T21D-01 08:05h
An emerging field of high-velocity friction and its implication for dynamic fault motion during large earthquakes
In order to understand not only the mechanisms of earthquakes, but also the origin of diverse behavior of faults and plate boundaries, one must integrate (1) field studies on faults to understand deep intrafault processes, (2) laboratory work to reproduce those processes and determine mechanical and transport properties of fault zones, (3) theoretical and numerical studies analyzing fault motion, including earthquake generation processes, based on the constitutive properties determined by laboratory studies, and (4) seismological and geodetic studies revealing dynamic fault motion during earthquakes and diverse aseismic fault behavior. Ideally, such integrated studies should be carried out for a selected fault that produced an earthquake with good seismic/geodetic records so the prediction from (1) to (3) can be fully tested with (4), rather than selecting favorite data in the literature. Present session is organized to promote such integrated fault and earthquake studies. This presentation will focus on high-velocity frictional properties of faults for which frictional heating plays crucial roles, with special reference to dynamic fault motion during large earthquakes. Recent progress in high-velocity friction studies on (1) frictional melting, (2) thermal pressurization and (3) high-velocity weakening of fault gouge are rapidly filling the gap between field/laboratory studies on faults and seismological/geodetic studies on earthquakes. Permeability and concentration of shearing deformation within fault zones determines relative significance of those processes. Accumulation of data on transport properties of fault zones has made it possible to perform realistic calculation of thermal pressurization processes, with predicted Dc values in quantitative agreement with seismically determined values. I also show highlight data on frictional melting and argue that effect of frictional melting on dynamic fault property can be predicted by solving a Stefan problem with moving boundaries [Hirose and Shimamoto, 2004; Satomi and Shirono, 2003, 2004; Matsuzawa and Takeo, 2004]. Remaining task is to include incipient frictional melting, characterized by melt-patches formation, and melt loss into fractures in the host rocks in the analyses of frictional melting [Hirose and Shimamoto, this session]. High-velocity friction data on Nojima fault gouge data [Mizoguchi and Shimamoto, this session] and intermediate-velocity data on rock-on-rock friction by the Brown group have revealed that there are unknown slip-weakening mechanisms, besides those two well-studied mechanisms. Tribochemical effects on high-velocity friction, i.e., the effects of interfacial chemical changes promoted by frictional heating under fluid-rich environments, are very important area for future systematic studies. Despite these unexplored areas, seismic fault motion will be predicted not so long in the future based on the measured properties on a fault that caused an earthquake. Transition from ordinary friction to high-velocity friction, poorly explored at present, should control the initial phase of earthquake generation and perhaps is critical to understand the physical bases of earthquake prediction. This is probably the most important area for systematic studies in the near future.
T21D-02 INVITED 08:20h
Results Of A Pilot Study To Investigate The Feasibility Of Using New Experimental Techniques To Measure Sliding Resistance At Seismic Slip Rates
Determining the shear resistance on faults during earthquakes is a high priority concern for researchers involved with fault and rock mechanics. Knowledge of shear resistance and how it depends on slip velocity, slip distance, normal stress, etc. is required fundamental information for understanding earthquake source physics. Reliable laboratory data on slip resistance of geological materials at high slip speeds, realistic displacements, and elevated normal stresses do not currently exist in spite of progress in this direction. This is because it is technologically difficult to conduct the relevant experiments. More recently, at CWRU, as a part of a pilot study funded by SCEC, two relatively new experimental techniques have been employed to investigate high-speed friction in analog materials and rocks. These experimental techniques are (a) plate impact pressure-shear friction experiment, and (b) the modified torsional Kolsky bar friction experiment. The plate impact experiments were employed to study a variety of friction states with normal stress varying from 0.5 to 1 GPa and slip speeds ranging from 1 to 25 m/s. The torsional Kolsky bar experiments were employed to study interfacial friction at normal stress ranging from 20 to 150 GPa and slip velocities of up to 10 m/s. Using these techniques plate impact pressure-shear friction experiments were conducted on soda-lime glass and fine grained novaculite rock, while the modified torsional Kolsky bar was used to conduct experiments on borosilcate glass specimens. The choice of soda-lime and borosilicate glass were dictated by a number of previous studies which have shown that the frictional behavior of glass is almost identical to that of rock. These experiments have provided the time resolved history of interfacial tractions, i.e. the friction stress and the normal stress, and the interfacial slip velocity. For the glass-on-glass experiments the interface shows no-slip initially, followed by slip weakening, strengthening and then seizure. For the novaculite rock the initial no-slip and the final seizure conditions are absent. It is interesting to note the range of the measured coefficient of friction for glass, which varies from 0.5 to as high as 1.3. These high values of coefficient of friction are in sharp contrast to metals where the coefficient of friction has been observed to be between 0.1 and 0.25 under similar loading conditions. Moreover, interfacial temperatures as high as 1500 degrees centigrade are estimated, which are close to the melting point of glass.
T21D-03 INVITED 08:40h
Numerical simulation of frictional melting: dependence of shear stress on viscosity
Frictional melting might lead to stress drop during slippage of a fault. McKenzie and Brune (1972) investigated frictional melting as a one dimensional heat conduction problem. They concluded that if the driving stress are of the order of 1 kbar, melting could occur for fault slips as small as 1 mm. Formation of a melting layer is also observed in laboratory experiments (Tsutsumi and Shimamoto 1997, Spray 1995). Once a melting layer is formed, the thickness of the layer increases or decreases (melting or solidification, respectively) according to a heat generation rate in the layer, and shear stress is determined by the thickness and viscosity of the layer. On the other hand, the heat generation rate depends on the viscosity, which strongly depends on temperature. Therefore, we have to solve a heat conduction problem with moving boundary condition (Stefan problem). In this study, we numerically solved this problem and determined shear stress evolution after a melting layer formed. A melting layer is sandwiched between two blocks moving at a constant sliding velocity. Viscosity of a melting layer is given by $\eta(T)=\eta_0\exp(E/T)$, where $E$ is an activation energy and $\eta_0$ is a temperature independent constant. The main results are summarized as follows: 1) dependence of the shear stress on both $\eta_0$ and $E$ is small. The stress increases only by a factor of three even if $\eta_0$ is increased by five orders of magnitude. The shear stress decreases as $1/\sqrt{t}$, and an approximate formula of shear stress evolution is derived. 2) the thickness of a melting layer increases as $\sqrt{t}$ and weakly depends on the viscosity parameters. For parameters simulating gabbro, the thickness is 1 mm after 1 s slippage. Comparison with an experiment reveals importance of escaping of a melt layer. Possible effect due to fracturing is discussed. References McKenzie, D. and Brune, J. N., Melting on fault planes during large earthquakes. {\it Roy. Astron. Soc. Geophys. J.}, {\bf 29}, 65--78, 1972. Spray, J. G., Pseudotachylyte controversy: Fact or friction? {\it Geology}, {\bf 23}, 1119--1122, 1995. Tsutsumi, A., and Shimamoto, T., High-velocity frictional properties of gabbro, {\it Geophys. Res. Lett.}, {\bf 24}, 699--702, 1997.
T21D-04 09:00h
``Partial Melting'' Of Fault Zones: A Mechanism Of Seismic Slip Termination
Our stick-slip experiments demonstrated that frictional melting terminates fault slips, and our numerical simulations demand that there should be a mechanism of enforced cooling (Koizumi and Otsuki, in this session). We present this mechanism referring the previous pin-on-disc experiments under severe conditions where friction is thermally controlled (Montgomery 1976; Ettles, 1986). Friction coefficient $\mu$ and wear rate W show a spectrum depending on the product of slip velocity V and normal stress $\sigma$. 1) Flash-melting: small $\mu$ (ca. 0.3) and W at small VP. Blobs of scratched debris play new asperities. The asperity contacts are easily flash-melted, but the temperature cannot rise up further, because the most of the heat flows to the counter face and the melt materials are immediately removed from the contacts. $\mu$ is self-adjusted by the equilibrium between the heat generation and cooling rates and formulated as, $\hspace*{10mm}$$\mu$ = 1.88 (Tm/Ph) (k$\rho$c/V)$^{1/2}$ (n*/$\sigma$)$^{1/4}$ ------- (1) where Tm: melting temperature, Ph: penetration hardness, k: thermal conductivity, $\rho$c: heat capacity of unit volume, n*: number density of asperities, and $\sigma$: normal stress. Eq. (1) represents velocity weakening and normal stress weakening. 2) Partial melting: abnormally large $\mu$ (up to 1.4) and W at moderate VP. The majority of the apparent contact surface remains cool in the flash-melting regime, but the temperature increases albeit slowly. Once it increases beyond a critical temperature, the penetration hardness decreases significantly and W increase abruptly. Wear debris produced at a high rate can cool the small amount of the melt materials, resulting in very high frictional resistance. If the external force is not sufficiently large to overcome this mechanical barrier, fault slips will stop. The elapsed time Te to the partial melting regime is expressed as, $\hspace*{10mm}$Te = 40 k$\rho$c Tm$^{2}$ (V$\sigma$)$^{-2}$ --------- (2). Applying eq. (2) to our stick-slip experiments (Koizumi and Otsuki, this session), Te is calculated at 27 $\mu$s, close to the observation of 17 $\mu$s. Assuming V=1 m/s and $\sigma$=275 MPa for seismic slips, Te=7.5 ms (7.5 mm slip), but it will be much larger when the slip is diffused in the fault zone with a finite width. 3) Full melting region: small $\mu$ (ca. 0.4-0.1) and W at large V. When the barrier of the partial melting is overcome, fault slip will run away.
T21D-05 09:15h
High-velocity Frictional Behavior of Dunite, Biotite Gneiss, Phyllite and Coal Show Evidence for Melting and Thermal Degasing
We conducted high-velocity frictional experiments on dunite, biotite gneiss, phyllite gouge and coal gouge at Kyoto University using a rotary high-velocity frictional testing machine. The purpose was to examine the effect of frictional melting in various rock types and to explore the effect of thermal degassing using coal as an analogue for a volatile fault zone. Experiments were conducted dry at equivalent slip rates of 1 m/s (1200 rpm) at normal stresses of 0.6-16 MPa for distances up to 90 m. Solid cylinders (25 mm diameter) of dunite and biotite gneiss were sheared with aluminum-alloy jackets at high stress, whereas phyllite and coal gouges were sheared with Teflon sleeves at low stress. The metal jackets allow high stress experiments to be performed and are inferred to melt before rock melting occurs. Dunite sheared at 10-16 MPa shows a weakening-strengthening followed by second weakening on melting, similar to previous experiments on gabbro without a metal jacket. Dunite melting is confirmed by, as yet unidentified, dendritic microlites, and a rapid reduction of steady-state frictional strength to 0.15. Under similar conditions, biotite gneiss shows apparent melting, but undergoes continuous strengthening without reaching steady state. Bituminous coal gouge sheared at 0.6 MPa undergoes a highly reproducible rapid weakening from 0.75 to 0.2, with odorous white gas emissions, sometimes accompanied by liquid hydrocarbons. Shear stress decreases prior to gasification and rapidly oscillating sample shortening/elongation occurs during gas emission. A slowly sheared sample (15 rpm) did not show weakening or gas emission. This is the first experimental demonstration of weakening associated with devolatilization during rapid slip. Vitrinite reflectance measurements on sheared coal samples may provide constraints on the temperature during gasification. Phyllite gouge sheared under the same conditions shows a gradual weakening to a steady-state strength of about 0.2; the weakening mechanism is not yet understood. Video presentations with mechanical and petrological data will illustrate these different behaviors.
T21D-06 09:30h
The Role of Silica Content in Dynamic Fault Weakening Due to Gel Lubrication
Little is known about the frictional behavior of rocks at sliding speeds ($\sim$1 m/s) and slip distances (1-10 m) characteristic of earthquakes, despite the importance of knowing the magnitude of the shear stress during seismic slip for understanding dynamic stress drops and strong ground motions. Recent experiments have demonstrated that the friction coefficient $\mu$ for monominerallic quartz rocks decreases to values as low as 0.1 at slip rates up to 0.1 m/s over meters of slip ({\it Goldsby and Tullis}, 2002; {\it DiToro et al}., 2004). Amazingly, the trend of $\mu$ versus log velocity for quartz rocks extrapolates to a value of zero at 1 m/s ({\it DiToro et al}., 2004). The mechanism deemed responsible for the low friction of quartz rocks is the formation of hydrated silica ('silica gel') on the sliding surface, which acts as a lubricant, lowering the shear resistance. Here we investigate the high speed sliding behavior of other important crustal rocks and assess the role that silica content plays in 'silica gel' formation. Tests were conducted on gabbro ($\sim$50 wt.% SiO$_{2}$), Tanco albite (68.6 wt.% SiO$_{2}$) and Westerly granite (69.2 wt.% SiO$_{2}$) in a 1-atm rotary shear apparatus. In each test, an annulus 54 mm in outer diameter and 45 mm in inner diameter was rotated against a flat circular plate of the same rock, at a constant normal stress of 5 MPa. Tests were begun by sliding at 10 $\mu$m/s for several mm of slip, followed by sliding at a constant velocity in the range 1 mm/s to 0.2 m/s for a displacement of $\sim$4 m. The value of the friction coefficient for all three rocks is $>$0.7 at 10 $\mu$m/s, in agreement with previous studies. A large value of $\mu$$\sim$0.8 is observed for gabbro at all velocities, whereas $\mu$ decreases dramatically with increasing velocity for granite and Tanco albite above 1 mm/s. At a given velocity, friction for granite and Tanco albite decreases rapidly over the first $\sim$0.2 to 1 m of slip (similar in magnitude to the slip weakening distances inferred from seismological data) followed by 'steady state' shear resistance. Trends of 'steady state' $\mu$ versus log velocity for granite and Tanco albite extrapolate to values of $\sim$0.3 and $\sim$0.35, respectively, at a seismic slip rate of 1 m/s. The friction coefficients for gabbro, granite, feldspar and quartz rocks decrease systematically with increasing silica content (for a given velocity and cumulative slip), suggesting that silica content plays a primary role in gel formation. Our results suggest that silica gel lubrication may be an important dynamic fault weakening mechanism during many crustal earthquakes.
T21D-07 09:45h
Texture and Energetics of Gouge Powder from Earthquake Rupture Zones
Standard methods for the analysis of particle-size-distribution (PSD) in fine-grained gouge (microscopic, sieving and laser particle analyzer) have several pitfalls. For example, our analysis of 145 gouge samples from the San Andreas fault zone shows that mean grain size dropped by 26 percent during 0.5 hr of continuous measurement with a laser particle size analyzer. This time drift reflects disaggregation of extremely fine particles into primary grains, and thus questions remain open with regard to size, PSD, and surface area of primary particles in fault gouge. To determine these properties accurately, we developed a new procedure for extended PSD measurements (up to 190 hr in the laser particle size analyzer), validated by SEM observations of gouge grains. Pristine gouge is difficult to find as chemical alteration and lithification may alter the gouge texture in exhumed, inactive fault zones. The present analysis is conducted on gouge from two fault systems that partly remove these limitations: (1) the active segment of the San Andreas fault-zone in the Tejon Pass region, southern California, with about 160 km slip and uplifted from a 2-4 km depth; and (2) the rupture zone of a "new borne" fault formed during the 1997 M=3.7 earthquake in Hartebeestfontein gold mine, South Africa, with about 0.4 m slip. The present study includes PSD measurements of approximately 250 gouge samples from both faults; 155 samples were measured for 0.5 hr or more, with eight samples being measured for 45-190 hr. Both faults display strikingly similar gouge characteristics: grain size distribution is non-fractal, fine grains approach the nanometer scale, and gouge surface areas approach 80 $m^2/g$. These observations challenge common precepts that gouge texture is fractal and that gouge surface energy (evaluated here as 2<sup><small>a</small></sup>10 $MJ/m^2$) is a negligible contributor to the earthquake energy budget. We propose that the observed fine-grain gouge is not related to quasi-static cumulative slip, but rather formed by dynamic rock pulverization during the propagation of a single earthquake (Reches and Dewers, this meeting, session S03).