Brittle faults of the upper crust play an important role in seismic hazards, fluid flow, and fluid storage. Characterizing the structure of faults furthers our understanding of the geometry, boundary conditions, and mechanical behavior of the components of faulted crust. Fault zones generally are internally zoned (Fig. 1) with a relatively narrow, localized slip zone (core) surrounded by a transitional zone of deformed rock (damaged zone), which is then surrounded by the undeformed host rock (Sibson, 1977; Flinn, 1977; Aydin, 1978; Flinn, 1979; Chester, 1983; Dengo, 1982; Mitra, 1984; Chester and Logan, 1986; Sibson, 1986; Scholz, 1987; Bruhn et al., 1990; Byerlee, 1990; Evans, 1990; Cowie and Scholz, 1992b; Chester et al., 1993; Scholz et al., 1993; Bruhn et al., 1994; Evans and Chester, 1995; Little, 1995; Caine et al., 1996; Chester and Chester, 1998a). Several studies have identified an inward intensification of microfractures, subsidiary faults, cataclasis, and alteration toward the main fault surface (Flinn, 1977; Chester, 1983; Chester and Logan, 1986, 1987; Sibson, 1986; Bruhn et al., 1990; Little, 1995; Caine et al., 1996), as well as a characteristic fault and fracture fabric with respect to the main fault and to the principal stress orientations (Friedman, 1969; Brock and Engelder, 1977; Chester and Logan, 1986, 1987; Chester et al., 1993; Anders and Wiltschko, 1994). Internal zones have been defined on the basis of several mesoscopic and microscopic scale fabrics, including microfracture fabric. Although microfractures occur throughout fault zones, they tend to define the broadest zone of fault-related deformation. Definition of fault zones on the basis of microfracture fabric may be most consistent with the geophysical definition of faults using seismic wave propagation or gravity techniques of Stierman (1984), Mooney and Ginzburg (1986), Wang et al. (1986), and Bruhn et al. (1994) (Chester et al. 1993). The origin of the internal structure of fault zones in general, and of the microfracture fabric in particular, are poorly understood.
     The main goals of this study are to characterize the microfracture fabric of the damaged zone of the Punchbowl fault and to use this information to test predictions of mechanical models of fault evolution. The fabrics will be analyzed in light of the stress states associated with fault growth and wear along fault surfaces (Scholz et al., 1993; Saucier et al., 1992; Chester and Fletcher, 1997; Chester and Chester, 1998b). I will infer relative timing of microfracture development with respect to the type of microfracture, folding of the host rock, and diagenesis, and reconstruct the evolution of microfracture formation for distinct phases of deformation.

     It is generally understood that microfractures form in an orientation perpendicular to the minimum compressive principal stress direction (Friedman, 1963; Scholz, 1968; Engelder, 1974; Dula, 1981; Lespinasse and Pecher, 1986; Kowallis et al., 1987; Laubach, 1989; Fig. 2 ). Thus, microfracture fabric may be used to study paleostress conditions at the time the microfractures formed. It is assumed that microfractures along large-displacement fault zones occur in preferred orientations and record the spatial and temporal variations in local stress states during fault initiation and growth. Thus, we can use the information about stress states to test the predictions of fault evolution models.
     Several models of fault zone structure and fault growth have been proposed. Most models of the internal structure of faults describe a narrow, localized slip zone, surrounded by a transitional zone of damaged rock, which is then surrounded by the undeformed host rock (Fig. 1; Sibson, 1977; Flinn, 1977; Aydin, 1978; Flinn ,1979; Chester, 1983; Dengo, 1982; Mitra, 1984; Chester and Logan, 1986; Sibson, 1986; Scholz, 1987; Bruhn et al., 1990; Byerlee, 1990; Evans, 1990; Cowie and Scholz, 1992a; Chester et al., 1993; Bruhn et al., 1994; Evans and Chester, 1995; Little, 1995; Caine et al., 1996; Chester and Chester, 1998a). The slip zone, also referred to as the fault core (Chester and Chester 1998a) , is usually composed of a mechanically weak, foliated cataclasite and ultracataclasite layer that is meters thick (Chester et al., 1993). Beyond the fault core is the damaged zone, also referred to as the process zone (Scholz et al., 1993; Little, 1995), the transition zone (Bruhn et al., 1994), or crush zone (Sibson, 1977), which consists of mesoscopically and microscopically fractured and subsidiary faulted rock and cataclasites (Flinn, 1977, 1979; Chester, 1983; Chester and Logan, 1986; Sibson, 1986; Chester et al., 1993; Bruhn et al., 1994; Caine et al., 1996). The undeformed host rock contains little or no brittle deformation features associated with the fault.
     Fault initiation and growth models have been developed primarily on the basis of experimental rock deformation and mechanics. The Andersonian model of faulting assumes simple, homogeneous stress states in the crust and Coulomb failure behavior. Faults are predicted to form at acute angles (~30 ^o) to the maximum principal compressive stress axis (s ₁ ) for most rock types (e.g., Anderson, 1942; Scholz, 1990). In this case, microfractures associated with fault development are expected to parallel the maximum principal compressive stress axis, display a single preferred orientation, and to have formed early in the faulting history.
     Fracture mechanics based fault growth models consider the inelastic deformation at a fault tip as it propagates through intact rock (e.g., Cowie and Scholz, 1992a, 1992b). In brittle faults, the inelastic deformation is represented, in part, by zones of intense microfracturing that are referred to as the frictional breakdown zone (Cowie and Scholz 1992a, 1992b) or process zone (Cox and Scholz 1988; Scholz et al., 1993; Reches and Lockner, 1994). Fractures in the process zone gradually coalesce to form a continuous slip surface. As the fault tip and associated process zone propagate further, the damage associated with the previously formed process zone remains unchanged. The deformation noted around slip surfaces thus is interpreted to reflect the process zone wake (Fig. 3a). This model for fault development predicts that the microfracture fabrics should have formed early, during fault growth. The predicted orientation of fractures depends on the mode of fault growth (Scholz et al., 1993). For mode II tips, fractures should occur at 20 ^o and 70^o to the fault on the compressional and tensional sides of the fault, respectively (Fig. 4b). For mode III tips, the fractures should be symmetric and at 45^o to the fault (Fig. 4c).
     Wear models focus on the progressive accumulation of damage resulting from displacement on an established surface. Fault surfaces are rough at all scales (Scholz and Aviles, 1986; Power and Tullis, 1991). Displacement along a rough surface produces geometric mismatch and local stress concentrations. With continued slip, stress cycling in the rock adjacent to the fault surfaces occurs as the result of the passing of geometric irregularities. This cycling leads to continued fracturing and wear. The assumption that increasingly larger irregularities are juxtaposed with increased displacement, lead the wear models to predict an outward extension of damage from the fault surface with displacement (e.g., Scholz, 1987). Thus, the wear models predict that the damage zone of a fault records the entire history of faulting, with the greatest deformation associated with the final stages of fault movement (Fig. 3b). Rock adjacent to large displacement faults will experience many repeated stress cycling events which may tend to homogenize the fabric in the damage zone.
     Analyses of stress states induced by slip along irregular frictionless faults show that the maximum compressive stress will locally vary from fault normal to fault parallel in orientation (Saucier et al., 1992; Chester and Fletcher, 1997). On the basis of mechanical analysis of slip along irregular frictional faults, fracture most likely will occur at angles to the fault that are slightly greater than the far-field maximum compressive stress (Chester and Chester, 1998b). Thus, the wear model for a frictional fault predicts that the fabric of the damage zone will be relatively homogeneous, with a diffuse but preferred orientation of fractures oriented at high angles to the fault (Fig. 4d).
     A particular case of wear-models of faults that may be appropriate for this study, is the case describing a large displacement fault that progressively weakens with displacement (e.g., Tjia, 1972; Little, 1995). Because the Punchbowl fault is an ancient fault of the San Andreas system, it may have been relatively weak, analogous to the modern San Andreas fault (e.g., Zoback et al., 1987; Chester et al., 1993). A model of progressive weakening would predict that the angle between the fault and the maximum compressive stress would increase with time as the fault weakens. For this case, late-stage microfractures oriented at high angles to the fault may be expected to overprint early-formed fractures that are oriented at lower angles to the fault surface (Fig. 4e). The characterization of damaged zone microfracture fabric in the well-exposed Punchbowl fault zone will test the predictions of these mechanical models of fault evolution.

     The Punchbowl fault is a segment of the Punchbowl-Nadeau system, located at the juncture of the San Gabriel Mountains and Mojave Desert (Noble 1954). The Punchbowl fault is oriented approximately N66W, 70SW, and merges with the San Andreas fault both to the northwest near Portal Ridge and to the southeast near Cajon Valley. Of the assumed 300 km of right-lateral strike-slip displacement within the San Andreas system, the Punchbowl fault accounts for roughly 44 km of slip since the late Miocene (Dibblee, 1968; Woodburne, 1975).
     The study area is located in the Devil's Punchbowl Los Angeles County Park, California (Fig. 5). At this location, the Punchbowl fault is a right-lateral oblique-reverse fault that juxtaposes the Punchbowl Formation and a zone of intensely fractured basement. The Punchbowl Formation has been described in detail by several authors (Noble, 1954; Woodburne and Golz, 1972; Woodburne, 1975; Barrows and Barrows, 1975; Chester, 1983; Dibblee, 1987). The type section is located in the Devil's Punchbowl County Park, and consists of Miocene-Lower Pliocene conglomerate, arkosic sandstone, and siltstone strata (Woodburne, 1975). The sediments were derived from granitic and gneissic basement rocks and marine strata from the San Francisquito Formation to the north and east (Pelka, 1971; Woodburne, 1975; Barrows and Barrows, 1975; Dibblee, 1987), and were deposited in alluvial fans of playas of a northwest-southeast trending fault-bounded trough (Larsen, 1959; Woodburne and Golz, 1972; Woodburne, 1975). The Punchbowl strata are folded into a large asymmetric syncline (and several smaller folds), with the steeper, slightly overturned limb terminating against the Punchbowl fault (Fig. 6; Chester and Logan, 1986, 1987).
     The depth of faulting, estimated from the stratigraphic thickness of the Punchbowl Formation, is 2-4 km, corresponding to temperatures and overburden pressures of 75-125^o and 22-45 MPa, respectively (Chester and Logan, 1986). The mineralogy of the Punchbowl Formation and ultracataclasite consists of quartz, feldspar, calcite, laumontite, clinoptilotie, illite, and smectite, which is consistent with the inferred pressure and temperature conditions. These conditions favor deformation by cataclasis with minor pressure solution (Chester and Logan, 1986).
     The Punchbowl fault displays an internally zoned structure with distinct rock fabrics (Chester, 1983; Chester and Logan, 1986; Chester and Chester, 1998a). The fault zone is composed of an ultracataclasite layer, a damaged zone, and at its boundaries, relatively undeformed host rock (Fig. 1). Displacement along the Punchbowl fault was localized to discrete shear surfaces within the ultracataclasite (e.g., Chester and Logan, 1986; Chester and Chester, 1998a).
     Mesoscopic and microscopic features within the damaged zone have been described to a limited extent by Chester (1983) and Chester and Logan (1986, 1987). Mesoscopic deformational features in the damaged zone of the Punchbowl Formation include subsidiary faults, disrupted bedding, and cataclasis. Subsidiary fault intensity measured along three traverses across the damaged zone records an increase in intensity toward the ultracataclasite layer (Chester, 1983; Fig. 7). Microscopic deformational features include microfractures. These features also increase in intensity toward the ultracataclasite layer. Microfractures record episodic deformation and cementation (Chester and Logan, 1986). The damaged zone of the deformed basement exhibits subsidiary faults, microfractures, and cataclasis. Similar to the deformation in the Punchbowl Formation, deformation intensity increases toward the ultracataclasite layer. Less extensive analysis was conducted for the deformed basement because pre-existing brittle deformation obscures the deformational features associated with movement on the Punchbowl fault (Chester, 1983).
     Subsidiary fault sets are mutually offsetting and record several episodes of deformation followed by cementation (Chester and Logan, 1986). Preferred orientations exist for the subsidiary faults, but these vary between locations along the fault (Chester and Logan, 1987). Overall, subsidiary faults form a quasi-conjugate set with bisector oriented nearly normal to the Punchbowl fault. This fabric implies that the maximum principal compressive stress axis was oriented at high angles to the Punchbowl fault, and thus the Punchbowl fault was weak similar to the modern San Andreas fault. Microfracture fabrics were characterized in three samples from one locality in the damaged zone, and display preferred orientations nearly parallel and nearly perpendicular to the fault (Chester, 1983).
     Because of the location, large amount of displacement, and excellent exposure of the Punchbowl Fault zone, it was chosen to collect the critical field data with which to test the fault zone structure and evolution models previously discussed.

     Oriented hand samples of the Punchbowl Formation sandstone were collected from several locations in the damaged zone of the Punchbowl fault along traverses oriented parallel and perpendicular to the fault. (Fig. 6; Table 1). Sample traverses were located to the northwest of previous sample sites (Chester, 1983) in order to check for spatial variations along fault strike. Five traverses follow stratigraphic bedding away from the fault, with sample locations as close as 2.5 meters from the fault. One of these traverses follows the south limb and across the axis of the Punchbowl Syncline, to a distance of 1042 meters from the fault. Rock type, bedding orientation, and description of subsidiary faults, fractures, and veins were taken at each sample location. Size, shape, and friability were noted for each hand sample collected. Three oriented and mutually perpendicular petrographic sections were prepared from each sample.

     In all cases, microfractures are classified as intragranular, transgranular or grain boundary fractures and as to whether they are open, healed, partially healed, or sealed. Intragranular fractures lie totally within a grain (Kranz, 1983; Kowallis et al., 1987). Transgranular fractures cut across two or more grains and are sharp-tipped with matching sides (Kowallis et al., 1987). In thin section, grain boundary microfractures separate neighboring grains along their boundaries. Open fractures are filled with epoxy, healed fractures occur as planar arrays of fluid inclusions that result from diffusion along fracture surfaces (Smith and Evans, 1984). Partially healed microfractures share characteristics of both open and healed microfractures. They may be sharp-tipped with matching sides, but along some portion of the fracture, consist of planar arrays of fluid inclusions characteristic of healed microfractures. Sealed microfractures are cemented by secondary minerals (Kranz, 1983; Kowallis et al., 1987).
     Microfracture orientations were measured for the intragranular microfractures and the portions of transgranular microfractures that lie within grains. Orientation measurements were grouped as open, healed, partially healed, or sealed microfractures.
     Microfracture orientations were measured in forty grains per thin section, 120 grains per sample. Subrounded quartz grains were chosen for microfracture measurements. Enlarged photocopy images of each thin section were used to ensure that grains were spatially distributed. The orientation of every microfracture in each grain was recorded. This method was compared to that of Friedman (1969), in which a single orientation was recorded for each set of microfractures within a grain. The effect of sampling only sets of microfractures was explored by extracting representative orientations of sets from the measurements taken for each grain. A comparison of each sampling method (Fig. A1) illustrates that the fabric is similar, and either method is an acceptable technique to define fabric.

     In order to ensure that a statistically valid data set of microfracture orientations is being measured in each sample, one must correct for sampling bias, if significant. I explored several possible measurement biases. First, the observation area, or field of view, is reduced when the universal stage is tilted to high angles. This reduction in observation area reduces the number of high angle microfractures that can be seen and measured by about one third. To account for the area reduction, an inclination correction was applied to the data (Appendix B). The inclination correction is an inverse cosine function that applies a larger weight to high inclination microfractures, and a smaller weight to low inclination microfractures.
     Second, a potential bias is the general difficulty in observing microfractures that have a high inclination. In order to see and measure these microfractures, the operator must tilt the microscope stage to a high inclination and search more diligently than when measuring at a low inclination (e.g., Anders and Wiltschko, 1994). Even with experience and extra effort, a correction for operator bias may be needed.
     An analysis of the frequency of microfractures measured as a function of inclination suggests that an inclination correction for operator bias and observation error is necessary for the data set herein (Fig. B1). I used an inverse cosine cubed function to correct for area and operator bias. The correction weights microfractures at the highest inclination (45^o) about three times that of microfractures at the lowest inclination (0^o). Comparison of a sample data set with and without this correction illustrates only a slight change in fabric (Fig. B2). This change is acceptable given the importance of correcting for not being able to sample high and low angle microfractures equally.
     Third, the number of microfractures measured per grain is variable. This variability may lead to fabric concentrations that reflect the fabric of a single, or of a few, heavily fractured and/or large grains. In order to account for microfracture orientation as a function of microfracture density per grain, each microfracture orientation is weighted as a fraction of the number of microfractures measured in a given grain. For instance, a microfracture orientation from a grain with only two microfractures is assigned a weighting factor of 0.5; a microfracture orientation from a grain with ten microfractures is assigned a weighting factor of 0.1 (Fig. B3). This grain-weighting correction limits a few large and/or heavily fractured grains from defining the overall sample fabric and ensures that every grain measured contributes equally to the total fabric.
     Fabric elements in certain orientations may be measured multiple times when making measurements from three mutually perpendicular thin sections. Areas of the stereonet corresponding to these orientations are referred to as overlap zones. Assuming that microfractures are observed over the full range of inclinations of 0^o to 45^o for each mutually perpendicular thin section, all zones in the stereonet are sampled at least twice (Fig. D1), supporting Friedman's (1969) statement that if measurements are made from three mutually perpendicular thin sections, then no blind zones of measurement should exist. However, four triangular regions are sampled three times; thus the sampling is not uniform. A comparison of the effects of non-uniform sampling and inclination bias is presented in Appendix D (Fig. D2). If no inclination correction is applied, the triple-sampled region is extremely underrepresented relative to other regions (Fig. D2a). The most homogeneous sampling is achieved when the cubic cosine inclination correction is applied to the data. After this correction, the triple-sampled areas are slightly oversampled and small regions surrounding these areas are slightly undersampled relative to the average sampling (Fig. D2b). The uneven sampling could generate artificial preferred orientations in the stereoplots of microfracture fabric. The smoothing function in the StereoNett program (Appendix C; Duyster, 1998) alleviates some of this problem. However, in some samples with relatively random fabrics, weak point maxima occur at each of the triple overlap zones. Accordingly, the effect of the non-uniform sampling was taken into consideration when point maxima were picked from each fabric plot by qualitatively judging the relative strengths of the point maxima and minima with respect to their position within or outside the triple overlap zones.
     In summary, two corrections have been applied to the microfracture orientation data set, an inclination correction to account for the decreasing field of view with universal stage tilt and for the difficulty in seeing and measuring microfractures that require high angles of universal stage tilt, and a grain-weighting scheme to minimize the effect of a few heavily fractured grains. For a few of the stereoplots, the selection of point maxima was adjusted qualitatively when weak point maxima occurred at the triple overlap zones. Orientations of normals to microfractures from each section are plotted on lower hemisphere, equal-area stereographic projections, and then rotated into a single, horizontal, plane aligned with geographic directions (Fig. E1). These data were analyzed using both Almendinger's (1995) Stereonet software and Duyster's (1998) StereoNett software. All contour and scatter plots presented herein are generated using the StereoNett program (Appendix C).

     The samples contain different types and percentages of pore cement and fracture sealing cement (fill). In order to investigate the timing relationships between fracturing and the various cementation events, several samples have been analyzed using the electron microprobe and a Cold Cathode Luminescence Model 8200 Mk II attached to a Nikon Optiphot Pol microscope. Conditions on the electron microprobe include a beam current of 9.98 nA and 15.03 kV gun potential. Conditions on the luminiscope were an average gun potential of 15 kV, ~0.01 Torr vacuum, and 300 mA beam current.
     One sample (P41C) contains laumontite cement and fracture fill with no evidence of other abundant cement or fill minerals. Timing relationships between open fractures and cement, open fractures and sealed fractures, and sets of sealed fractures are the focus of observation in this sample. Another sample (P52C) contains both laumontite and calcite cement and fracture fill. This sample is used to establish the timing relationships between the laumontite and calcite cements, laumontite and calcite fracture fill, and open microfractures and cement or fracture fill. One of the three mutually perpendicular thin sections from P52C contains mostly calcite cement with very little laumontite cement. This thin section, and those from several other samples (P1B, P44, P45), which contain almost all calcite cement and fracture fill, are analyzed for determining the number of episodes of calcite cementation relative to fracture history.

     Elemental x-ray detection maps for silicon, potassium, sodium, and calcium were made to identify quartz, potassium feldspar, plagioclase feldspar, laumontite, and calcite. Individual elemental maps were superposed using a color scheme to make identification and distribution of each mineral easier to distinguish. Electron back-scatter and gray-scale cathodoluminescence images using the microprobe were also taken to identify open (epoxy-filled) fractures and possible timing relations between laumontite and calcite cements and fracture fill. However, because of the very strong luminescence of calcite, the gray-scale cathodoluminescence images of calcite-rich samples are of very poor quality. Also, the intense electron beam focused on the calcite, necessary for imaging, breaks down the calcite in a very short time. For these reasons, samples with abundant calcite (P52C, P44, P45, and P1B) were further imaged on the cold luminiscope that has a color cathodoluminescence detector.
     Variation in luminescence intensity of calcite is studied using the luminiscope. Samples rich in calcite cement (P52C and P1B) were analyzed in order to describe timing relationships between fractures and episodes of calcite cementation and fracture fill. Additional analysis has been done on samples outside the damaged zone (P44 and P45) in order document the extent of calcite cementation and fracture fill, as well as different stages of calcite cementation and fracture fill (strong vs. weak luminescence) beyond the damaged zone.
     In order to test the usefulness of images obtained from the microprobe, the limits of resolution of mineral phases in all x-ray elemental maps have been explored. Small or thin mineral grains, or minerals that have different compositions but similar proportions of elements for the phases of interest (laumontite and calcite) are displayed as various colors, including those that represent laumontite and calcite. The occasional discrepancy in color representation of mineral phases of interest has been confirmed optically for larger grains. Small or thin mineral grains are too difficult to distinguish optically, so the presence and distribution of these mineral phases can not be checked against the microprobe image analysis with much confidence. Accordingly, small or thin mineral grains observed in x-ray elemental maps are used in analysis of timing relationships cautiously.

     The samples of Punchbowl Formation collected are all arkosic sandstones, consisting of quartz, potassium feldspar, and plagioclase feldspar (Table 2). Quartz grains display relict undulatory extinction and no evidence of silica overgrowths under normal cathodoluminescence viewing. Potassium feldspar grains are mostly unaltered, but commonly display twinning and cleavage-related microfractures. Plagioclase feldspar grains are rich in sodium and commonly altered, containing numerous mica inclusions. Grains range from very fine to coarse in size, and vary in shape from angular to subrounded with aspect ratios typically 1.2:1. Samples are poorly sorted.
     The sandstones are grain-supported, with laumontite and calcite cement. Laumontite is a calcium-rich zeolite, and is assumed to be the by-product of alteration of plagioclase feldspar. Laumontite commonly forms aggregates of prismatic crystals as intergranular pore cement and microfracture fill, and displays cleavage-related microfractures. Calcite cement and microfracture fill is observed in many Punchbowl Formation samples. Calcite cement occurs as void fill between grains. The calcite cement has varying angles of extinction, orientation of twins, and two levels of cathodoluminescent intensity. The distribution of calcite cement varies from patchy to very extensive in a given sample and among sample locations (Table 3). In addition to laumontite and calcite cement, clay minerals are observed in several Punchbowl Formation samples. The clays occur as intergranular pore cement as alteration of biotite and chlorite crystals. The distribution of clay cement varies from isolated occurrences to extensive among samples.

     Open, healed, partially healed, and sealed microfractures are observed in all samples. Most microfractures are intragranular or grain boundary fractures, with transgranular microfractures common only very close (within 5 meters) to the fault.
     Open, healed, and partially healed intragranular microfractures are the most common microfracture types observed optically, and account for 22.2%, 61.1%, and 14.4%, respectively, of the optically measured microfractures. Sealed microfractures are much less abundant (2.3%). Open and sealed grain boundary microfractures are difficult to distinguish optically, but are clearly apparent and abundant in the samples studied by microprobe and cathodoluminescence techniques.

     The microfracture density of each sample has been quantified following the methods described by Anders and Wiltschko (1994) to determine the intensity of microfracture occurrence relative to distance from the ultracataclasite layer. Microfracture density is calculated by counting the number of microfracture traces that intersect a line of length 1.5 times the average grain diameter of each sample. On each of the three mutually perpendicular thin sections, 30 evenly spaced locations were selected for counting. At each location, the microscope stage was rotated an arbitrary amount in order to randomize the counting line orientation. The number of microfracture intersections along the counting line at each location was summed and then divided by the number of counting locations. This number was then converted to a linear density in terms of microfracture intercepts per millimeter for each sample.
     The microfracture density was determined for 15 samples taken from distances of 0.025m to 1.04km from the ultracataclasite layer (Table 4 ; Fig. 8) . Microfracture density decreases with distance from the ultracataclasite layer from a high of 69 microfractures/mm at 0.025 meters to 25 microfractures/mm at 1.04 km. Generally, fine-grained samples have a lower microfracture density than the coarse-grained samples. This difference may arise because smaller grains are harder to fracture and/or because it is easier to see fractures in larger grains than in smaller grains. Cement type appears to have no effect on microfracture density.
     Beyond a distance of about 100 meters from the ultracataclasite layer the microfracture density becomes relatively constant with increasing distance. Assuming that the constant density represents the regional, background level of microfracturing in the host rock, then the damaged zone boundary, as defined by microfracture density, is approximately 100 meters from the ultracataclasite layer. The boundary between the damaged zone and host rock determined by microfracture density is at a greater distance from the ultracataclasite layer than the boundary designated by the relative intensity of subsidiary faulting (Fig. 7; Chester and Logan, 1986). On the basis of subsidiary fault density, the boundary is about 15-30 meters from the ultracataclasite layer. These data indicate that there is a relatively narrow zone surrounding the ultracataclasite layer that is characterized by more intense deformation, recorded by subsidiary faults and microfractures. This is consistent with models of fault zone structure (Chester et al., 1993; Little, 1995).

     Cross-cutting relationships between different types of microfractures and between microfractures and pore cement are observed in all of the samples analyzed. Intragranular open microfractures cut healed and sealed microfractures (Fig. 9a), and sealed microfractures cut healed microfractures. Also, adjacent or cross-cutting open and sealed microfractures are observed in which the open microfractures show no evidence of partial fill (Fig. 9b). Open transgranular and grain boundary microfractures cut intragranular healed and sealed microfractures (Fig. 10). Some grain boundary and transgranular microfractures are partially filled by laumontite (Fig. 11) or calcite (Fig. 12c). Some sealed grain boundary and sealed transgranular microfractures are bounded by open grain boundary and open transgranular microfractures (Fig. 12b,c). These relations confirm that, in general, open microfractures are the youngest and healed are the oldest.
     Cross-cutting relationships are also observed between microfractures and pore cement (Figs. 12-16). Open grain boundary and open transgranular microfractures cut both laumontite and calcite cement (Figs. 12a, 12b/c, 13 and 16). In some cases, intragranular sealed microfractures connect with the surrounding pore cement (Fig. 14).
     Evidence for timing relations between types of cement are observed in samples containing both laumontite and calcite. Evidence that laumontite predates calcite includes prismatic crystals of laumontite cement occurring directly adjacent to quartz and feldspar grains, with calcite cement filling the void space surrounding these grains (Figs. 13 and 16). In addition, some small fragments of laumontite are entirely surrounded by calcite cement (Figs. 13 and 16). Very close to the ultracataclasite layer, laumontite cement is heavily damaged by microfractures that are sealed with calcite cement (Fig. 15(a,b), 15(c,d).
     Laumontite cement is nonluminescent and therefore, multiple episodes of laumontite cementation can not be distinguished. In contrast, at least two episodes of calcite cementation are observed using color cathodoluminescence analysis. Samples within and beyond the damaged zone display both weakly and strongly luminescent calcite cement (Figs. 17b, 18b, and 19b,d). Generally, the weakly luminescent calcite cement resembles a matrix or void filling cement, while the strongly luminescent calcite cement fills both pores and transgranular and grain boundary microfractures (Fig. 17b). The strongly luminescent calcite cement also fills microfractures that cut through the weakly luminescent calcite cement (Fig. 19d). These relations indicate that the strongly luminescent calcite post-dates the weakly luminescent calcite. The relative extent of twinning in each type of calcite cement is variable (Fig. 19c). The fact that both types of calcite cement are twinned and are cut by open microfractures implies that all calcite cementation occurred during deformation and was not the result of near-surface weathering.
     The possibility of more than just two episodes of calcite cementation is suggested by calcite veining observed in a sample very close to the fault (Fig. 20). Chester and Logan (1987) report supporting evidence of several episodes of synfaulting and a post-faulting episode of calcite cementation in the ultracataclasite layer of the Punchbowl fault. The ability to determine the timing relationships among these episodes of cementation is limited due to the observation of only two distinct phases of calcite under the luminiscope.

     The microfracture fabric of the individual types of microfractures within individual samples varies from nearly random to non-random. Non-random with multiple preferred orientations are typical fabrics of samples taken from very close to the Punchbowl fault surface (0.025 to 2.4m), while non-random with single preferred orientations are typical fabrics of the rest of the samples in the damaged zone. Individual density plots of the normals to microfractures for each microfracture type at each sample location were generated to compare fabrics (Figs. 21-32; Table 5). In general, the microfracture fabric of the different types of microfractures is similar, though not exactly the same, within individual samples.
     The normals to the open and sealed microfractures in sample DP10B (0.025 meters from the ultracataclasite layer) are non-random (Fig. 21) and are characterized by relatively strong point maxima. The normals to open microfractures display two point maxima (Table 5); one forms a broad girdle. The normals to sealed microfractures form a single point maximum in the SW quadrant that is aligned with the non-girdle point maximum of the open microfractures. The normals to healed and partially healed microfractures are more random.
     The normals to open, healed, and partially healed microfractures in sample P41A (2.5 meters from the ultracataclasite layer) are non-random, with similar point maxima (Fig. 22). The normals to the open microfractures are rotated counterclockwise relative to those of the healed and partially healed microfractures. The point maximum defined by the normals to the healed microfractures is inclined relative to that of the open and partially healed microfractures. The orientations of the normals to the sealed microfractures are similar to those of the open microfractures.
     The fabrics of the open and healed microfractures are similar in sample P52A (3.0 meters from the ultracataclasite layer), where the normals form a concentration in the NW quadrant (Fig. 23). The fabric of the partially healed microfractures is slightly more random; the normals forming a weak point maximum that is rotated clockwise relative to that of the open and healed microfractures. No sealed microfractures were observed in this sample.
     The normals to the healed and partially healed microfractures of sample P41C (5.0 meters from the ultracataclasite layer) form a strong point maximum in the SE quadrant (Fig. 24). Open microfracture fabric is more diffuse, displaying a preferred orientation that is rotated clockwise relative to the healed and partially healed microfractures. Sealed microfractures display similar concentrations.
     The normals to the open, healed, and partially healed microfractures in samples in P52B (6.0 meters from the ultracataclasite layer) and P52C (10.0 meters from the ultracataclasite layer) form similar point maxima that are aligned in the SE quadrant (Figs. 25 and 26). Sealed microfractures display a slight preferred orientation in sample P52B (Fig. 25), and display more random orientations in sample P52C (Fig. 26).
     At P18 (12 meters from the ultracataclasite layer), the normals to the open, healed and partially healed microfractures display a point maximum in the SE quadrant (Fig. 27). The normals to open and healed microfractures also display additional point maxima in the W, N and NE quadrants. Sealed microfractures do not display any distinct preferred orientations.
     At P1B (36 meters from the ultracataclasite layer), the normals to the open, healed, and partially healed microfractures are non-random, but form many local point maxima (Fig. 28). Sealed microfractures do not display any distinct concentrations.
     The normals to the open and partially healed microfractures in sample P54 (127 meters from the ultracataclasite layer) form broad point maxima in the NW quadrant (Fig. 29). The normals to the healed microfractures form a weaker point maximum concentration that is rotated clockwise relative to the point maxima defined by the normals to the open and partially healed microfractures. Sealed microfractures are randomly oriented.
     The normals to the open microfractures in sample P7B (479 meters from the ultracataclasite layer) display a very strong point maximum concentration in the E quadrant (Fig. 30). The normals to the healed and partially healed microfractures are more dispersed in orientation, displaying concentrations in the SE and NE quadrants that form a diffuse girdle. No sealed microfractures were observed in this sample.
     The normals to the open, healed, and partially healed microfractures of sample P10 (897 meters from the ultracataclasite layer) are non-random, displaying many local point maxima (Fig. 31). Point maxima for the three microfracture types are not similar. No sealed microfractures were observed in this sample.
     At P31B (1042m), the normals to the open, healed, and partially healed microfractures are non-random, forming moderately strong point maxima (Fig. 32). The normals to the open microfractures define a broad girdle that contains two point maxima, one in the NE and one in the SW quadrants. The normals to the healed and partially healed microfractures also display two point maxima each. The point maxima for normals to the healed microfractures are in the NW and SW quadrants, and those for the partially healed microfractures are in the NE and NW quadrants.
     Overall, the following observations can be made: 1) the normals to most microfractures form one or two point maxima within each sample, 2) the point maxima for different types of microfractures within a particular sample often are similar, and 3) the point maxima occur at a variety of orientations.
     For comparison purposes, two sets of summary plots are constructed. One comparison shows the point maxima of normals to microfractures, as distinguished by microfracture type, for all samples (Fig. 33). The second comparison consists of contour plots of the normals to microfractures for all of the healed, partially healed, and open microfractures combined, distinguished by sample location (Figs. 34, 35(a-d), 35(e-h),and 36; Table 6). The latter comparison is valuable because the different types of microfractures for a particular sample location often are similar.
     When grouped by microfracture type, the point maxima representation of the normals to microfractures form very diffuse clusters (Fig. 33). Most point maxima are moderately inclined and occur in the NW and SE quadrants. An additional, more minor cluster of small inclination, occurs in the SW quadrant. Point maxima also occur outside these clusters, creating a diffuse distribution of point maxima. Many or all microfracture types are represented in each cluster. Overall, the data show no spatial correlation by microfracture type.
     When all microfractures are grouped by sample, versus type, the data illustrate the overall fabric of each sample location, and can be compared to the fabric of the four samples measured by Chester (1983). Overall, the microfracture fabric at 0.025 meters from the ultracataclasite layer is nearly random, and displays only a weak preferred orientation of steeply dipping microfractures striking NW (Fig. 34a). The three samples between 0.76 and 2.4 meters from the ultracataclasite layer have similar microfracture fabric and well-defined preferred orientations (Fig. 34b,c,d). Each sample displays a set of steeply inclined microfractures with NW strikes and a set of N-NE striking microfractures that dip approximately 15^o to 50^o to the W-NW. The steeply dipping microfractures are subparallel to the Punchbowl fault, and similar in orientation to that in the sample that is 0.025 meters from the ultracataclasite layer (DP10B). The microfracture fabric at 2.5 and 3.0 meters from the ultracataclasite layer is more diffuse and lacks the steeply inclined microfractures (Fig. 35a,b). The overall fabrics in samples from 5.0, 6.0, 10, 12, and 14.2 meters from the ultracataclasite layer (Fig. 35c-g) are similar and non-random and is characterized by moderately inclined microfractures that strike to the NE. The overall microfracture fabric of samples at distances greater than 36 meters from the ultracataclasite layer are weak, with a variety of preferred orientations (Figs. 35h and 36). Of all of these latter samples, the sample at 479 meters from the ultracataclasite layer displays the strongest preferred orientation that is similar to that of the samples at moderate distances from the ultracataclasite layer (Fig. 36b).

     It is critical to determine the relative timing of microfracture generation and folding in order to analyze the relation of microfracture fabric to movement on the Punchbowl fault because variable rotations associated with folding modify pre-existing fabrics.
     Although cross-cutting relationships indicate the relative timing between microfractures types, the timing relative to folding of the Punchbowl Formation must determined indirectly through fold tests. Several types of fold tests were performed in order to investigate the timing relationship. The fold tests involve grouping microfracture orientation data based on the position of the sample in the fold and with respect to the fault, the orientation of bedding at each sample locality, and the type of microfracture. Rotation of the fabric of samples that simulate the process of unfolding (Table 7) are used to compare the microfracture fabrics of domains on opposite sides of the fold axis, and of different domains relative to the Punchbowl fault. If microfracture fabrics in a given domain are more homogeneous before folding, then it would be assumed that microfracture formation occurred before folding, and subsequent folding of the strata would have rotated the microfractures by different amounts and in different directions depending on location relative to the fold axis.
     The four samples closest to the ultracataclasite layer were not used in the fold tests because bedding is disrupted very near the fault by cataclasis, and representative bedding orientations could not be determined. The bedding orientation and rotations for unfolding for all other samples are given in Table 7. The rotations for unfolding are based on the orientation of the Punchbowl fold axis orientation reported by Chester and Logan (1987), and on the orientation of bedding. Unfolding rotations involve first rotating the fold axis to horizontal, followed by rotation of bedding to horizontal about the strike line (Table 7).

     The first fold test combines data from all samples for open and healed microfractures ôas isö to represent post-folding fabric and after the unfolding rotations to represent pre-folding fabric (Fig. 37). The post-folding fabric of all healed microfractures displays a stronger preferred orientation than the pre-folding fabric which suggests that most of the healed microfractures formed after folding. However, the fabric of all open microfractures is nearly random for both the pre- and post-folding cases.
     The second test compares pre- and post-folding fabrics of combined open, healed, and partially healed microfractures grouped by position in the Punchbowl Syncline (two samples on the south limb, P54 and P7B, and two samples on the north limb, P10 and P31B). In general, the combined fabrics on the north and south limbs display stronger preferred orientations for the post-folding case than for the pre-folding case (Fig. 38). In addition, the post-folding fabric on the north and south limbs are more similar. These relations suggest that most microfracturing occurred after the folding event.
     The third fold test analyzes the microfracture fabric of the damaged zone. Post-folding microfracture fabric from each sample in the damaged zone for open and healed microfractures treated separately, and for combined open and healed microfractures are consistently non- random with preferred orientations at high angles to the Punchbowl fault (Fig. 39a-d, e). The pre- folding fabrics of the combined and healed microfractures are much more random, such that a preferred orientation cannot be established. The stronger preferred orientations evident in the post-folding fabrics suggest that the observed fabric was largely established after folding.
     The final fold test is based on the observation that the bedding orientations fall into four basic orientation domains: 1) striking NW with moderate dips to the SW (P10, P31B), 2) striking NE with steep dips to the NW (P41A, P41C, P11I), 3) striking ENE with steep dips to the NW (P18, P1B, P54, P7B), and 4) striking ENE with overturned dips to the SE (P52A, P52B, P52C). The point maxima for normals to microfractures of the combined data set (open, healed, and partially healed microfractures) for each sample are grouped according to bedding orientation and compared for the pre-folding and post-folding orientations. The post-folding point maxima cluster in the SE quadrant regardless of the bedding domain (Fig. 40a). In contrast, the same data sets in pre-folding orientations display weaker preferred orientations overall, and more clustering relative to bedding domain (Fig. 40b). The results from this and the other three fold tests suggest that the microfracture fabric observed in the Punchbowl Formation was established largely after the folding event that produced the Punchbowl Syncline.

     Microfracturing in the laumontite-cemented Punchbowl Formation sandstones occurred before, during, and after Punchbowl faulting and folding (Fig. 41). The cross-cutting relations among types of microfractures observed under the petrographic microscope, electron microprobe, and the luminiscope suggest that 1) open grain boundary, open transgranular, and open intragranular microfractures are the youngest, 2) healed intragranular microfractures are the oldest, and 3) sealed grain boundary and sealed transgranular microfractures are intermediate between the open and healed microfractures. Among intragranular open, sealed, and healed microfractures, healed microfractures are oldest and open microfractures are youngest. A comparison of pre- and post-folding microfracture fabrics further indicates that the microfracture preferred orientations developed subsequent to Punchbowl Formation folding (Fig. 41).
     Cross-cutting relations between microfractures and cement indicate that open transgranular and open grain boundary microfractures formed intermittently with laumontite and calcite cementation. Sealed microfractures are assumed to be contemporaneous with pore cementation because both microfracture fill and pore cement are composed of the same minerals. Cross-cutting relations also indicate that at least some healed microfractures predate cementation (Fig. 41).
     The coarse and sometimes euhedral crystals of laumontite adjacent to clastic grains, and fragmentation of laumontite cement within calcite cement suggests that the laumontite cement largely predates the calcite cement (Fig. 41). Differences in luminescence intensity of calcite provide evidence of at least two episodes of calcite cementation. The strongly luminescent calcite cement postdates the weakly luminescent calcite cement, based on the presence of strongly luminescent calcite-filled microfractures in the weakly luminescent calcite cement.
     The calcite cement, grain boundary microfractures, and transgranular microfractures were probably formed contemporaneously with Punchbowl faulting. First, both types of calcite cement are mechanically twinned, suggesting that all of the calcite has undergone deformation associated with movement along the Punchbowl fault, and thus is not due to post-faulting or surficial processes. Further, disrupted calcite cemented veins in the Punchbowl ultracataclasites documented by Chester and Logan (1987) suggest that calcite veining is contemporaneous with faulting.
     Grain boundary microfracture orientations were not recorded to any great extent during universal stage traverses because these features were extremely difficult to distinguish. Therefore their orientations are not reflected in the fabric plots. Whether the orientations of the grain boundary and transgranular microfractures correlate with observed microfracture fabric needs to be addressed. Several microprobe and cathodoluminescence images clearly illustrate the abundance, preferred orientations, and concentration of grain boundary and transgranular microfractures within the damaged zone. The higher density of grain boundary and transgranular microfractures in the damaged zone suggests that these types of microfractures are related to deformation along the Punchbowl fault. Some evidence of partial calcite fill within a single grain boundary also suggests that the formation of the partially filled transgranular and grain boundary microfractures pre-dates calcite cementation events, which are assumed to be related to movement along the Punchbowl fault. These observations suggest intermittent microfracture and calcite cementation, in which an early deformation event produced large transgranular and grain boundary microfractures, which were subsequently sealed with calcite. A later deformation event then produced similar large transgranular and grain boundary microfractures, sometimes bounding the older sealed microfractures, and that were not subsequently filled with calcite.
     Nonetheless, there is the distinct possibility that some open grain boundary and open transgranular microfractures formed after Punchbowl faulting and are related to unloading associated with exhumation and erosion (Fig. 41). This is suggested by some images that display open grain boundary microfractures that seem to encircle entire grains. Post-faulting episodes of calcite cementation may also have occurred.

     The microfracture fabrics of samples are analyzed as a function of distance from the ultracataclasite layer in order to compare the fabrics to the structurally defined zones of the Punchbowl fault (i.e., the "undeformed" host rock, damaged zone, and fault core; Figs. 42-45). In general, microfracture fabrics for ten of the eleven samples within 36 meters of the ultracataclasite layer display a strong preferred orientation of microfractures that are oriented at high angles to the fault (Figs. 34, 35(a-d,e-h) ,42, and 44; Table 8). In addition, the four samples closest to the ultracataclasite layer (within 2.4 meters of the ultracataclasite layer) display an additional preferred orientation of microfractures that are subparallel to the fault (Figs. 34, 42a, and 44; Table 8). Overall, samples at 36 to 1040 meters from the ultracataclasite layer display more random fabrics with dissimilar preferred orientations (Figs. 35g, 36, 43, and 44; Table 8).
     The structurally defined zonation is also evident when looking at the point maxima defined by the normals to microfractures from each sample location (Fig. 45). Point maxima from the fault core cluster at orientations approximately perpendicular and subparallel to the Punchbowl fault, while those from the damaged zone cluster only at orientations approximately perpendicular to the fault (Table 9). Point maxima from the "undeformed" host rock are not clustered in any distinct orientation. The differences in clustering of point maxima imply that a representative microfracture fabric exists for each structurally defined zone.
     Although the boundaries between the fault core and damaged zone, and between the damaged zone and "undeformed" host rock are gradational and can not be located precisely on the basis of microfracture density (e.g., Chester et al., 1993), the structural boundaries do correlate approximately with changes in microfracture fabrics. Fabrics within each of the three domains are fairly homogeneous. Combining microfracture orientations of samples within the three domains provides a representation of the overall fabric of each domain (Figs. 42, 43, and 44).
     On the basis of the microfracture density measurements, the sample P1B at 36 meters from the ultracataclasite layer (P1B) would be expected to display fabrics representative of the damaged zone. However, the nearly random fabric of this sample is more consistent with that of the "undeformed" host rock (Figs. 35(a-d, e-h) and 36). Closer-spaced samples will be needed to investigate how abruptly the microfracture fabric changes in the transition zone between the damaged zone and "undeformed" host rock.
     To first approximation, the combined fabric stereoplots imply the fabric of the damaged zone is characterized by a single set of microfractures with preferred orientations approximately perpendicular to the Punchbowl fault and approximately perpendicular to the inferred direction of slip on the fault (Figs. 42b,c,d, and 44). The fabric of the fault core is characterized by a similarly oriented microfracture set showing even stronger preferred orientation and another microfracture set with a preferred orientation subparallel to the Punchbowl fault ultracataclasite layer (Figs. 42a and 44). The ôundeformedö host rock is characterized by a nearly random fabric overall (Figs. 43 and 44).

     The microfracture fabrics of damaged zone samples are characterized with respect to orientation in order to test the predictions of fault growth and wear models of fault evolution. Each set of models predicts a zone of intense microfracturing close to the fault (e.g., Cowie and Scholz, 1992a, 1992b; Scholz et al., 1993), similar to that defined by the microfracture densities in the Punchbowl fault zone. Both the Andersonian model of fault formation and fault growth models predict a homogeneous microfracture fabric which records the early-stage deformation associated with formation of the fault surface. In these cases, the fabric should be characterized by a preferred orientation of microfractures at relatively low angles to the fault (Fig. 4a,b,c). For the case of Mode II fault growth, microfractures are expected to form either at 20^o or 70^o to the fault surface. For the case of Mode III fault growth, the microfractures are expected to form at 45^o to the fault surface. The predictions of these models of microfracturing associated with fault formation do not agree with the finding that microfracture fabrics in the Punchbowl fault zone display a preferred orientation at about 90^o to the fault surface.
     The wear models predict relatively homogeneous microfracture fabrics that record later-stage deformation associated with displacement on an existing fault surface. Wear fabrics should have preferred orientations that are consistent with late-stage stress states. Assuming that stress along faults is homogeneous, the observation for the Punchbowl fault, that microfractures have preferred orientations approximately normal to the fault, imply that the late stage maximum compressive stress direction was oriented nearly perpendicular to the fault Fig. 4d). Such an inference was made by Chester and Logan (1987) on the basis of subsidiary fault fabrics along the Punchbowl fault (Fig. 46). However, homogeneous stress with the maximum compressive stress direction nearly normal to the fault can not explain the second set of microfractures in the fault core, that are oriented nearly parallel to the fault.
     Mechanical modeling of irregular, frictional faults by Chester and Chester (1998b) indicates that stress is not homogeneous near a frictional fault due to the juxtaposition of geometric irregularities along the fault surface with displacement. For elastic properties and a fault geometry appropriate to crustal faulting, the normal stress across a fault will vary locally to near zero values even for the case of nearly fault-perpendicular loading in the far-field. Zero normal stress locally along the fault surface produces local separation of the fault surface and of microfractures oriented parallel to the fault. Such a process could explain the observation of fault-parallel microfractures in the fault core of the Punchbowl fault.
     Near the fault surface, stress directions and magnitudes should vary greatly, particularly for faults with low coefficients of friction (Chester and Chester, 1998b; Saucier et al., 1992; Chester and Fletcher, 1997). However, for the case of frictional faults, inelastic deformation is expected to occur only locally within the rock near the fault surface. Chester and Chester (1998b) show that within the failure regions, particularly for the case of faults with a low coefficient of friction, the orientation of the local maximum principal compressive stress is at a greater angle to the fault than that of the far-field state of stress. Under many loading conditions, the orientation of the maximum principal compressive stress in the failure region near the fault is nearly normal to the fault surface. The microfracture fabric in the damaged zone of the Punchbowl fault is compatible with this mechanical model, particularly for the case of a fault with a low coefficient of friction and loaded by a far-field maximum principal compressive stress that is oriented at a high angle to the fault.
     The observation of random microfracture orientations in the region beyond the damaged zone of the Punchbowl fault may result because local features (e.g., subsidiary faults or folds) dominate the fabric locally. However, observed and mapped local structures, e.g., fold hinges and subsidiary faults, do not appear to have a consistent relationship with microfracture fabric in samples collected from locations near these structures. Open and partially healed microfracture fabric at one location (P7B), which is close to the fold hinge and near large subsidiary faults, is non-random with a strong preferred orientation at 20^o from the hinge (Fig. 30a,c). In contrast, healed microfracture fabric at this location is almost random, with a very weak preferred orientation (Fig. 30b). At a location outside the hinge zone and further from subsidiary faults (P54), the combined open, healed, and partially healed microfracture fabric is close to random, with only a very weak preferred orientation (Fig. 36a). Within the hinge zone (P10), the combined open, healed, and partially healed microfracture fabric is weak to random (Fig. 36b), even though it is very close to a large subsidiary fault. The microfracture fabric very far from subsidiary faults, (P31B), is non-random, with moderately strong open microfracture preferred orientations at both high and intermediate angles to the Punchbowl fault (Fig. 36d).
     To test the possibility that microfractures record progressive weakening of the Punchbowl fault, the point maxima for normals to open and healed microfractures are plotted for each sample from the damaged zone. Fabrics from the fault core and from beyond the damaged zone are not used in this analysis because of the double point maxima and nearly random fabric, respectively. Given that healed microfractures are older than open microfractures, the difference in preferred orientation of the normals to the healed and open microfractures could record changes of the maximum compressive stress direction through time. Small (~5^o) to intermediate (~40^o) clockwise rotations from the healed to open microfractures, respectively, exist for almost all of the damaged zone samples (Fig. 47). However, at P41A, the open microfracture fabric is oriented about 45^o counterclockwise relative to the healed microfracture fabric (Fig. 47) . The counterclockwise rotation at this location may not be significant, however, given that the fabric at this location is weak relative to the other damaged zone samples (Fig. 22).
     An overall clockwise rotation of open microfractures relative to healed microfractures and the observed relative rotation of age of the two types of microfractures would record a clockwise rotation of compressive stress direction as deformation progressed. This clockwise rotation would record an increase in the angle between the Punchbowl fault and the maximum principal compressive stress direction and would be consistent with the effects of fault weakening on the stress state near a fault (e.g., Tjia, 1972). However, the directions of rotation of open relative to healed microfractures from this data set are variable. In addition, the overall fabrics of open and healed microfractures in the damaged zone do not indicate a significant rotation. Therefore, a rotation of the maximum principal compressive stress direction with time can not be demonstrated convincingly.
     The preferred orientations very close to the Punchbowl fault reflect local reoriented stress directions. The preferred orientations are consistent with the stress states predicted by models of irregular, frictional faults, which suggests that damage around large-displacement faults, such as the Punchbowl fault, is due largely to wear processes (progressive deformation) rather than the initial localization and growth of the fault surface (Chester and Chester, 1998b). The fact that the oldest microfractures (healed) and the youngest microfractures (open) have similar preferred orientations further implies that the average stress conditions along the fault were relatively constant in orientation.

     Measurement of microfracture orientations called for addressing the sampling technique of measuring all vs. sets of microfractures. A sampling technique of measuring all microfractures was used, from which data were organized into sets according to similar orientation. From each set, a representative orientation was chosen and plotted. This manipulation of the complete data set creates data sets similar to those collected by Friedman (1969). The complete and the abbreviated data set were plotted on a lower hemisphere, equal-angle projections to determine if the sampling technique affected the fabric of the sample (Fig. A1). Only very subtle changes are evident, indicating that either sampling technique gives a representative fabric.

     Several possible measurement biases were explored. First, there is the reduction in number of high angle microfractures that can be seen and measured when the universal stage is tilted to high angles. The reduction is due to the reduction in observation area. A related measurement bias is the general difficulty in observing microfractures that have a high inclination. In order to see and measure these microfractures, the operator must tilt the microscope stage to a high inclination and search. Even with experience and extra effort, a correction for this operator bias may be needed.
     The frequency of all microfractures measured in all samples at intervals of 1^o inclination were plotted to compare the sampling of high vs. low angle microfracture inclinations (Fig. B1a). Combining measurements from many sections for many samples should produce a random data set with uniform distribution of low and high inclination measurements. However, I found a significant difference between the frequency of low and high angle microfractures. Accordingly, a correction bias was applied which applies larger weight to high angle microfractures, and smaller weight to low angle microfractures. This correction is an inverse cosine cubed function, weighting microfractures with inclinations of 45^o about three times that of microfractures with inclinations of 0^o. Intermediate weights are applied to intermediate inclination values. The correction evens out the distribution of low and high angle microfractures (Fig. B1b), giving a more representative sampling of microfracture orientations. The effect of the correction on the microfracture fabric is minimal (Fig. B2), so even if the assumptions of the correction are not entirely accurate, they are relatively unimportant.
     A second weight correction was applied to the data, which ensures that fabric concentrations do not reflect the fabric of a single, or a few, heavily fractured and/or large grains. Each microfracture orientation is weighted as a fraction of the number of microfractures measured in a given grain. The effect of this correction on a sample data set (Fig. B3) illustrates the subtle changes in fabric. Although the change in fabric is subtle, the correction is still applied to all data in order to limit the effect of a few large and/or heavily fractured grains on the overall sample fabric.

     After combining the data sets for each of three mutually perpendicular sections, overlap zones of measurement can be identified. All regions within the stereonet are sampled at least twice, with four triangular regions sampled three times (Fig. D1). However, because of the operator bias in inclination measurements, the triple-sampled regions are still very underrepresented relative to other regions (Fig. D2a). The effects of this non-uniform sampling are combined with the inclination corrections used in this study are illustrated in Fig. D2. The sampling under the cubic cosine inclination correction is relatively homogeneous, with triple-sampled areas that are slightly oversampled, and small regions surrounding these areas that are slightly undersampled relative to the average sampling (Fig. D2b). This uneven sampling is somewhat smoothed in the contouring of the data (Appendix C), and in some cases, point maxima and minima are qualitatively judged with respect to their position within or outside one of the overlap zones.