DI51A-0282
Spontaneous thermal runaway as an earthquake mechanism at elevated pressure: insights from petrological and numerical studies
Convergent margins are characterized by strong seismic activity with earthquakes occurring at depths of up to 700km. Shallow earthquakes (<60km) are explainable by the brittle failure of rocks. At greater depths, the increased ambient pressure should inhibit brittle failure and two main hypotheses have been proposed to explain intermediate-depth (60-300km) seismicity: (1) dehydration embrittlement and (2) melt shear instabilities.During dehydration embrittlement elevated fluid pore pressures counteract the lithostatic pressure and thereby lower the effective pressure to values at which tectonic stresses can lead to seismic failure. The fundamental unknown in this scenario is the pore pressure. Ductile shear instabilities are a different failure mode of rocks. Rapid deformation rates lead to frictional heating and melt lubrication resulting in self-accelerating deformation at seismogenic strain rates. The fundamental unknowns in this failure mode are the conditions at which a perturbation of the system will self-amplify instead of decaying and thereby lead to extreme localization of deformation and frictional heating. Pseudotachylytes are the only certain geological evidence for paleo- earthquakes and more and more eclogite-facies pseudotachylytes localities have now been discovered. They were found in rocks of the deeply exhumed continental roots and in exhumed fragments of subducted slabs. Their existence indicate that frictional melting is possible and may even be required for earthquakes under pressure-temperature conditions reasonable for depths >60 km.We present the results of a joint petrological, analytical and numerical study of thermal runaway as an intermediate earthquake mechanism. Field evidence from the Krakenes Gabbro in Western Norway show coexisting narrow eclogite-facies shear zones and pseudotachylytes. Using an analytical and numerical model for thermal-runway (Braeck and Podladchikov, 2007), we explore under which conditions (e.g. differential stress, P--T, and rheology) these shear zones/pseudotachylytes may have formed and why they coexist along strike. Our modeling results demonstrate that thermal runaway mechanism is applicable for intermediate-depth and deep earthquakes in subduction zones. In fact, at depths greater than 60 km thermal-runaway occurs at lower differential stresses than brittle failure and involves realistic stress drops. Braeck, S. and Podladchikov, Y. Y., 2007. Spontaneous Thermal Runaway as an Ultimate Failure Mechanism of Materials. Physical Review Letters. DOI: 10.1103/PhysRevLett.98.095504.
DI51A-0283 INVITED
A periodic shear-heating mechanism for intermediate depth earthquakes in the mantle
We first summarize results from a recent paper (Kelemen & Hirth, Nature 07). Intermediate depth earthquakes at 50-300 km in subduction zones occur below the brittle-ductile transition, where high pressures render frictional failure unlikely. Their location approximately coincides with 600 to 800 C isotherms in thermal models, suggesting a thermally activated mechanism. Some earthquakes may occur by frictional failure when high pore pressure results from metamorphic dehydration. However, because some intermediate depth earthquakes occur ~ 30 to 50 km below the paleo-seafloor, the hydrous minerals required for the dehydration mechanism may not be present. We present an alternative mechanism, the onset of highly localized viscous creep in pre-existing, fine-grained shear zones. Our numerical model uses olivine flow laws for a fine-grained, viscous shear zone in a coarse-grained, elastic half space, with initial temperatures from 600-800 C and background strain rates of 1E-12 to 1E-15 per second. When shear heating becomes important, strain rate and temperature increase rapidly to > 1/s and 1400 C. Then, stress drops dramatically, followed by low strain rates and cooling. Continued far-field deformation produces a quasi-periodic series of instabilities. Instability at the lowest strain rates is only predicted when active shear zones are a few millimeters wide. Using data from natural samples, we are exploring the hypothesis that localization in periodically deformed ductile shear zones increases over time, because grain size reduction during high stress deformation is retained at low stress in mixed olivine-pyroxene layers due to second-phase pinning, but heals via relatively rapid grain growth in olivine-rich layers. This is consistent with the observation that, in oceanic and ophiolite peridotite mylonites, olivine grain size in millimeter- scale, mixed phase bands is much smaller than in dunite bands.Our published models do not explicitly include the effect of enthalpy of fusion on the heating rate, nor weakening due to melting, and instead use a fixed peak temperature. New models will include heat of fusion and melt weakening to avoid the arbitrary temperature cap. New models will also investigate this mechanism using rheology for compositions other than olivine.
DI51A-0284
Serpentine Rheology and Dehydration at High-Pressure, Implications for Intermediate-depth Seismicity
Serpentinites have a lower viscosity than other mantle and slab materials within subduction zones. Serpentine dehydration is believed to play a major role in intermediate-depth seismicity, and several mechanisms have been proposed such as dehydration embrittlement and shear heating. However, quantifying the influence of serpentine rheology and its dehydration on strain rates and stress distribution within subduction zones has remained beyond reach, because of the lack of experimental data on deformation of the high-pressure variety antigorite, at relevant P and T conditions. Antigorite deformation experiments were carried out both within its stability field and during dehydration, over a pressure temperature (P-T) range of 1 - 4 GPa and 200-600 /deg C, at strain rates between ~10-4 and 10-6 s-1, in a D-DIA apparatus at GSE-CARS (Advanced Photo Source). Strain rates and stresses were obtained respectively from in-situ monitoring the sample length with X-ray radiographs, and azimuthal dependence of d- spacings on diffraction patterns. The determined stress-strain curves within antigorite stability field were fitted to a power-law equation including both temperature and pressure dependence. At the lowest strain rate investigated and nominal T within the antigorite stability field, localization occurred accompanied by local dehydration and a moderate increase in strain rate. Whatever the reaction and the sign of the volume change, dehydration induced an increase in strain rate. The present results show that antigorite rheology is likely to govern stress building-up and relaxation at the slab surface during interseismic time. We will discuss the implications of the results from the dehydration experiments for the role of serpentinites in intermediate-depth seismicity within subduction zones.
DI51A-0285
Nanoscale Properties of Rocks and Subduction Zone Rheology: Inferences for the Mechanisms of Deep Earthquakes
Grain boundaries are the key for the understanding of mineral reaction kinetics. More generally, nanometer scale processes involved in breaking and establishing bonds at reaction sites determine how and at which rate bulk rock properties change in response to external tectonic forcing and possibly feed back into various geodynamic processes. A particular problem is the effects of grain-boundary energy on the kinetics of the olivine-spinel phase transformation in subducting slabs. Slab rheology is affected in many ways by this (metastable) mineral phase change. Sluggish kinetics due to metastable hindrance is likely to cause particular difficulties, because of possible strong non-linear feedback loops between strain-rate and change of creep properties during transformation. In order to get these nanoscale properties included into thermo-mechanical models, reliable kinetic data is required. The measurement of grain-boundary energies is, however, a rather difficult problem. Conventional methods of grain boundary surface tension measurement include (a) equilibrium angles at triple junction (b) rotating ball method (c) thermal groove method, and others (Gottstein & Shvindlerman, 1999). Here I suggest a new method that allows for the derivation of grain-boundary energies for an isochemical phase transformation based on experimental (in-situ) kinetic data in combination with a corresponding dynamic scaling law (Riedel and Karato, 1997). The application of this method to the olivine-spinel phase transformation in subducting slabs provides a solution to the extrapolation problem of measured kinetic data: Any kinetic phase boundary measured at the laboratory time scale can be "scaled" to the correct critical isotherm at subduction zones, under experimentelly "forbidden" conditions (Liou et al., 2000). Consequences for the metastability hypothesis that relates deep seismicity with olivine metastability are derived and discussed. References: Gottstein G, Shvindlerman LS (1999) Grain Boundary Migration in Metals, CRC Press, 385 pp., New York. Riedel MR, Karato S (1997) Grain-Size Evolution in Subducted Oceanic Lithosphere Associated with the Olivine- Spinel Transformation and Its Effects on Rheology. EPSL 148: 27-43. Liou JG, Hacker BR, Zhang RY (2000) Into the forbidden zone. Science 287, 1215-1216.
DI51A-0286 INVITED
Aseismic anomalies in the mantle transition zone: subduction versus continental collision
Deep earthquakes are always associated with zones of recent convergence where cold temperature is expected in the mantle. Since low temperatures will raise both P- and S-wave speeds ( VP and VS), aseismic anomalies of high V in the transition zone of the mantle (TZ) provide unique insights into the origin of deep seismicity: Low temperature is only one necessary factor in causing deep earthquakes. Along the Tonga subduction zone, where over 1/3 of all deep earthquakes occur, a collection of complementary evidence indicates that a petrologic anomaly must be present to counteract the effect of very low temperature associated with extremely fast subduction of cold lithosphere: High VP and VS are conspicuously absent in the source zone of outboard earthquakes that extends about 1000 km farther to the west of the Wadati-Benioff zone. Meanwhile, preferred alignment of highly anisotropic material such as metastable olivine must be present in the source zone to account for over 1% of polarization anisotropy in the TZ. Moreover, the occurrence of large earthquakes in sub-horizontal slabs where a regional pattern of strain is generally absent calls for a localized source of stress, such as stress concentration associated with transformation-induced faulting, in the nucleation of deep earthquakes. Over time, petrologic anomaly due to metastable olivine would gradually dissipate with rising temperature and only a pure thermal anomaly of modest amplitude is expected to remain. This is precisely the case as an aureole of isotropic anomaly, characterized by modest-raised high VP and VS, is observed just outside of outboard earthquakes beneath Tonga-Fiji. A similar case is observed in the TZ beneath the Northern Philippine Sea, apparently the western extension of the Izu-Bonin subduction zone. A contrasting example is the absence of high VS where a large-scale anomaly of high VP was recently recognized in the TZ beneath central Tibet. A likely cause of the discrepancy between anomalies in VP and VS is a minor amount of water in nominally anhydrous polymorphs of olivine. Prior to thickening by continent collision, the Tibetan lithospheric mantle was part of a mantle wedge which has been hydrated during past episodes of subduction. The aseismic nature of the Tibetan anomaly is consistent with the fact that a small amount of water is likely to eliminate any metastable olivine. Furthermore, convective removal of thickened sub-continental lithospheric mantle is potentially a new pathway for water to enter the TZ.
DI51A-0287
Relationship Between the Focal Depth Distribution of Earthquakes and Rheological Structure of Lithosphere in Tibetan Plateau
It has been shown that almost all earthquakes on the continents are confined within the crustal layer. However, earthquakes of the Tibet plateau occur not only in the upper part of the thickened crust, but also in the mantle portion of the lithosphere. The lower crust appears to be aseismic. It is still unclear regarding the distributions of the focal depth and mechanism of earthquakes in Tibetan plateau. To study the distributions of the focal depth of the earthquakes in Tibet, high-quality databases from International Seismological Centre (ISC), National Earthquake Information Centre (NEIC) and China Seismological Network (CSN) are used in the period from 1964 to 2006 including 7157 events. The statistical results show that earthquakes are not confined only in the upper part of the crust, but extends all the way down to the lower crust and about 400 km depth in the mantle. Our preliminary results also suggest that the peaks of earthquake frequency are concentrated at the depths of about 10 km, 15 km, 25 km, 30 km and 90~100 km, respectively. Based upon the geophysics structure, thermal structure and composition of the lithosphere, the rheological structure of the lithosphere of the Tibetan Plateau is calculated to elucidate the mechanism of earthquakes at different depths. Our results indicate that there are three significant low-viscosity layers located at the base of the middle crust, within the lower crust and at the bottom of the lithosphere, respectively. The magnitude of the viscosity varies from 1024 to 1021 Pa·s, 1022 to 1020 Pa·s and 1023 to 1019 Pa·s in these three low-viscosity layers, respectively. Combining with the results of the focal depth distribution of earthquakes, our results suggest that most earthquakes are in the upper part of the crust with high mechanical strengths. The occurrence of shallow-focused earthquakes (<25 km) is closely related to the brittle fracture and frictional sliding. The earthquakes concentrated at the depths of about 30 km may relate to the brittle-ductile transition zone. Earthquakes also occur within the lower crust and the bottom of the lithosphere with low viscosity. The occurrence of earthquakes in the lower crust may be caused by fluids, partial melts or dehydrated reactions. The mechanism of mantle earthquakes (>300 km) may relate to the mantle phase transformation.
DI51A-0288
Determination of Rupture Velocities of Deep-focus Earthquakes Using Combination of Teleseismic and Regional Data
Although rupture velocity is an important parameter in understanding the physics of the earthquake source, it is often difficult to determine. For shallow earthquakes, not only seismic waveform data but also geodetic or surface displacement data are frequently available, so that they can be constraints for the estimation of source parameters including rupture velocity. However these closely observed data are not available for deep-focus earthquakes and this fact makes it more difficult to accurately estimate earthquake source parameters. One common method to estimate rupture velocity of deep-focus earthquakes is to use relative locations between the hypocenter and subevents, using time differences between the initiation and later phases in waveforms depending on the direction from the source to stations. This method has many uncertainties especially when it is difficult to pick the arrivals from subevents in the waveforms and when the event has a simple source time function. Another method is with seismic waveform inversions, which also often used for shallow earthquakes. This, however, does not have strong constraints on the fault dimension and rupture velocity because of the trade- off between them. To estimate rupture velocities of deep-focus earthquakes with better resolution, we used a new method, which is combination of teleseismic waveform inversion and forward modeling of regional data. We first carried out teleseismic waveform inversions for rupture velocities of 0.5 \symbol{"7E} 6 km/s on both nodal planes obtained by Harvard CMT solutions, varying grid size for the change of rupture velocity. Next, forward modeling of regional data was performed with each slip distribution obtained by teleseismic inversions, using empirical Green function method. Then we estimated the rupture velocity when the synthetics calculated by forward modeling best explain the observed regional data. Using this method we were able to estimate the rupture velocities of three deep-focus earthquakes surrounding Japan. The estimated rupture velocities are about 1 \symbol{"7E} 2 km/s, which are equivalent to 20 \symbol{"7E} 40 % of shear wave velocity. These values of rupture velocities are quite slow, compared to typical values for shallow earthquakes, which are 70 \symbol{"7E} 80 % of shear wave velocity.
DI51A-0289
High-precision relative earthquake locations in the Tonga-Kermadec subduction zone and implications for the mechanisms of deep-focus earthquakes
High-precision determination of earthquake location is important to the understanding of the physical mechanisms of the earthquakes. Using repeating seismic events is a powerful tool for determining high- precision relative locations of the events and studying temporal property change of the Earth. In this study, we focus on studying the repeating events occurring in the Tonga-Kermadec subduction zone, the most seismically active region for producing deep-focus earthquakes. We assemble seismic data for more than 7600 earthquakes occurring in the Tonga-Kermadec subduction zone in the time span of 1990-2006, with mb greater than 4.6. 119 high-quality repeating events are found among 115000 potential event pairs. The depth distribution of the repeating events is bimodal, with 64 in the top 100 km, 50 at a depth below 500 km, and 5 between 100 km and 500 km. For the shallow repeating earthquakes (event depth less than 100 km), the minimum and maximum recurrence intervals are 3 months and 14.5 years, respectively. The closest separation of the repeating events is in the same location, but different in depth by 0.5 km. For the deep-focus repeating events (event depth greater than 500 km), the recurrence interval ranges from 7 days to 13 years. The closest separation of the events is 2.4 km in horizontal space and 1 km in depth. For the intermediate-depth repeating earthquakes, the minimum and maximum recurrence intervals are 1.3 years and 7 years, respectively. The closest separation of the repeating events is 7.8 km in horizontal space and 6 km in depth. Among all the discovered deep-focus repeating events, five have different earthquake radiation patterns. The best one has occurrence interval of 5.1 years, separated by 10 km in horizontal space and 16 km in depth. We will discuss in detail the geographic and depth distributions, repeating pattern, relative location of the repeating earthquakes, as well as implications for various proposed mechanisms for deep-focus earthquakes.
DI51A-0290
Precise hypocenter distribution and earthquake generating and stress in and around the upper-plane seismic belt in the subducting Pacific slab beneath NE Japan
1. Introduction We found an intraslab seismic belt (upper-plane seismic belt) in the upper plane of the double seismic zone within the Pacific slab, running interface at depths of 70-100km beneath the forearc area. The location of the deeper limits of this belt appears to correspond to one of the facies boundaries (from jadeite lawsonite blueschist to lawsonite amphibole eclogite) in the oceanic crust [Kita et al., 2006, GRL]. In this study, we precisely relocated intraslab earthquakes by using travel time differences calculated by the waveform cross-spectrum analysis to obtain more detailed distribution of the upper plane-seismic belt within the Pacific slab beneath NE Japan. We also discuss the stress field in the slab by examining focal mechanisms of the earthquakes. 2. Data and Method We relocated events at depths of 50E00 km for the period from March 2003 to November 2006 from the JMA earthquake catalog. We applied the double-difference hypocenter location method (DDLM) by Waldhauser and Ellsworth (2000) to the arrival time data of the events. We use relative earthquake arrival times determined both by the waveform cross-spectrum analysis and by the catalog-picking data. We also determine focal mechanisms using the P wave polarity. 3. Spatial distribution of relocated hypocenters In the upper portion of the slab crust, seismicity is very active and distributed relatively homogeneously at depths of about 70-100km parallel to the volcanic front, where the upper-plane seismic belt has been found. In the lower portion of slab crust and/or the uppermost portion of the slab mantle, seismicity is spatially very limited to some small areas (each size is about 20 x 20km) at depths around 65km. Two of them correspond to the aftershock area of the 2003 Miyagi (M7.1) intraslab earthquake and that of the 1987 Iwaizumi (M6.6) intraslab earthquake, respectively. Based on the dehydration embrittelment hypothesis, the difference of the spatial distribution of the seismicity in the slab should correspond to the difference of the spatial distribution of the hydrated minerals and their dehydration reactions. In the upper slab crust, the upper-plane seismic belt is found because the hydrated minerals could be distributed homogeneously and the dehydration reaction (from jadeite lawsonite blueschist to lawsonite amphibole eclogite [Hacker et al., 2003b]) occurs perhaps largely at depth of 70-100km. Our result also suggests that in the lower portion of the slab crust and/or the uppermost portion of the slab mantle, the hydrated minerals could be inhomogeneously distributed and the seismicity occurs at depths around 65km, where another dehydration reaction may exist. 4. Characteristics of the focal mechanisms We examined the stress distribution within the slab by using focal mechanisms of the upper plane, interplane and lower plane events. From the plate interface to about 20 km below it, downdip-compressional (DC) type events are dominant. Below 20km from the plate interface, downdip-tensional (DT) type events are dominant. Many of interplane events have DC type focal mechanisms because of their locations in the uppermost portions of the slab mantle. These results indicate that the stress neutral plane from the DC type to DT type could be located at depth of about 20km from the plate interface.
DI51A-0291 INVITED
Source Rupture Process of the 2005 Tarapaca Intermediate Depth Earthquake
We investigate the details of the rupture process of the large (Mw 7.7) intermediate-depth earthquake that occurred on 13 June 2005 in the Tarapaca region of the Northern Chile seismic gap, using different data sets and different methods. The high quality and variety of seismic and geodetic data available for this event provided an unprecedented opportunity to study its source in detail. This earthquake is a slab-pull event with down dip extensional source mechanism. The aftershock distribution, determined from a post-seismic temporary array, indicates a sub-horizontal fault plane lying between the upper and lower planes of the double seismic zone. This earthquake was also recorded by a permanent digital strong-motion network operated by the University of Chile. These records have absolute time and high dynamic range so that they contain direct information about the rupture process. We used a systematic, fully nonlinear inversion method based on the neighbourhood algorithm to invert for the kinematic slip distribution using the accelerometric data set. This low frequency inversion provides a relatively smooth image of the rupture history. The kinematic inversion shows that the earthquake occurred by the rupture of two asperities. Based on the kinematic inversion result, we propose dynamic rupture models in order to quantify the dynamic rupture process. We simulate the dynamic rupture process and the strong ground motion using a 3D finite-difference method. In our simulation, dynamic rupture grows under the simultaneous control of initial stress and rupture resistance by friction. We constrain dynamic rupture parameters of the Tarapaca earthquake by simple trial and error. Large intraplate earthquakes in subduction zone are quite common although very few have been studied in detail. These earthquakes occurred at depth where the mechanism by which they are triggered remains poorly understood. Consequently, the determination of source rupture for intermediate intraslab events is important to understand the rupture process of these earthquakes.
DI51A-0292
Deep Earthquake Mechanics Inferred From Fault-Plane Orientations in Central South America
To place constraints on the physical mechanisms of deep earthquakes, we analyze the rupture properties of >30 intraslab earthquakes with MW >5.7 in central South America (15°--25°S). For all earthquakes, we perform a directivity analysis to estimate the rupture vector and identify the fault plane. After comparing the results with synthetics, we can distinguish the fault plane of the focal mechanism for ~1/3 of these earthquakes. For the largest earthquakes, we also invert for the slip distribution on the fault plane. At intermediate depths, we test whether earthquakes result from dehydration embrittlement reactivating the steep, trenchward-dipping faults of the outer rise. After accounting for the angle of subduction, these faults would be approximately vertical. This prediction contrasts with the orientation of faults identified between 100--300 km depth, which are all subhorizontal and instead suggest the creation of a new system of faults. The exclusive occurrence of subhorizontal faults agrees with previous studies in the Tonga-Kermadec and Middle America subduction zones. The similarity in results between the three subduction zones despite large differences in temperature, subduction velocity, and subduction angle suggests that the earthquake-generating mechanism is controlled by pressure rather than tectonic parameters. Deeper than 300 km, earthquakes occur along both subhorizontal and subvertical fault planes.