Laboratory analysis of restrenghtening on simulated faults

Abstract

Seismic data show that stress drop increases with the logarithm of earthquake recurrence interval (1-5 MPa per decade time). If stress drop is taken as a proxy for fault strength, then the data suggest that faults restrengthen during the quiescent period of the earthquake cycle. Seismic surveys across natural faults show that seismic velocities increase with elapsed time after an earthquake. These data also indicate that faults lithify and heal during the interseismic period. Hence, this thesis is centered on laboratory research designed to investigate fault restrengthening. Experiments have been conducted to study the factors that effect stick-slip behavior of bare rock surfaces, and healing of simulated fault gouge under a range of physico-chemical conditions. We have performed experiments to investigate the repetitive stick-slip behavior of initially bare granite surfaces (nominal contact area 25 cm2). The tests were conducted in a double-direct shear apparatus at room-temperature and humidity, and normal stress was held constant throughout. Shear was induced by controlling the velocity of the loading piston (0.5-300 µmis). Data from individual stick-slip cycles show that shear stress increases quasi-linearly without considerable displacement across the sliding surface (stick), and sample failure (slip) is accompanied by a rapid stress drop. The amount of stress drop ranges from 0.1 to 3.1 MPa (or 4-49% of the failure strength). Prior to each instability samples exhibit yielding, consistent with previous laboratory observations of premonitory slip. Quasi-periodic instabilities occur repeatedly and we study the effect of loading rate and normal load on stress drop and recurrence interval (the time since the last event). We measure recurrence interval from the time when reloading begins after a slip event to the time that the next instability occurs. At a given loading rate, our data show a positive correlation between stress drop and recurrence interval, indicating healing rates of -3.5 MPa per decade increase in recurrence time. However, the combined data from all velocities show a lower rate, suggesting an apparent healing rate of -0.8 MPa per decade increase in recurrence time. We observe a similar velocity-dependent correlation between the peak shear stress prior to failure ( a measure of the ultimate strength of the material) and recurrence interval. We find a consistent scaling between different loading rates when the stress drop data are compared to the load.point displacement prior to failure ...