheological properties of fault zone materials under high-pressure and high-temperature conditions. Laboratory data on frictional properties of fault surfaces of fault zone rocks are useful for understanding the slip process in deep extensions of seismogenic zones. Frictional properties of feldspar and quartz gouges under high-pressure and high-temperature conditions were obtained. In the laboratory study, we use the information on the natural fault materials based on the results of the geological study. We conducted geological studies on the exhumed seismogenic fault zone in Japan, the Hatagawa Fault Zone, to characterize fault zone materials and fault zone structure. We conducted frictional experiments (the velocity-stepping test) by using feldspar and quartz gouges (about 3 micron diameter) under high-pressure and high-temperature in the wet and dry conditions. Temperature varied from room temperature to 600C. In the dry conditions, experiments were conducted under the confining pressure of 150MPa. In the wet conditions, pore water pressure was applied up to 50MPa under the confining pressure of 200MPa. Sample was put between upper and lower sawcut alumina cylinders. The values for a-b of quartz and feldspar were positive under the dry conditions from room temperature to 600C. On the other hand, in the wet conditions, velocity weakening of quartz gouge is seen at around 300C, and negative region is from 200C to 400C. Velocity weakening of feldspar gouge is observed from 150C to 450C in the wet conditions. Velocity weakening of feldspar gouge is more apparent than that of quartz gouge. If this is true, feldspar plays more important role than quartz, in the velocity weakening region observed with granite gouge. These frictional properties are discussed with the texture of samples by the SEM observations.
T23A-0563 1340h
Periodic Viscous Shear Heating Instability in Fine-Grained Shear Zones: Possible Mechanism for Intermediate Depth Earthquakes and Slow Earthquakes?
Localized ductile shear zones with widths of cm to m are observed in exposures of Earth's shallow mantle (e.g., Kelemen & Dick JGR 95; Vissers et al. Tectonophys 95) and dredged from oceanic fracture zones (e.g., Jaroslow et al. Tectonophys 96). These are mylonitic (grain size 10 to 100 microns) and record mineral cooling temperatures from 1100 to 600 C. Pseudotachylites in a mantle shear zone show that shear heating temperatures can exceed the mantle solidus (e.g., Obata & Karato Tectonophys 95). Simple shear, recrystallization, and grain boundary sliding all decrease the spacing between pyroxenes, so olivine grain growth at lower stress is inhibited; thus, once formed, these shear zones do not "heal" on geological time scales. Reasoning that grain-size sensitive creep will be localized within these shear zones, rather than host rocks (grain size 1 to 10 mm), and inspired by the work of Whitehead & Gans (GJRAS 74), we thought these might undergo repeated shear heating instabilities. In this view, as elastic stress increases, the shear zone weakens via shear heating; rapid deformation of the weak shear zone releases most stored elastic stress; lower stress and strain rate coupled with diffusion of heat into host rocks leads to cooling and strengthening, after which the cycle repeats. We constructed a simple numerical model incorporating olivine flow laws for dislocation creep, diffusion creep, grain boundary sliding, and low T plasticity. We assumed that viscous deformation remains localized in shear zones, surrounded by host rocks undergoing elastic deformation. We fixed the velocity along one side of an elastic half space, and calculated stress due to elastic strain. This stress drives viscous deformation in a shear zone of specified width. Shear heating and thermal diffusion control temperature evolution in the shear zone and host rocks. A maximum of 1400 C (where substantial melting of peridotite would occur) is imposed. Grain size evolves during dislocation creep and grain boundary sliding as a function of stress and strain, and undergoes diffusive growth during diffusion creep. For strain rates ca E-13 per second and initial temperatures ca 600 to 850 C, this model produces periodic viscous shear heating events with periods of 100's of years. Strain rates during these events approach 1 per second as temperatures reach 1400 C, so future models will incorporate inertial terms in the stress. Cooling between events returns the shear zone almost to its initial temperature, but ultimately shear zone temperature between events exceeds 850 C resulting in stable viscous creep. Back of the envelope calculations based on model results support the view that viscous deformation in both shear zone and host will be mainly via grain-size sensitive creep, and thus deformation will remain localized in shear zones. Similarly, we infer that inertial terms will remain small. Future models will test and quantify these inferences. The simple model described above provides an attractive explanation for intermediate-depth earthquakes, especially those in subduction zones that occur in a narrow thermal window (e.g., Hacker et al JGR 2003). We think that a "smoother"periodic instability might be produced via the same mechanism in weaker materials, which could provide a viscous mechanism for some slow earthquakes. By AGU, we will construct a second, simple model using quartz rheology to investigate this. Finally, coupling of viscous shear heating instabilities in the shallow mantle with brittle stick-slip deformation in the weaker, overlying crust may influence earthquake frequency.
T23A-0564 1340h
Scaling Off-Fault Damage from Field to Laboratory
Rice, Sammis and Parsons \(BSSA, 2004\) derived analytical expressions for a dynamical stress field in the vicinity of a propagating slip pulse on a fault plane. They found that the magnitude of off-fault stresses was primarily determined by the velocity of the propagating slip pulse and the normal and shear stresses applied to the fault. Coulomb failure on favorably oriented cracks was possible to a distance on the order of 1 to 2 times a scaling distance Ro*, the slip weakening distance in the limit of low velocity and infinite pulse length. Using parameters measured by Heaton \(1990\), Ro* for earthquakes, and hence the widths of the damage zones, were calculated to be 1 to 80 m. We scaled Ro* to laboratory dimensions by writing it as a function of Dc and strength drop. An experimental normal load of 10 Mpa and a characteristic displacement of 10 microns, yields Ro* of 14 cm. We also calculate a fracture energy of 42.9 J/m2 and locked in slip displacement of 12.9 microns. These values compare favorably with data from Abercrombie and Rice \(2004\), Heaton \(1990\) and Okubo and Dieterich \(1984\).
T23A-0565 1340h
First seismic observation of a fluid pulse propagating along a fault
Several mechanisms have been put forward to explain the lubrication of slipping faults. Among these, the presence of high pore pressure inside fault zones sparks considerable interest in both the fields of exploration and earthquake seismology. For instance, high pore pressure along the San Andreas fault can explain the apparent lack of heat produced at the slip plane. In sedimentary basins, growth faults cutting through young, poorly consolidated rocks provide a means for overpressured hydrocarbons generated in deep source areas to migrate into shallower, economically producible reservoirs during times of microearthquake activity. Though these two examples cover different spatial scales, a collaborative effort between exploration and earthquake seismologists can help to gain a better understanding of fault dynamics. Using seismic reflection data from the prolific South Eugene Island Block 330 field, offshore Louisiana, we study the possibility of geologically fast, pressure-driven fluid flow along growth faults. From the data, we find what is, to our knowledge, the first unequivocal image derived from seismic reflectivity of a fluid pulse inside a fault zone that, based on geochemical evidence and drilling data, is ascending the fault with time. The fluid pulse is confined to a growth fault known as the B-fault, for which there are associated fault-plane reflections in the seismic data. Though the fault zone may be relatively thin at the location of the fluid pulse, it is detectable because the fluid pulse is of high fluid pressure and, hence, extremely low P-wave velocity. We extract the amplitude of the fault-plane reflection from the B-fault by applying a processing technique known as dip-filtering to migrated 3D seismic data gathered by Shell in 1992. The reflectivity at the location of the fluid pulse is greater than at an unremarkable part of the B-fault where a sonic log passed through the fault in 1993. As a further test of the hypothesis that the high amplitude anomaly results from an ascending fluid pulse, we compare the reflections from the B-fault in data sets from 1985 and 1992. Areas of high reflectivity systematically move up the fault plane 950 m (for an average speed of 130 m/yr). In contrast, the reflectivity of the nearly horizontal layer boundaries do not change as significantly in the two data sets. The pulse speed we observe can be explained with a model of a permeable fault zone connecting the shallow, normally pressured sediments to a deep, overpressured compartment.
http://www.mines.edu/~rsnieder/Publications.html
T23A-0566 1340h
Temperature Measurements and Active Faulting
Temperature measurements associated with active faults can be useful for studying the total energy balance, and especially the dynamic frictional levels during faulting of large earthquakes. However, temperature anomalies across faults or temperature changes associated with earthquakes are relatively rare. We make some simple calculations to estimate the temperature changes that should be observed across a fault for large earthquakes. For example, a temperature profile at 500 m depth across a fault that slipped 2 meters, at a time 6 months following the earthquake, shows a temperature anomaly of about 0.2 degrees, assuming an apparent coefficient of friction of 0.6. For an apparent coefficient of friction of 0.3, the anomaly reduces to about 0.05 degrees. The differences in the apparent coefficient of friction should be resolvable with the current temperature sensor instruments we are developing for borehole measurements. Estimating the apparent coefficient of friction is important for understanding the mechanics of faulting. The level of friction, and thus the amount of heat produced during an earthquake, has been a controversial issue in seismology for several decades. Timely measurements of the temperature profile across the fault following large earthquakes may be able to answer these long-standing questions about the level of dynamic friction.
T23A-0567 1340h
Zircon fission-track thermochronology of the Nojima Fault zone, Japan
Fission track (FT) thermochronologic analysis was performed on zircon separates from rocks in and around the Nojima fault, which was activated during the 1995 Kobe earthquake. Samples were collected from the University Group 500 m (UG-500) borehole, Geological Survey of Japan 750 m (GSJ-750) borehole, the fault trench at Hirabayashi, and nearby outcrops. Zircon FT data from the UG-500 borehole record ~2.5 Ma cooling in the zircon partial annealing zone (ZPAZ) for samples within ~3 m from the fault plane, whereas those of the GSJ-750 borehole record ~35 Ma cooling within ~25 m from the fault. On the basis of one-dimensional heat conduction modeling as well as the consistency between the degree of FT annealing and the degree of deformation/alteration of borehole rocks, these cooling ages in both boreholes are interpreted as consequences of ancient thermal overprints by heat transfer or dispersion via fluids in the fault zone. For the fault trench samples, zircon FTs of the 2 - 10 mm thick pseudotachylyte layer were totally reset (or remained reset) and subsequently cooled at ~56 Ma, which is interpreted as the time of final cooling through ZPAZ immediately after the pseudotachylyte formation. It is, therefore, suggested that the present Nojima fault system was reactivated in the Middle Quaternary from an ancient fault initiated at ~56 Ma at mid-crustal depths.
T23A-0568 1340h
Constraints on heat generation along two (paleo)seismogenic faults in California from U-Th/He and fission-track data
Although the stress / heat-flow paradox of the San Andreas fault has been debated for several decades, there is yet no widely accepted resolution to the problem. New results from the SAFOD drilling project on the San Andreas and similar drilling projects of other active faults around the world will help resolve this paradox. Similarly, our study and other detailed structural and thermochronologic studies of exhumed faults are providing important constraints on the evolution, thermal history, and strength of (paleo)seismogenic faults. We have obtained U-Th/He data for zircons from one inactive strike-slip fault in the San Andreas system (the San Gabriel) that has been exhumed approximately 2 to 5 kilometers. This data complements our current fission-track data set of 41 apatite and 8 zircon samples. The San Gabriel fault samples were collected in traverses near Whitaker Peak, the "Earthquake locality", and along Bear Creek. Samples were also collected from one traverse across the associated Punchbowl fault. Six zircon U-Th/He dates from the Bear Creek locality range from 18 to 23 Ma on the north side of the fault and from 36 to 51 Ma on south side of fault. Zircon FT dates from the same locality range from 39 to 46 Ma on the north side and 42 to 61 Ma on south side of the fault. Apatite FT dates from the same locality range from 9 to 22 Ma and 18 to 44 Ma (excluding one sample) from the north and south side of the fault, respectively. These data are consistent with an approximately 20 Ma offset in U-Th/He and FT dates across the fault and "cooling" rates between 3 and 6 oC / Ma. Apatite FT dates are between 51 and 25 Ma from Whitaker Peak, 54 and 24 Ma from the Earthquake locality, and 15 and 7 Ma at the Punchbowl fault. U-Th/He and FT dates vary systematically with distance from the fault cores along individual traverses; though, they do not young monotonically towards the faults, which would be expected if there had been sustained elevated temperature along the faults. Until we obtain more U-Th/He data, we are unable to definitively determine the factors controlling the variations in dates, though there is a moderately good correlation between dates and concentration of mobile elements in the fault rocks. Ultimately, this data set will help constrain the heat production and strength of these faults when they were active.
T23A-0569 1340h
The temperature during the high speed friction experiments estimated by ESR signals in quartz
ESR (electron spin resonance) dating method has been developed to obtain ages of quaternary geological events using calcite, aragonite, hydroxyapatite, and quartz. In natural quartz, paramagnetic (ESR sensitive) defects such as Al center (an Al atom replacing a Si, trapping an electronic hole) and Ti-Li center (a Ti atom replacing a Si, trapping an electron together with Li$^{+}$ as a charge compensator) are stable for the geological time scale while they decay on heating according to the thermal activation processes. In the present study, we use these ESR signals as indicators of temperature during the high speed friction experiments. The present experiment will also tell the conditions of faulting which completely zero the ESR signals, which is necessary for ESR dating of faulting to work. The gamma ray irradiated quartz grains of 0.5 to 1 mm were sandwitched by two gabrro columns of 25 mm in diameter with a tephron sleeve. The friction expereiments were performed with a load of 30 kg and with speeds of 75 to 300 rotation per minutes. After removing the columns, the crushed quartz powder was divided into three parts, outer, intermediate, and inner parts. ESR measurements were performed by an ESR spectrometer, JEOL PX-2300, at 83-87K, with a microwave power of 5 mW, and with an modulation amplitude of 0.1 mT. The estimated temperatures are higher for outer part where the frictional speed is higher while lower for inside, with systematic difference for different ESR signals, when assuming that the temperature was constant during friction experiments. The temperatures will be estimated again, in the presentation, with considering the temperature change with time while friction experiments, also with taking into account the results of heating experiments.
T23A-0570 1340h
A numerical simulation of the interaction between seismic slip and frictional melting
Frictional heat generated in the faulting process causes melting of the fault rocks. This may also drastically change the mechanical property of the fault. It is often suggested that the melt plays a role of lubricant in such case. However, after the result of Tsutsumi and Shimamoto (1997), some researchers point out the possibility that the melt behaves like a brake to decelerate the slip due to the high viscosity at the beginning of melting (e.g. Fialko, 2004). We simulate the interaction between seismic slip and viscous friction of melting layer in 2-D elastic medium, numerically. We aim to simulate an earthquake from the beginning of rupture. Thus, to treat the frictional behavior as a constitutive law of friction on the fault surface, we introduce following three friction regimes. At the initial stage of rupture, frictional stress obeys a slip weakening friction law (slip weakening friction regime). We assume the boundary between moving fault surfaces is filled by fault gouge and the frictional work heats the gouge layer. Then friction regime changes into a transitional stage from slip weakening to viscous friction (transitional friction regime). In this stage, the fault surface is not fully covered by viscous material though the melting started partially in the gouge layer. Finally, when the melt fraction becomes large, the viscosity of the viscous layer (hereafter, melt layer) controls friction on the fault (viscous friction regime). For the slip weakening friction regime, we assume a slip-dependent friction law with a critical slip distance. We assume heat is generated homogeneously in the fault gouge layer, and calculate the thermal field around the fault zone. For the viscous friction regime, we estimate the frictional stress from the thickness of the melt layer, slip velocity, and melt viscosity, assuming laminar flow in the melt layer. We calculate the evolution of thickness of the melt layer and the thermal field solving Stefan problem. In this regime we assume melt viscosity is temperature dependent (Vogel equation type). For the transitional friction regime, we gradually change the frictional behavior between slip weakening and viscous friction regime as linear function of averaged temperature in the gouge layer. We simulate rupture process of an earthquake in 2-D elastic medium regarding the above regimes as a constitutive relationship of friction. In this study, 2-D anti-plane rupture is simulated by finite difference method, and rupture starts when the shear stress exceeds the peak strength. In our typical result of simulation, after the shear stress decreases at the slip weakening regime, the frictional stress increases at the transient friction regime in the case of high slip velocity. This corresponds to viscous braking stage. If the elastic medium continues to supply energy into the gouge layer, the frictional regime changes to a viscous friction, then the melt layer is heated to high temperature, and large stress drop occurs. This is a viscous lubrication stage. For example, when the gouge thickness is 2 mm, the initial stress level is 25 MPa, and the frictional stress level drops to 20 MPa at slip weakening regime, frictional stress level increases to 23 MPa at viscous braking stage. Then the seismic slip continues and finally the frictional stress level drops to 5 MPa. On the other hand, melt lubrication doesn't occur in the case of low frictional stress level. Available energy supplied from the surrounding elastic medium determines whether the subsequent viscous lubrication arises or not. This is characteristic behavior in 2- (or 3-) dimensional elastic medium.
T23A-0571 1340h
Dynamic Traction Evolution and Fracture Energy On Extended Faults Inferred From Kinematic Slip Models
We estimate fracture energy density $(G)$ for moderate-to-large earthquakes by retrieving dynamic traction evolution at each point on the fault plane from slip history imaged by inverting ground motion waveforms. We use rupture models of the 1979 Imperial Valley, the 1992 Landers, the 1994 Northridge, 2000 Tottori, 1984 Morgan Hill, 1997 Colfiorito and the 1995 Kobe events. Our numerical approach uses slip velocity as a boundary condition on the fault in an elastic medium. We employ a 3-D finite difference algorithm to compute the dynamic traction evolution in the time domain during the earthquake rupture. We estimate fracture energy by calculating the scalar product between dynamic traction and slip velocity vectors. This approach does not require specifying a constitutive law and does not require the dynamic traction to be collinear with slip velocity. If these vectors are not collinear, the inferred fracture energy depends on the initial traction level. Fracture energy is taken to be the excess of work over the frictional work, where the kinetic friction (heat level) is assumed to be the minimum value of traction reached during slip. While the traction versus slip curves might be significantly degraded by poor resolution, other studies have shown that the inferred fracture energy might be more reliable. Our calculations reveal that the spatial distribution of $G$ is correlated with the total slip, the peak slip velocity and the rupture velocity $V_{r}$. For each heterogeneous slip model we compute the average of fracture energy density on the whole fault plane as well as for fault patches having slip exceeding various fractions of the maximum slip. Our estimates of average fracture energy density range between 1.4e6 and 2.e7 $J/m^2$ for earthquakes having moment magnitudes between $6.0$ and $7.3$, and are in agreement with the values proposed in the literature. We find that the total fracture energy $E_{g}(J)$ scales with seismic moment $Mo$ and, as a consequence, a similar scaling exists between average $G(J/m^2)$ and the average slip. The observed dependence of $G$ on $V_{r}$ is more complex than that expected from ruptures in 2D, and the dependence might be biased by slip heterogeneity.
T23A-0572 1340h
Thermal pressurization in 3-D dynamic spontaneous rupture models with cohesive zone
We simulate an earthquake rupture through a 3-D finite difference algorithm using the traction-at-split-nodes fault boundary condition. The dynamic rupture propagation is governed by an assigned constitutive law, which controls the breakdown processes within the cohesive zone. Our numerical procedure allows the use either of time- and slip-weakening or rate- and state-dependent (R $&$ S) friction laws. Seismic slip on faults produces temperature perturbations. Fault heating is controlled by the mechanical properties of the fault surface and by the rheological properties of the gouge layer. We model the temperature evolution on the fault through the heat flow equation and we couple these thermal variations with the fluid pressure changes by using the Darcy's law for fluid flow in porous media and the continuity equation of fluid mass in a solid. We assume that the increase of temperature does not change the adopted R $&$ S constitutive parameters during the dynamic instability. In a first set of simulations, we consider a constant porosity within the slip zone and we model the temporal variations of effective normal stress by considering the Terzaghi law. Subsequently, we use the evolution equation for the state variable proposed by Linker and Dieterich (1992), which accounts for normal stress variations. In this way we model the state variable evolution as a function of the constitutive parameters and the effective normal stress changes. Finally, we link this constitutive model with the evolution law for porosity proposed by Segall and Rice (1995). The goal of this study is to investigate dynamic fault weakening caused by shear heating and thermal pressurization of pore fluids. We show how these phenomena may complicate the dynamic traction evolution and affect dynamic fault strength. Our simulations reveal that the effect of frictional heating and temperature increase strongly depend on the thickness of the slip zone. Thus, our 3-D simulations confirm that thermal pressurization is a viable mechanism to explain earthquake ruptures.
T23A-0573 1340h
Stick-Slip as a Mechanism for Earthquakes Revisited
38 years ago Brace and Byerlee (Science, 1966) proposed that shallow earthquakes represent stick-slip sliding along old or newly formed faults in the earth and that observed stress drops represent release of a small fraction of the stress supported by the rock at the earthquake focus. We have investigated the generality of this proposal by relating the results of stick-slip friction experiments on large granite samples to counterpart observations for earthquakes. The stick-slip experiments, entailing the measurements of fault slip, frictional stress, and loading stress as functions of time, can be analyzed to determine the maximum slip rate, the apparent stress, the dynamic stress drop and the static stress drop, which, in turn, can be related to the corresponding earthquake parameters using a stress adjustment factor. That is, the laboratory loading stresses must be multiplied so as to approximate stresses measured in the seismogenic crust. For example, a stick-slip experiment run at a normal stress of 2.76 MPa was adjusted for the state of stress measured at a depth of 6.8 km at the KTB site, Germany, the deepest available in situ measurement. Applying the resulting stress adjustment factor of 41 to laboratory measurements of static stress drop, dynamic stress drop, apparent stress, and peak slip rate yielded estimates of 12.7 MPa (compared to a total shear stress of 65 MPa), 10.2 MPa, 3.3 MPa, and 3.1 m/s, respectively. These stress-adjusted parameters, independent of earthquake size, are all typical of those observed or inferred for major earthquakes. In particular, maximum slip rates within the fault zones of earthquakes appear to be several m/s independent of magnitude or moment. To relate the laboratory slip to the maximum slip within the fault zone of an earthquake, differences in stiffness as well as the loading stresses must be taken into account; the stiffness adjustment results in maximum slip scaling according to the cube root of seismic moment. Applying both the stress and stiffness adjustments to a laboratory slip of 93 microns yielded a maximum slip of 11 m for the M7.9 Denali, Alaska earthquake, in good agreement with peak values of about 12 m from slip models developed for this event. Thus, the physics of stick-slip friction as observed in biaxial laboratory tests at modest loading stresses appears to govern the rupture processes of crustal earthquakes in general, even those with large magnitudes. If so, then there is no need to invoke rupture processes unique to large earthquakes, such as thermal pressurization.
T23A-0574 1340h
Rupture Process and Energy Budget of Some Reservoir-Induced Earthquakes
The study of earthquake rupture process and energy budget through the analysis of seismic waveforms can yield insights into the effect of such factors as hypocentral depth and the presence of fluids on rupture process. A long-standing issue is whether the rupture processes of induced earthquakes differ from those of tectonic earthquakes. For example, Abercrombie and Leary (1993) noted that hydro-fractures, mining and reservoir-induced earthquakes appear to have lower average stress drop than natural tectonic earthquakes. This difference might be a result of a different tectonic setting or the shallower hypocentral depths of induced earthquakes. Alternatively, simple source models and the assumption of constant rupture velocity might have resulted in underestimated stress drops. We analyze earthquakes induced by seasonal oscillations in the water level in the A\c{c}u Reservoir, NE Brazil for comparison with previous studies of tectonic and induced earthquakes. 286 earthquakes 0$\leq$M$\leq$2.2, were recorded at 200 sps, by 8 3-component digital seismographs. We use 3 different approaches to calculate the source parameters of the 6 largest earthquakes (M$\geq$1.8). First, we fit the individual spectra using an $\omega^{-2}$ source model to find corner frequency ($f_{c}$), frequency-independent $Q$, and long-period amplitude. Second, we use collocated small earthquakes as empirical Green's functions of the large one and calculate the spectral ratios. We fit the spectral ratios solving for the $f_{c}$ of the largest earthquake. Third, we use relative source time functions (i.e. pulse widths) to determine source radius ($r$), and the rupture velocity. Estimates of source duration and $f_{c}$ imply stress drops in the range of 10 to 100 MPa. These are similar to tectonic earthquakes suggesting that hypocentral depth and the presence of water did not strongly affect stress drop. The source time functions vary systematically with azimuth implying a rupture velocity of $\geq$0.6$\beta$, consistent with that of large tectonic earthquakes. To continue the analysis, seismically radiated energy can be estimated from the source spectra; this information bears on whether the ratio of energy to moment changes with moment. Such a change may indicate a difference in the physics of large and small earthquakes and has been seen in some studies, but remains controversial. The radiated energies, considered together with rupture velocities can provide scarce constraints on the efficiency of rupture. It is unusual to have source spectra, static stress drops, radiated energies, and rupture velocities all available for small earthquakes.
T23A-0575 1340h
Rupture Process of the 2003 Bam, Iran, Earthquake: Did Shallow Asperities on a Fresh Fault Cause Extreme Ground Motions?
The Bam, Iran, earthquake on December 26, 2003 caused heavy damage to the city of Bam including the historic heritage of Arg-e-Bam. This Mw6.5 earthquake rupture created fresh faults 5 km westward away from the Bam fault. The Bam strong-motion station recorded 992 gal in the UD component and two directivity pulses in the horizontal components with a dominant frequency of 1 Hz. We inferred the rupture process of the 2003 Bam earthquake from strong motion data observed by BHRC, together with teleseismic data to constrain global features of the source. Waveform inversions using teleseismic data (e.g. Yamanaka, 2003; Yagi, 2003) have suggested the existence of a shallow asperity. Nakamura et al. (2004) estimated aftershock distribution with vertical dipping that superimposed the fresh faults, not the Bam fault. They proposed fault planes consisting N-S alignment with northward branches beneath the city of Bam. Our preliminary analyses show that two directivity pulses are created by northward rupture near the hypocenter and north-eastward rupture beneath the city. Recent earthquakes occurred on immature faults with shallow asperities have also generated localized extreme-strong motions (e.g., 2003 Miyagi-ken Hokubu, Japan, with Mw6.1; 2000 Tottori, Japan, with Mw6.6). Larger fracture energy is expected for shallow asperities on immature faults than those on mature faults. For example, the 2000 Tottori earthquake has several times larger fracture energy than expected by the scaling between seismic moment and fracture energy. When considering the energy budget, are radiated energy from the immature faults enough to generate the extreme ground motions? Detailed source process inversions might be able to answer this question.