Fault strength and stability: Lessons learned from the San Andreas fault in central California

In order to provide insight into the processes and controls that dictate fault strength and slip behavior, detailed laboratory investigations on samples of fault material, both outcrop and borehole, are necessary. Both material types are available for the San Andreas Fault (SAF) in central California where a borehole crossed the actively creeping fault at a depth of 2.7 km. I will use results from this case study to make larger determinations about the controls of fault strength and behavior.;The San Andreas Fault Observatory at Depth (SAFOD) project is a fault zone drilling initiative that was undertaken to fully characterize the behavior of materials and determine in situ conditions of the San Andreas Fault in central California at hypocentral depths. Results from experiments designed to characterize material in the 3D volume surrounding the borehole indicate that crust surrounding the fault is strong. Samples of granodiorite, arkosic sandstone, and siltstone returned from the borehole, wall rock on either side of the fault, are frictionally strong (mu=0.56-0.66). These samples come from within ~2 km of the active fault strand. Additionally, the results from experiments on outcrop samples representative of the lithologies found surrounding the fault at depth also show frictionally strong behavior (mu=0.56-0.68). A sample of serpentinite, thought to abut the fault at depth, showed low friction, mu = 0.18-0.26, and velocity-strengthening friction, consistent with observations of fault creep. The implications of these experiments are that the crust surrounding the fault is strong, whereas material thought to be in (or involved) with the fault at depth is weak.;To further investigate fault behavior, I performed experiments on well-located samples of cuttings and core returned from the SAFOD borehole at a vertical depth of 2.7 km. Material was recovered surrounding and from within 3 active faults penetrated by the SAFOD borehole. Two strands are actively creeping and were cored recovering intact fault material for analysis and experimental work. Experiments performed on cuttings from the main creeping strand are the first to indicate that mineralogy, specifically the group of clays, smectite, is responsible for observed behavior. These experiments showed that material from the fault is weak and exhibits near zero rates of frictional restrengthening or healing. Frictional restrengthening, along with velocity-weakening friction behavior (unstable slip), are both required for repeated earthquake rupture along the same fault plane or patch. I also performed experiments on intact and powdered core from the two actively creeping fault strands. Experiments performed on intact wafers of fault core indicate that the creeping faults are weaker than previously observed, mu = 0.09. In addition, this weakness is extremely localized, friction increases to mu > 0.4 over distances less than 1m outside of each fault. Furthermore, our results suggest that the faults would likely be weak in the upper 5-8 km of the crust, consistent with long-term observations of fault strength. Moreover, results from rock surrounding the faults show positive rates of frictional healing and velocity-weakening friction behavior, consistent with observed repeating earthquakes on the downdip extension of one fault and also on a nearby fault strand. I characterized the behavior of the third intersected fault strand through experiments performed on cuttings. This fault strand exhibited no active creep, and is the location of two repeating earthquake clusters. Results from these experiments indicate material surrounding the fault is strong, mu > 0.4, and exhibits complex frictional stability behavior. Our results indicate that material near this fault is frictionally unstable over short displacements and frictionally stable over larger displacements.;The observation of near zero healing rates in SAFOD material prompted the study of how mineralogy controls healing behavior and how the observed healing behavior relates to observations of fault slip style. To do this, I performed experiments on 9 natural fault samples and 8 synthetic gouge samples. Experimental procedure ensured that the only variable in this suite of experiments was the mineralogical composition of the samples used. Experiments show that phyllosilicate-rich gouges exhibit zero, or near zero rates of frictional healing, indicating the presence of these materials would lead to low fault strength and stably sliding behavior. Quartzo-feldspathic dominated gouges show positive healing rates whereas calcite-rich gouges show the largest magnitudes of frictional healing. Results for healing behavior and velocity dependence indicate that the majority of fault samples studies would fail under multiple modes of fault slip. (Abstract shortened by UMI.)...