Laboratory measurements of the frictional properties of the Zuccale low-angle normal fault, Elba Island, Italy

Transcription

1 Click Here for Full Article JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115,, doi: /2008jb006274, 2010 Laboratory measurements of the frictional properties of the Zuccale low-angle normal fault, Elba Island, Italy S. A. F. Smith 1,2 and D. R. Faulkner 3 Received 23 December 2008; revised 24 June 2009; accepted 8 September 2009; published 10 February [1] Using a case study from the island of Elba, Italy, we seek to test the hypothesis that the presence of minerals with low frictional strengths can explain prolonged slip on low-angle normal faults. The central core of the Zuccale low-angle normal fault contains a distinctive fault rock zonation that developed during progressive exhumation. Most fault rock components preserve microstructural evidence for having accommodated deformation entirely, or partly, by frictional mechanisms. One millimeter thick sample powders of all the major fault rock components were deformed in a triaxial deformation apparatus under water-saturated conditions, at room temperature, and at constant effective normal stresses of 25, 50, and 75 MPa. Pore fluid pressure was maintained at 50 MPa throughout. Overall, the coefficient of friction (m) of the fault rocks varies between 0.25 and 0.8, emphasizing the marked strength heterogeneity that may exist within natural fault zones. Also, m is strongly dependent on fault rock mineralogy and is <0.45 for fault rocks containing talc, chlorite, and kaolinite and >0.6 for fault rocks dominated by quartz, dolomite, calcite, and amphibole. Localization of frictional slip within talc-rich portions of the fault core can potentially explain movements along the Zuccale fault over a wide range of depths within the upper crust, although the mechanical importance of the talc-bearing fault rocks likely decreased following their dismemberment into a series of poorly connected fault rock lenses. Additionally, slip within clay-bearing fault gouges with m between 0.4 and 0.5 may have facilitated movements in the uppermost (<2 km) crust. For several other fault rock components, m varies between 0.5 and 0.8, and mineralogical weakening alone is insufficient to account for low-angle slip. In the latter fault rock components, other weakening mechanisms such as the development of high fluid pressures, or dissolution-precipitation creep, may have been particularly important in reducing fault strength. Citation: Smith, S. A. F., and D. R. Faulkner (2010), Laboratory measurements of the frictional properties of the Zuccale low-angle normal fault, Elba Island, Italy, J. Geophys. Res., 115,, doi: /2008jb Introduction 1 Reactivation Research Group, Department of Earth Sciences, University of Durham, Durham, United Kingdom. 2 Now at Istituto Nazionale di Geofisica e Vulcanologia, Rome, Italy. 3 Rock Deformation Laboratory, Department of Earth and Ocean Sciences, University of Liverpool, Liverpool, United Kingdom. Copyright 2010 by the American Geophysical Union /10/2008JB006274$09.00 [2] The strength of crustal fault zones represents an outstanding debate in the Earth sciences [e.g., Holdsworth, 2004; Scholz, 2000; Zoback, 2000]. Arguments focus on whether major fault zones support shear stresses consistent with a majority of laboratory-derived friction coefficients ( ; see Byerlee [1978] and Townend and Zoback [2000]), or whether they are able to slip at much lower shear stresses [e.g., Lachenbruch and Sass, 1980; Zoback et al., 1987]. Low-angle normal faults continue to draw widespread attention because their existence is not predicted by standard rock mechanics, assuming a vertical trajectory of the maximum principal stress, s 1, and values of the coefficient of sliding friction of 0.6 < m < 0.85 [Anderson, 1942; Byerlee, 1978]. Under these conditions, normal faults are predicted to experience frictional lock-up at dip angles <30 [Sibson, 1985]. Slip along low-angle normal faults suggests that they may be classified as weak faults, and exposures of low-angle normal faults provide the opportunity to investigate the physical mechanisms leading to fault weakness. [3] Potential mechanisms of fault zone weakening in the middle to upper crust include: (1) the generation and maintenance of high fluid pressures [Axen, 1992; Faulkner and Rutter, 2001; Reynolds and Lister, 1987; Rice, 1992]; (2) the presence of fault rock materials with a coefficient of friction <0.6 [Imber et al., 2001; Moore and Lockner, 2004; Morrow et al., 2000]; (3) activation of stress-enhanced dissolution and precipitation processes [e.g., Bos et al., 2000; Niemeijer and Spiers, 2005; Rutter, 1983]; (4) dynamic fault weakening during coseismic rupture [Di Toro et al., 2004; Han et al., 2007; Rice, 2006]; or a combination of the above. The absence of unambiguously identified moderateto-large earthquakes along low-angle normal faults [e.g., 1of17

2 Collettini and Sibson, 2001; Jackson and White, 1989] suggests that dynamic weakening may be less important in this case, although Wernicke [1995] has argued that unusually long recurrence intervals may occur along lowangle normal faults, and rare examples of pseudotachylyte have been reported [e.g., John, 1987]. [4] Previous work suggests that high fluid pressures and stress-enhanced dissolution and precipitation processes are plausible mechanisms of weakening along some low-angle normal faults [Axen, 1992; Collettini and Holdsworth, 2004; Healy, 2009; Reynolds and Lister, 1987]. Additionally, Numelin et al. [2007] investigated the frictional properties of natural fault gouge from the Panamint Valley low-angle normal fault located in eastern California, which accommodated slip within a m thick, clay-bearing fault core. They determined that samples with a significant proportion of clay minerals possessed a friction coefficient of <0.4, and combined their experimental results with broad estimates of exhumation depth to suggest that localization of strain within the weaker portions of the fault core may explain low-angle slip within the upper 5 km of the crust. [5] Natural fault zones often contain a complex internal structure, reflecting the heterogeneous nature of wall rock lithologies, a wide range of physical and chemical fault zone processes, and the kinematic history of the fault zone [e.g., Chester and Logan, 1987; Chester et al., 2004; Childs et al., 2009; Cowan et al., 2003; Faulkner et al., 2003; Hayman, 2006]. Despite this, most previous studies concerning the frictional properties of natural fault rocks do not elaborate on the details of fault zone structure, or the connectivity of different fault-rock materials, both of which are critical in determining whether grain-scale processes can influence the bulk rheological behavior of the fault zone [Holdsworth, 2004]. In this paper, we use triaxial deformation experiments to determine the frictional strength and stability of natural fault rock materials collected from the Zuccale low-angle normal fault on the Island of Elba, Italy. The internal structure and kinematic history of the Zuccale fault have previously been constrained using field and microstructural relationships [Collettini and Holdsworth, 2004; Smith et al., 2007]. We combine the experimental results with interpretations of fault kinematic history to determine a strength profile for the Zuccale fault, and test the hypothesis that mineralogical weakening may have facilitated prolonged slip. 2. Geological Setting and Previous Work [6] The Zuccale fault is exposed on the island of Elba in the northern Tyrrhenian Sea (Figure 1), an area dominated by recent postcollisional extension driven by eastward rollback of the Ionian-Adriatic subduction zone [Doglioni et al., 1999; Jolivet et al., 1998]. Extension was accommodated along a system of shallowly east-dipping low-angle normal faults, in combination with synthetic and antithetic hanging wall structures [Barchi et al., 1998; Chiaraluce et al., 2007; Collettini et al., 2006b; Jolivet et al., 1998; Pauselli et al., 2006]. [7] Temporary seismic networks in the Umbria-Marche Apennines have highlighted the presence of an active, shallowly east-dipping (15 ) normal fault termed the Altotiberina fault [Chiaraluce et al., 2007]. The Altotiberina fault appears to creep at a rate of 1 mma 1, and produces abundant microseismicity (M L < 2.3) in a stress field characterized by a subvertical s 1 and a subhorizontal, NE-SW trending s 3 [Chiaraluce et al., 2007; Montone et al., 2004]. Exhumation to the west of the Umbria-Marche area has exposed older, inactive low-angle normal faults, including the Zuccale fault on the island of Elba [e.g., Jolivet et al., 1998]. [8] Elba contains diverse continental and oceanic units juxtaposed within a thrust stack during Cretaceous early Miocene collision between the Corsica-Sardinia microplate and the Adriatic plate (Figure 1; see Alvarez [1972]). Trevisan et al. [1967] recognized 5 main thrust complexes dipping to the west, all of which are crosscut by the Zuccale fault (Figure 1b). Displacement across the Zuccale fault determined from offset of preexisting thrust faults is 6 km, and shear sense is uniformly top to the east (Figure 1b; see Collettini and Holdsworth [2004], Keller et al. [1994], and Keller and Coward [1996]). [9] There are several lines of evidence to suggest that the Zuccale fault slipped at a low angle, and thus represents a weak fault (Figure 1c). These are: [10] 1. Preexisting thrust faults in the hanging wall of the Zuccale fault dip on average 30 west, and do not appear to have experienced significant back-rotations [Collettini and Holdsworth, 2004]. [11] 2. There are a suite of high-angle footwall normal faults that were active synchronously with slip along the main Zuccale fault. The orientation of such footwall faults and their interaction with the fault core suggests that only minor reorientations of the Zuccale fault have occurred [Smith et al., 2007]. Additionally, some of the footwall faults are part of conjugate sets that suggest a subvertical maximum principal stress, and a large angle (60 80 ) between the maximum principal stress and the Zuccale fault. [12] 3. The latest set of fault veins are represented by subvertical tensile fractures, consistent with a subvertical maximum principal stress, whereas earlier veins are sheared synthetically within the fault core [Collettini and Holdsworth, 2004]. [13] 4. Steep (60 ) normal faults in the hanging wall sole directly in to the main Zuccale fault [Collettini and Holdsworth, 2004]. [14] 5. The fault zone can be traced beneath extensional basins to the east of Elba, and appears to dip uniformly at a low angle (15 ; see Keller and Coward [1996]). There does not appear to be any reorientation of synextensional sedimentary strata within the basins, nor is there any noticeable offset of the Zuccale fault along later higherangle normal faults [Keller and Coward, 1996]. [15] The Zuccale fault is particularly well exposed in a series of large coastal outcrops at the type locality Punta di Zuccale (Figures 1a and 1b). At Punta di Zuccale, complete sections can be observed through the fault zone, and the collection of samples used in this study focused exclusively on these outcrops (Figures 2a and 2b). [16] Collettini and Holdsworth [2004] suggested that deformation along the Zuccale fault initiated by cataclasis, followed by fluid influx and progressive localization of strain within a central fault core. Within the fault core, they recognized a distinctive fault rock zonation, which they 2of17

3 Figure 1 3of17

4 Figure 2. (a) Photograph of the Zuccale fault at Punta di Zuccale. The dashed lines mark the upper and lower margins of the central fault core, and labels L2 L5 refer to the fault rock components discussed in section 2. (b) Schematic profile through the Zuccale fault at Punta di Zuccale [after Collettini and Holdsworth, 2004; Smith et al., 2007]. The original sample numbers are given on the right-hand side, corresponding to the experiment numbers in Table 1. Note that the lower margin of the fault core is displaced by a series of high-angle footwall normal faults, resulting in fault rock components 2 and 3 occurring as discontinuous lenses of material that are not interconnected within the fault core [Smith et al., 2007]. (c) Schematic diagram illustrating the sequence of fault rock development observed at the Punta di Zuccale outcrop [after Smith et al., 2007]. During the early stages of slip, amphibole schists (component L2) and talc-chlorite-amphibole phyllonites (component L3) accommodated strain and were likely interconnected within the fault core, but these were progressively down-faulted by broadly synchronous movement on footwall normal faults. Progressive exhumation led to the development of foliated cataclasites (component L4), fault breccias (component L5.1), and foliated fault gouges (component L5.2), all of which are present as continuous layers. labeled as layers L2 L5. On the basis of field and microstructural analysis, the fault rock zonation is interpreted to have formed during progressive exhumation of the fault zone from depths of <6 8 km, although this depth remains poorly constrained at present (Figures 2a and 2b; see Collettini and Holdsworth [2004] and Smith et al. [2007]). Figure 1. Geological setting of the Zuccale fault. (a) Location of Elba in the northern Tyrrhenian Sea and detail of the eastern side of Elba, where the Zuccale fault outcrops at the surface. Roman numerals I V refer to the tectonic complexes of Trevisan et al. [1967], which are crosscut and displaced by the Zuccale fault. (b) Cross section through central and eastern Elba, highlighting the geometry of the Zuccale fault at depth, determined from structure contour analysis and borehole records [Bortolotti et al., 2001]. The stars in Figures 1a and 1b mark the location of the Punta di Zuccale outcrop, where samples were collected for triaxial friction tests. (c) Schematic diagram illustrating the lines of evidence that suggest the Zuccale fault slipped at a low angle. See Collettini and Holdsworth [2004], Collettini et al. [2006a], Keller and Coward [1996], Pascucci et al. [1999] and Smith et al. [2007] for details. 4of17

5 Table 1. Summary of Triaxial Friction Experiments a Experiment Sample Fault Rock Component b c Effective s n (MPa) Velocity Steps d (mm/s) Behavior Thickness After Experiments e (mm) PZ24_1 fault gouge L5.2 25, 50, 75 stable sliding 0.45 PZ24_2 fault gouge (dry) L5.2 25, 50, 75 stable sliding PZ5_1 foliated cataclasite L4 25, 50, , 1 stable sliding 0.7 PZ5_2 foliated cataclasite L4 25, 50, 75 stable sliding PERM1_1 chlorite phyllonite L3.3 25, 50, , 1 stable sliding 0.3 PERM1_2 chlorite phyllonite L3.3 50, 75 stable sliding T2 talc phyllonite L3.2 25, 50, , 1 stable sliding 0.38 PZ11_1 amphibole schist L3.1 50, , 1 stable sliding PZ23_1 amphibole schist L2 25, 50, 75 stick-slip PZ23_2 amphibole schist L2 25, 50, 75 stick-slip PZ23_3 amphibole schist L , 1 stick-slip 0.5 a Note that one experiment on the fault gouges was carried out under nominally dry conditions. b Corresponds to the fault rock components labeled in Figures 2 and 10. c Pore fluid pressure was maintained at 50 MPa in all experiments, except experiment PZ24_2, which was carried out under nominally dry conditions. d Initial run-in performed at 0.3 mm/s for 300 mm. e Initial thickness of sample powder = 1 mm. Throughout this paper, the terminology of Collettini and Holdsworth [2004] is adopted to refer to the various fault rock components. [17] Amphibole schists (component L2) and talc-chloriteamphibole phyllonites (component L3) are interpreted to represent the earliest formed fault rock assemblages within the core of the Zuccale fault (Figure 2c, left). The amphibole schists are probably derived from amphibolite pods within the basement of thrust complex I or from ophiolitic assemblages in thrust complex IV. The phyllonites are most likely derived from thrust complex I or from carbonate lithologies of thrust complex II. The phyllonites are interlayered with pervasively recrystallized calcite mylonites, suggesting minimum temperatures of deformation of 150 C. [18] Slip along the Zuccale fault occurred broadly contemporaneously with slip along the network of smalldisplacement footwall normal faults mentioned previously [Smith et al., 2007]. Progressive movement along the footwall faults led to isolation of the amphibole schists and phyllonites within down-faulted sections of the fault core, so that they presently occur as discontinuous lenses of material up to 5 m thick [Smith et al., 2007] (see Figure 2c, middle). [19] During continued exhumation, foliated cataclasites developed within the fault core (component L4), and were probably derived from carbonates and siliciclastic sequences of thrust complex IV. Currently, the foliated cataclasites are represented by a continuous layer of fault rock material up to 3 m thick (Figure 2c, middle). [20] Finally, coarse fault breccias (component L5.1) and incohesive fault gouges (component L5.2) developed during the latest stages of exhumation, and were probably derived from siliciclastic flysch sequences of thrust complex V. Both the fault breccias and fault gouges occur as continuous layers with a total thickness up to 2 m (Figure 2c, right). [21] We sampled four of the main fault rock components for mechanical tests (Figure 2b), all of which exhibit grainscale evidence for having deformed entirely, or partly, by frictional mechanisms. The fault breccias (component L5.1) are chaotic in nature and consist of clasts up to 20 cm in diameter predominantly reworked from the underlying foliated cataclasites. The clasts lie within a finer-grained matrix of comminuted material and carbonate cement. Owing to the size of the clasts and the cemented nature of the matrix, the fault breccias were not suitable for mechanical tests. The phyllonites (component L3) contain layers that are dominated by varying proportions of talc, chlorite, amphibole and calcite. For this reason, we collected 3 samples from different parts of fault rock component 3 to assess its strength heterogeneity (Figure 2b; L ). 3. Experimental Methodology [22] The experimental apparatus used in this study was a high-pressure high-temperature triaxial deformation apparatus with a servo-controlled axial loading system and fluid pressure pump [Mitchell, 2006; Mitchell and Faulkner, 2008]. Silicone oil was used as a confining medium, and distilled water as a pore fluid. During the course of the experiments, the upstream pore pressure was servocontrolled to a resolution of 50 Pa, and the axial load was measured using a calibrated internal force gauge with a resolution of 10 Pa [Mitchell, 2006; Mitchell and Faulkner, 2008]. Axial displacement was measured externally using a displacement Linear Variable Differential Transformer (LVDT) that is in contact with the moving piston, and later corrected for machine compliance [Mitchell, 2006; Mitchell and Faulkner, 2008]. [23] Frictional sliding experiments were conducted on powdered materials of fault rock components L2 L5, excluding the breccias of L5.1 (Table 1). Note that powdering of the samples does not retain any of the natural fault rock fabrics, which may influence the frictional strength (section 5.1). Duplicate tests were carried out to assess the reliability of our results. Sample powders were prepared by passing thumb-sized pieces of fault rock material through a porcelain jaw crusher until they had disaggregated. Fault rock components L2, L3 and L5 were friable and disaggregated easily, whereas component L4 was slightly more cohesive. Sieving techniques were used on each disaggregated sample to separate different grain size fractions. The experiments reported here were carried out on the mm grain size fraction. Semiquantitative X-ray diffraction (XRD) analysis was performed on material derived from the experimental grain size fraction, allowing us to place constraints on modal mineralogy (Table 2). 5of17

6 Table 2. X-Ray Diffraction Results a Sample Fault Rock Component Quartz (%) Amphibole (%) Calcite (%) Dolomite (%) Chlorite (%) Fault gouge L Foliated cataclasite L Chlorite phyllonite L Talc phyllonite b L Amphibole schist L Amphibole schist L a Error is typically ±5%. b Estimated from thin section and scanning electron microscope observations. Talc (%) Kaolinite (%) [24] A schematic diagram of the sample assembly is shown in Figure 3. An initially 1 mm thick layer of sample powder was placed between 18.5 mm diameter forcing blocks containing a 30 saw cut surface. The powder was first made in to a paste using deionized water. Powder layers were constructed on the saw cut surfaces using a precision leveling jig to ensure uniform thickness. Both forcing blocks were made from a circular cylinder of hardened steel. The elliptical saw cut surfaces on the forcing blocks were roughened prior to each experiment using European grade P400 (320 grit) silicon carbide paper with an average grit size of 35 mm, estimated to produce an RMS surface roughness around 0.36 mm. Although this produces a smooth surface finish, examination of the gouge microstructure following the experiments indicated that slip was not localized at the boundaries between the sample powders and the forcing blocks. The forcing blocks were inserted in to a polyolefin jacket, a 0.35 mm thick annealed copper jacket, and an outer 2.5 mm thick PVC jacket made from Nalgene tubing (Tygon 1 R-3603). This arrangement prevented any leakage of pore fluid between the forcing blocks and the copper jacket during the course of the experiments. The PVC jacket ensured that if the copper jacket ruptured during shear, confining fluid could not enter the sample. Inclusion of a Teflon spacer at the downstream end of the sample assembly allowed decoupling of the lower forcing block, preventing any rotation of the forcing blocks during the experiments. [25] We designed our experiments to assess the frictional strength and stability of the fault rock materials under watersaturated conditions at three different effective normal stresses. Each experiment consisted of loading an individual sample powder into the pressure vessel of the triaxial apparatus, deforming the sample at 25, 50, and 75 MPa effective normal stress, and removing the sample from the pressure vessel. All the experiments were conducted at room temperature. Confining pressures were initially increased to 75 MPa, and pore pressures to 50 MPa, equivalent to an effective normal stress of 25 MPa. Pore fluid was introduced through two small holes 1 mm in diameter that were drilled perpendicular to the saw cut surfaces in the steel forcing blocks (Figure 3). Pore fluid was allowed to equilibrate for approximately 60 min before commencing the experiments. In all the experiments we performed an initial run-in at an axial displacement rate of 0.3 mm/s for 300 mm, or until the sample had clearly experienced yield. Typically, we sheared each sample at each effective normal stress for a total of 1.4 mm, producing a total axial displacement during each experiment of 4.2 mm. This corresponds to a displacement along the saw cut surface of 4.9 mm. We calculated the frictional strength as the ratio of shear stress to normal stress resolved on the saw cut surface. In some of the experiments velocity stepping tests were performed at each effective normal stress to characterize the velocity dependence of the fault rock materials (Table 1). Normal stress was maintained at a constant value throughout using servo-controlled adjustments to the confining pressure. The mechanical data were corrected for the elastic axial distortion of the deformation apparatus. We have chosen to report the raw mechanical data without applying a correction for the strength of the jacketing materials, and without a correction for decreasing overlap area of the two elliptical forcing blocks during progressive displacement. However, we did carry out friction experiments using thin sheets of Teflon instead of sample powder. The frictional properties of Teflon are well known, so we Figure 3. Schematic diagram of the pressure vessel and sample assembly used in triaxial deformation experiments. 6of17

7 were able to calculate the approximate shear strength of the jacketing materials, reported in Figure 5. The strength of the jacketing materials at pressure does not make a significant contribution to the overall mechanical behavior. [26] Following the experiments, fragments of the deformed sample powders were immersed under vacuum in low-viscosity resin and polished samples were prepared for observation in the scanning electron microscope. The experimental samples were cut perpendicular to the deformed sample layer and approximately parallel to the shearing direction. 4. Results 4.1. Composition and Textural Characteristics of Natural Fault Rock Samples [27] Amphibole schists (component L2) are dominated by a matrix of platy, fine-grained (<50 mm) amphibole grains (Figure 4a), that surround larger fractured clasts of amphibole. A foliation is defined by a grain-scale preferred orientation of platy grains and by a compositional banding of amphibole with different Fe contents. The clasts contain intragranular fractures suggesting that grain size reduction was achieved by progressive brittle fragmentation, while the matrix is crosscut by numerous discrete cataclastic horizons containing Fe-rich amphibole (Figure 4b). [28] Talc and chlorite phyllonites (component L3) are typically characterized by a foliated matrix containing talc, chlorite, and minor amphibole grains, that surround polycrystalline clasts or individual fractured clasts of calcite and dolomite (Figures 4c and 4d). Collettini et al. [2009] suggested that talc in these fault rocks formed by the reaction of dolomite with silica-bearing hydrothermal fluids, and that deformation occurred in talc by frictional slip along (001) foliation planes. The phyllonites are also associated with layers containing only calcite and amphibole. We collected one sample containing 60% calcite and 39% amphibole (component L3.1, labeled as amphibole schists), one sample containing 70% talc (component L3.2, labeled talc phyllonites), and one sample containing a total of 28% talc and chlorite (component L3.3, labeled as chlorite phyllonites). [29] Foliated cataclasites (component L4) are dominated by subrounded to subangular polycrystalline clasts of dolomite that lie within a finer grained matrix of quartz and dolomite, with subsidiary kaolinite, calcite, and apatite (Figures 4e and 4f). Some of the polycrystalline dolomite clasts contain aggregates of crystals with blocky-elongate grain morphologies, suggesting derivation from preexisting dolomite veins within the fault core. The foliated cataclasites contain a mesoscopic P foliation defined by a preferred orientation of large reworked inclusions of fault rock components L2 and L3. Additionally, a microscopic P foliation is defined by a preferred orientation of the long axes of polycrystalline clasts (Figure 4e; see Rutter et al. [1986]). A pervasive matrix foliation is typically not visible (e.g., Figure 4f) except in areas with a relatively high concentration of clay phases. Grains of quartz and dolomite in the matrix are crosscut by intragranular fractures suggesting that grain size reduction occurred by brittle fragmentation (Figure 4f inset). [30] Fault gouges (component L5.2) are dominated by varying proportions of quartz, calcite, dolomite, and clay phases such as chlorite and kaolinite (Figures 4g and 4h). Large survivor grains, which are rounded, isolated grains that appear to have escaped grain fracturing processes [Cladouhos, 1999; Engelder, 1974], typically account for <15% of the gouges on a hand-specimen scale. The survivor grains are dominantly composed of polycrystalline dolomite, partly or completely replaced by calcite (Figure 4g), although survivor grains of sandstone and kaolinite derived from the hanging wall are also present. The surrounding matrix contains chlorite, kaolinite, and calcite grains that define a microscopic P foliation, enhanced by a preferred orientation of the long axes of small quartz, dolomite, and calcite survivor grains (Figure 4h). Kaolinite also occurs as stacked pseudohexagonal plates occupying matrix pore space Triaxial Friction Experiments Frictional Strength [31] Figures 5 and 6 present the results of triaxial friction experiments at 25, 50 and 75 MPa effective normal stress carried out on the six fault rock samples shown in Figure 2, in addition to the strength of the jacketing materials. For duplicate tests (Table 1), only one experiment is shown for clarity in Figure 5, but all experiments are summarized in Figure 6. Reproducibility of duplicate tests was generally good. The largest difference in the friction coefficients between duplicate experiments occurred for the chlorite phyllonites (component L3.3) at 75 MPa, and is equivalent to m = 0.07 after 500 mm displacement (Figure 6). For all the other repeat experiments at each confining pressure, the differences in m were 0.05, and are sufficiently small that they do not affect any comparisons of the relative strengths between the different fault rock components (Figure 6). [32] All the samples experienced yield at <300 mm axial displacement, and then followed similar strain hardening trends (Figure 5). Values of m discussed herein, and shown in Figure 6, were measured after 500 mm axial displacement (gray bar in Figure 5), and after m had been corrected to account for the strength of the jacketing materials at 500 mm displacement (Figure 5). [33] Overall, m ranges between 0.25 and 0.80, emphasizing the marked strength heterogeneity that can exist within natural fault zones. With the exception of the talc and chlorite phyllonites, m increases with increasing effective normal stress (Figure 6). [34] At 25 MPa effective normal stress, m varies between 0.28 and The talc and chlorite phyllonites, and the water-saturated fault gouges, have m The talc phyllonites are the weakest samples with m = The foliated cataclasites and the amphibole schists have m [35] At 50 MPa effective normal stress, m varies between 0.25 and The talc and chlorite phyllonites, and the water-saturated fault gouges have m The talc phyllonites remain the weakest samples with m =0.25.The foliated cataclasites and amphibole schists have m [36] At 75 MPa effective normal stress, m varies between 0.31 and The talc and chlorite phyllonites have m 0.5. The talc phyllonites are the weakest samples with m = 0.31, and the water-saturated fault gouges have m = The foliated cataclasites and amphibole schists have m of17

8 Figure 4 8of17

9 [37] One experiment was performed on nominally dry fault gouges (Table 1), which have a frictional strength greater than the water-saturated fault gouges (Figure 6) Velocity Dependence of Strength [38] The rate dependence of frictional strength can be characterized by the change in steady state friction coefficient, Dm ss, which occurs after imposing a logarithmic change in sliding velocity, Dm ss /DlnV (Figure 7). This is equivalent to the quantity (a b), where a represents the so-called direct effect in the framework of rate-and-state friction, and b represents an evolution effect that occurs over a critical slip distance (see Dieterich [1979] and Marone [1998] for a comprehensive review, as well as Ruina [1983]). When (a b) is positive, the material is said to be velocity strengthening, leading to inherently stable behavior. If (a b) is negative, the material is velocity weakening, potentially leading to unstable behavior [Dieterich, 1979; Ruina, 1983]. [39] Owing to the limited displacements that could be achieved during the triaxial experiments, only 2 or 3 velocity steps were carried out at each effective normal stress. In our experiments, we imposed velocity changes from 0.1 to 1 mm/s and from 1 to 0.1 mm/s. We observed that the 0.1 to 1 mm/s velocity steps were generally characterized by a poorly defined and delayed direct effect (point x in Figure 7). We believe this resulted from using impermeable steel forcing blocks, which meant that pore fluids required a relatively long period of time to equilibrate with the samples following an increase in sliding velocity. This delayed response is not observed following decreases in sliding velocity, because pore fluids had sufficient time to equilibrate with the samples (Figure 7). We have only used the 1 to 0.1 mm/s velocity steps to determine the values of (a b). This means that, typically, only one velocity step was available to determine (a b). Further experiments carried out to higher displacements, and with a larger number of velocity steps, are required to confirm our preliminary measurements. The program xlook was used to process the velocity-stepping friction data (obtained from Chris Marone, Pennsylvania State University). Blanpied et al. [1995] have suggested that the contribution of the copper jacket to the quantity (a b) is <0.0002, and we assumed that the other jacketing materials used in this study also had negligible effects. [40] All except one of the measured values of (a b) are positive, indicating that most of the fault rocks are velocity strengthening and prone to stable sliding under the experimental conditions reported here (Figure 8). After imposing a change in sliding velocity, most samples exhibited an instantaneous direct effect, followed by an evolution effect of the opposite sign, before returning to a new steady state frictional strength (Figure 7). However, the amphibole schists of component 3 (component L3.1), and the chlorite phyllonites at 75 MPa effective normal stress (component L3.3), displayed near-zero or negative b values, resulting in relatively large values of (a b) (Figure 8). The physical basis for the occurrence of near-zero or negative b values remains poorly understood, although they may be related to grain-scale contact saturation, in which individual mineral surfaces are in full contact with one another (see Ikari et al. [2009] for recent discussions and Saffer and Marone [2003]). Near-zero or negative b values have previously been reported for quartz and granite gouge at elevated temperatures [Blanpied et al., 1998; Karner et al., 1997], smectite gouge at room temperature [Saffer and Marone, 2003], and talc gouge over a wide temperature range [Moore and Lockner, 2008] Textural Characteristics of Experimental Gouge Layers [41] We observed no localization of slip along boundary shears at the margins between the sample layers and the steel forcing blocks. After loading and shearing at 75 MPa effective normal stress, the samples containing phyllosilicates compacted from 1 mm in thickness to 0.4 mm, while the samples dominated by amphibole, quartz, calcite, and dolomite compacted from 1 mm to mm (Table 1). [42] The samples of amphibole schist (components L2 and L3.1) and foliated cataclasite (L4) deformed predominantly by fracturing and grain size reduction of clasts (Figures 9a and 9b). Fracturing was most evident at the Figure 4. Optical photomicrographs and backscattered scanning electron microscope images illustrating the microstructural characteristics of natural fault rock samples. All samples were cut perpendicular to foliation and parallel to lineation. Figures 4a, 4c, and 4e show optical photomicrographs in crossed-polarized light, and Figure 4g shows an optical photomicrograph in plane-polarized light. Figures 4b, 4d, 4f, and 4h show scanning electron microscope images. Abbreviations are as follows: Am, amphibole; Ca, calcite; Cl, chlorite; Tc, talc; Q, quartz; Do, dolomite; Ka, kaolinite; and Py, pyrolusite. (a) Amphibole schists (component L2) containing a compositional banding up to several hundred microns thick, defined primarily by variations in iron content. (b) Two images of cataclastic horizons within the amphibole matrix. The white grains are relatively Fe-rich amphibole. (c) Phyllonites (component L3) containing a compositional banding of chlorite-, talc- and amphibole-rich layers with calcite-rich layers. (d) Detail of the chlorite- and talc-rich layer enveloping fractured clasts of calcite. The chlorite and talc grains are typically well aligned and present as a grain-scale interconnected network [see Collettini et al., 2009]. (e) Foliated cataclasites (component L4) dominated by quartz and dolomite. Dolomite occurs as large polycrystalline clasts that are extensively fractured within a dolomite and quartz matrix. Note the preferred orientation of the long axes of dolomite clasts that contribute toward a P foliation observed in the field. (f) Detail of quartz and dolomite matrix. No preferred orientation of grains is visible in this example. The inset image shows an intragranular fracture filled with relatively Fe-rich dolomite crosscutting a grain of relatively Fe-poor dolomite. (g) Foliated fault gouges dominated by a fine-grained matrix of quartz, calcite, dolomite, and kaolinite surrounding survivor grains which include polycrystalline dolomite and calcite. Note posttectonic growth of dendritic pyrolusite. (h) Detail of fault gouge matrix, showing the preferred orientation of small survivor grains of dolomite, within a foliated matrix of chlorite, kaolinite, and calcite. 9of17

10 Figure 6. Friction coefficient versus effective normal stress for all experiments. Note the addition of one experiment carried out at 60 MPa normal stress on component 2. The dry experiment on the fault gouges is also indicated. For samples that display identical friction coefficients, the symbols have been moved to one side for clarity. contacts between large clasts (e.g., Hertzian fractures in Figure 9a). There was no development of a clear shapepreferred orientation of individual clasts, but the foliated cataclasites were crosscut by poorly developed Riedel R 1 shears that lie at an angle of 15 to the margins of the deformed sample (Figure 9c; see Rutter et al. [1986]). The Riedel shears are defined by generally finer grain sizes than the surrounding matrix material, and appear to extend from one boundary of the sample layer to the other (Figure 9c). [43] The samples of chlorite and talc phyllonite (components L3.2 and L3.3) both contain phyllosilicate grains that have a preferred orientation approximately parallel to the margins of the deformed sample layers (Figure 9d). In places the chlorite and talc grains form layers parallel to the sample margins that appear connected on a grain scale, although the layers are not continuous along the full length of the sample layer. The chlorite and talc grains surround fractured clasts of calcite that show no preferred orientation 10 of 17 Figure 5. Summary of friction experiments. Each experiment consisted of three effective normal stress steps (s 0 n = 25, 50 and 75 MPa) that are presented separately. Displacement values at 50 and 75 MPa effective normal stress start at 0 mm, even though some displacement had already accumulated during the preceding normal stress steps. All experiments were done under saturated conditions with the pore fluid pressure constant at 50 MPa, except for one nominally dry experiment on the fault gouges of component 5. The strength of the jacketing materials, determined from tests using Teflon wafers, is also shown. Each of the curves is labeled with the number of the fault rock component shown in Figures 2 and 10. Reported friction coefficients were taken after 500 mm axial displacement (vertical gray bar). For duplicate experiments (see Table 1), only the results of one experiment are shown in Figure 5 for clarity.

11 Figure 7. Summary of a velocity step excerpted from experiment PERM1_1 at 50 MPa effective normal stress. Velocity steps from 0.1 to 1 mm/s are associated with a delayed and rounded direct effect (at point X) that we believe reflects our use of steel forcing blocks. Only the mm/s velocity steps were used to determine the quantity (a b). (Figure 9d). Additionally, there was no clear localization of slip within crosscutting Riedel shear bands. [44] The fault gouges (component L5) contain clasts of calcite and dolomite not exceeding 100 mm in size, although usually they are <50 mm (Figure 9e). Relatively large clasts sit within a finer-grained matrix of comminuted dolomite and calcite, and platy clay phases consisting of chlorite and kaolinite (Figure 9e). There is no clear preferred orientation of grains in the matrix, and no localization of slip within Riedel shear bands. 5. Discussion 5.1. Relationship Between Frictional Properties and Fault Rock Mineralogy [45] Table 3 provides a summary of rock friction experiments relevant to the fault rock materials discussed in this paper. Most of the experiments in Table 3 were carried out on layers of gouge material sandwiched between opposing forcing blocks, and thus the frictional strengths are analogous to those determined in this paper. [46] Amphibole schists at the base of the fault core (component L2), and mixed amphibole-calcite lithologies (component L3.1), are characterized by friction coefficients between 0.62 and 0.8. The authors are not aware of any experimental data on the frictional properties of pure amphibole, but it is not expected to be weak under the experimental conditions used in this study, and most likely possesses typical Byerlee friction values in the range [Byerlee, 1978]. Morrow et al. [2000] reported m 0.7 for wet calcite gouge deformed at room temperature (Table 3). [47] The weakest fault rock samples within the core of the Zuccale fault are the talc phyllonites (components L3.2), which have frictional strengths of The friction coefficients of the talc phyllonites are slightly elevated above those reported by Moore and Lockner [2008] for pure talc gouge deformed at room temperature (Table 3). This small difference is most likely due to our samples containing 30% calcite and amphibole. Moore and Lockner [2008] also found that the dependence of m on effective normal stress was smaller for talc than for other sheet silicate minerals such as montmorillonite and muscovite. Our results are broadly consistent with the observations of Moore and Lockner [2008], as we observe no obvious stress dependence of m for the talc phyllonites. [48] The friction coefficients of the chlorite phyllonites (L3.3), containing 12% chlorite and 16% talc, are These values are elevated above those reported by Moore and Lockner [2008] for pure talc (see above paragraph) and slightly elevated above those reported for pure chlorite gouge deformed at room temperature (Table 3). Nevertheless, talc and chlorite in our experimental samples appear to be strongly influencing the frictional strength, despite the fact that they account for <30% of the sample (more than 70% calcite and amphibole; see Table 2). This result supports the idea that phyllosilicates may significantly reduce the strength of fault rock materials if they form grain-scale layers that are able to support the applied normal load during shearing [e.g., Holdsworth, 2004; Imber et al., 2001]. Our microstructural observations indicate that the experimental samples of chlorite and talc phyllonite are characterized by platy phyllosilicate minerals aligned parallel to the sample margins, in places forming layers that separate isolated clasts of calcite and amphibole (Figure 9d). We suggest that the talc and chlorite are effectively lubricating clasts of the stronger mineral phases, preventing strong grain-to-grain contacts from developing and accounting for the low frictional strengths. Both the talc and chlorite phyllonites are dominantly velocity strengthening at the imposed slip rates, consistent with the documented velocity-strengthening behavior of both talc and chlorite (Table 3). [49] The above result prompts a brief discussion on the potential importance of foliations in the natural fault rock samples. Our experimental procedure is optimized for use with powdered material, but the fact that talc and chlorite Figure 8. Effective normal stress versus (a b). 11 of 17

12 Figure 9. Backscattered SEM images showing the microstructural characteristics of experimentally deformed fault rock powders. All samples were cut perpendicular to the deformed gouge layer and approximately parallel to the shearing direction. None of the samples show evidence for localization of slip along boundary shears. (a) Amphibole schists (fault rock components L2 and L3.1) showing cataclastic grain size reduction and grain rotation. Note the presence of Hertzian fractures at the contact of the two large grains in the center right of the image. (b) Foliated cataclasites (component L4) showing cataclastic grain size reduction and grain rotation. (c) Deformed layer of foliated cataclasites showing the development of an incipient Riedel R 1 shear that extends from the upper to the lower sample boundary. Because the Riedel shear is only weakly developed, the quartz and dolomite clasts have been highlighted in white, and the background matrix in black using the software program Scion Image. The sample layer is 700 mm thick. (d) Phyllonites (components L3.2 and L3.3) containing layers of platy chlorite and talc grains. In places the chlorite and talc line the boundaries of fractured calcite clasts. (e) Fault gouges (component L5) are characterized by small, rounded survivor grains of calcite and dolomite within a fine-grained matrix of chlorite and kaolinite. appear to strongly influence friction at low abundances in our experiments suggests that concentration of slip along throughgoing talc- and chlorite-bearing foliation planes in the natural samples could reduce their frictional strengths toward those of the single-mineral strengths. Thus, if deformation in the natural samples were to localize along foliation planes, the frictional strengths determined in our experiments may represent upper limits for the strengths of 12 of 17

13 Table 3. Summary of Friction Experiments Carried out on Minerals Relevant to This Study a Friction Coefficient (m) Effective Normal Stress (MPa) Axial Displacement Velocity (mm/s 1 ) Velocity Moisture Displacement Mineral Dependence b Content (mm) Reference Calcite saturated Morrow et al. [2000] Chlorite saturated Morrow et al. [2000] oven-dry <4 0.5 Moore and Lockner [2004] saturated < Brown et al. [2003] saturated < Kopf and Brown [2003] (0.02) saturated < Ikari et al. [2009] c Dolomite (0.005) 75 very dry Weeks and Tullis [1985] d Kaolinite saturated < Morrow et al. [1982] saturated Morrow et al. [2000] saturated Moore and Lockner [2004] oven-dry <4 0.5 Moore and Lockner [2004] (0.003) 100 room humidity < Ikari et al. [2007], M. Ikari, unpublished data [2009] saturated <5 0.3 Crawford et al. [2008] room humidity <5 0.3 Talc saturated Morrow et al. [2000] saturated Moore and Lockner [2004] e + (0.007) saturated < Moore and Lockner [2008] f + (0.01) saturated < e room humidity <3 3 Escartin et al. [2008] 0.1 g room humidity < Escartin et al. [2008] (0.008) room humidity Carpenter et al. [2009] Quartz saturated < Logan and Rauenzahn [1987] (0.004) saturated < Marone et al. [1990] (0.014) room humidity Mair and Marone [1999] ( 0.007) room humidity saturated Morrow et al. [2000] saturated <5 0.3 Crawford et al. [2008] Westerly granite (0.002) 25 room humidity < Beeler et al. [1996] ( 0.002) > (0.004) >100 a The results of Beeler et al. [1996] on a simulated gouge layer of Westerly granite are also included for discussion purposes. b Values in parentheses are the maximum or minimum values of (a b). c Sample was natural chlorite schist containing 46% chlorite, 35% plagioclase, 12% quartz, and 6% illite. d Starting assembly consisted of bare dolomite surfaces, but Weeks and Tullis [1985] suggest that the friction they report is appropriate for a layer of gouge material that developed during the course of the experiments. The sample is reported as very dry owing to confining gas leaking through the sample during the course of the experiment. e Room temperature experiments. f Experiments at C. g Experiments at C. One experiment was conducted under saturated conditions with a pore fluid pressure of 59 MPa. the natural samples, particularly given the relatively high laboratory strain rates. [50] Samples of foliated cataclasite (component L4) are dominated by quartz and dolomite, in roughly equal proportions, and possess m values between 0.53 and m appears to increase significantly with increasing normal stress. These strength values are consistent with previously reported strengths of both quartz and dolomite gouge deformed at room temperature (Table 3). [51] Our microstructural observations suggest that the experimental samples of foliated cataclasite accommodated slip by distributed cataclasis. Y shears were not observed in the experiments, while incipient development of Riedel-R 1 shears occurred at the highest displacements (5 mm; see Figure 9c). These observations may explain the observed velocity-strengthening behavior of the foliated cataclasites, and are consistent with the results of Beeler et al. [1996] and Mair and Marone [1999]. These authors found that velocity-strengthening behavior dominated in quartz and granite gouges at slip displacements of <10 20 mm, correlating with a lack of microstructural fabrics indicating slip localization (Table 3). At slip displacements >10 20 mm, they observed a transition to velocity-weakening behavior, which correlated with the development of boundary and Riedel shears. Beeler et al. [1996] observed a second transition back to velocity strengthening at high displacements (>100 mm), and used their observations to suggest that velocity weakening may be directly associated with the development and maintenance of boundary-parallel Y shears. Additionally, Weeks and Tullis [1985] observed velocitystrengthening behavior in one experiment performed on dolomite gouge at small displacements (Table 3). [52] XRD and SEM analyses suggest that the fault gouges (component L5) contain quartz, calcite and dolomite as the dominant mineral phases, with chlorite and kaolinite as important matrix phases. Water-saturated fault gouges are characterized by m between 0.38 and 0.54, while nominally dry fault gouges are typically 25 40% stronger, with m between 0.48 and m in the fault gouges increases significantly with increasing effective normal stress. Quartz, 13 of 17

14 Figure 10. Summary of the fault rock and mechanical zonation present within the core of the Zuccale fault. (a) Fault zone profile highlighting the fault rock zonation observed within the core of the Zuccale fault. (b) Friction coefficients of the different fault rock components. The range of typical Byerlee friction coefficients (m = ) is also shown. The bar labeled M&L shows the range in friction coefficients for pure talc gouge measured by Moore and Lockner [2008]. The hatched areas represent the range of friction coefficients required to explain slip along a normal fault dipping at 15, assuming hydrostatic pore pressure and reasonable values of cohesion and internal friction [Axen, 2004]. For the case of fault rock components L2, L3, and L4, depths of formation are poorly constrained, but a friction coefficient of <0.24 can explain slip over a wide range of depths in the upper crust. Fault rock component L5 likely formed at depths of <2.6 km, where a friction coefficient of <0.4 can facilitate low-angle slip. The open symbols associated with L5.2 are for the nominally dry experiment. (c) Quantity (a b) for the different fault rock components. The dominant velocity-strengthening behavior of the fault materials at these experimental conditions suggests that the most likely mode of failure along the Zuccale fault was stable fault creep. calcite, and dolomite have frictional strengths that are too high to explain the strength of the fault gouges (Table 3). In contrast, chlorite and kaolinite have frictional strengths that may adequately explain both the nominally dry and watersaturated strengths of the fault gouges (Table 3). We suggest that the frictional strengths of the experimental fault gouges may be explained by platy grains of chlorite and kaolinite in the matrix lubricating stronger clasts of quartz, calcite, and dolomite, by a similar mechanism to that proposed for the talc and chlorite phyllonites above. The velocity-strengthening behavior of the fault gouges at 50 and 75 MPa effective normal stress is consistent with the velocitystrengthening behavior of chlorite and other nonexpandable clays such as kaolinite (Table 3) Mineralogical Weakening Along Low-Angle Normal Faults [53] Figure 10 presents a summary of the experimental results, highlighting the fault rock zonation observed within the central core of the Zuccale fault, interpreted to record exhumation from relatively deep to relatively shallow crustal levels [Collettini and Holdsworth, 2004]. Microstructural analyses and crosscutting relationships suggest that strain within the fault core at the sampled outcrops was initially accommodated within fault rock components L2 and L3, before being superseded by component L4, and ultimately by component L5 (Figure 2c). [54] Our experiments, combined with previously determined field relationships, provide insights in to the frictional strength of the Zuccale fault over prolonged periods of time, and suggest that the importance of mineralogical weakening may have varied considerably in response to changes in fault zone structure. [55] During periods of time when strain within the core of the Zuccale fault was localized within the talc phyllonites, mineralogical weakening may explain slip along the Zuccale fault without the need to appeal to other potential weakening mechanisms, although we recognize that high fluid pressures and dissolution-precipitation mechanisms may have weakened the fault further (Figure 2c, left). Although the depths of formation of fault rock components L2 L4 remains poorly constrained, talc has a coefficient of friction sufficiently low to account for slip along a normal fault dipping at 15 over a wide range of depths in the upper crust [Axen, 2004; Collettini et al., 2009; Moore and Lockner, 2008]. [56] Similarly, the fault gouges (component L5) have a sufficiently low strength to account for slip during the late stages of exhumation. We believe the fault gouges deformed, at least in part, at depths of less than approximately 2.6 km on the basis of observations that: (1) the gouges contain survivor grains derived from the underlying fault rock components, most noticeably the foliated cataclasites (L4), suggesting that they were the last fault rock component to 14 of 17