Permeability and seismic velocity and their anisotropy across the Alpine Fault, New Zealand: An insight from laboratory measurements on core from the Deep Fault Drilling Project phase 1 (DFDP‐1)

The Alpine Fault, a transpressional plate boundary between the Australian and Pacific plates, is known to rupture quasiperiodically with large magnitude earthquakes (Mw ~8). The hydraulic and elastic properties of fault zones are thought to vary over the seismic cycle, influencing the nature and style of earthquake rupture and associated processes. We present a suite of laboratory permeability and P (Vp) and S (Vs) wave velocity measurements performed on fault lithologies recovered during the first phase of the Deep Fault Drilling Project (DFDP‐1), which sampled principal slip zone (PSZ) gouges, cataclasites, and fractured ultramylonites, with all recovered lithologies overprinted by abundant secondary mineralization, recording enhanced fluid‐rock interaction. Core material was tested in three orthogonal directions, orientated relative to the down‐core axis and, when present, foliation. Measurements were conducted with pore pressure (H2O) held at 5 MPa over an effective pressure (Peff) range of 5–105 MPa. Permeabilities and seismic velocities decrease with proximity to the PSZ with permeabilities ranging from 10−17 to 10−21 m2 and Vp and Vs ranging from 4400 to 5900 m/s in the ultramylonites/cataclasites and 3900 to 4200 m/s at the PSZ. In comparison with intact country rock protoliths, the highly variable cataclastic structures and secondary phyllosilicates and carbonates have resulted in an overall reduction in permeability and seismic wave velocity, as well as a reduction in anisotropy within the fault core. These results concur with other similar studies on other mature, tectonic faults in their interseismic period.


Introduction
The quantitative characterization of the physical properties of fault zones, such as permeability and the transmission of seismic waves, is essential in order to understand the influence they have on fluid flow and seismicity in the brittle crust. Upper-crustal fault zones can be divided into three primary components: the fault core, the damage zone, and the host protolith. These components may be lithologically and structurally distinct and vary in proportion according to geological setting [Chester and Logan, 1986;Caine et al., 1996;Faulkner et al., 2010]. The fault core can consist of singular or multiple anastomosing slip surfaces and is where the highest strain and most displacement are accommodated [Faulkner et al., 2003]. This core is usually nested within a damage zone, a region of kinematically related cracks and small faults, decreasing in density with distance from the core and transitioning into the local host protolith [Anders and Wiltschko, 1994;Wilson et al., 2003;Mitchell and Faulkner, 2009].
The Alpine Fault in New Zealand is a major fault that forms the active plate boundary between the Pacific and Australian plates. The exhumation of the Alpine Fault hanging wall enables us to make direct observations of the internal structure of a large plate-boundary fault, which is late in its seismic cycle and could therefore provide important information on the fault state at the time prior to an earthquake. The damage zone of the Alpine Fault of New Zealand is composed of diverse fault rock lithologies including clay-rich gouges, cataclasite, ultramylonites, and fault breccias, all overprinted, to different extents, by a geochemically altered zone, which, in this fault, masks evidence of a brittle-viscous transition [Sibson et al., 1981;Caine et al., 1996;Sutherland et al., 2012]. • Low permeability and seismic velocity is associated with the damage zone of the Alpine Fault Zone, New Zealand • The development of pervasive gouge-filled and calcite-filled fractures explains the low permeability and seismic velocity • Seismic velocity anisotropy is low (<30%) due to the lack of preferred orientation of the fracture network in the fault damage zone Supporting Information: • Supporting Information S1 • Table S1 • Table S2 • Table S3 • Figure S1 • Movie S1 • Movie S2 In this paper, laboratory measurements of permeability and ultrasonic wave velocity are made on fault rock samples recovered from the central Alpine Fault Zone. These specimens were sampled from core collected from two shallow (<150 m depth) boreholes that were drilled as part of the Deep Fault Drilling Project (DFDP-1). These cores provide a unique opportunity to sample through a continuous section across the Alpine Fault Zone, uninterrupted by thick vegetation and relatively unaffected by surface weathering effects. First, a brief introduction to the Alpine Fault Zone and the DFDP project is given. Then the techniques used to prepare the delicate samples analyzed are described. Finally, results of permeability, P and S wave velocity measurements, including their anisotropy, and microstructural controls are described, and their influence on fault zone processes during and after seismicity is discussed.

The Alpine Fault, New Zealand, and DFDP Drilling
The Alpine Fault Zone (AFZ), bounding the western edge of the Southern Alps on New Zealand's South Island, accommodates over 70% of the relative motion between the colliding Australian and Pacific Plates [Norris and Toy, 2014]. The ramping of the Pacific plate over the Australian plate is accommodated through dextral-reverse movement, offsetting basement rocks laterally by~470 km, with Quaternary horizontal displacement rates estimated at 21-27 mm yr À1 [Norris and Cooper, 2001;Lamb et al., 2015]. Uplift across the fault is rapid, exhuming crustal rocks from depths of 20-30 km at a rate of~6-9 mm yr À1 [Little et al., 2002]. Accompanying this rapid uplift is the advection of the regional geothermal gradient. Quartz fluid inclusion microthermometry places the gradient at 40°C/km [Toy et al., 2010], while titanium thermobarometry of quartz shows a gradient of 45°C/km [Cross et al., 2015] and recent shallow borehole measurements report gradients of 62.6 ± 2.1 and 125 ± 55°C/km .
Palaeoseismic evidence indicates that the Alpine Fault has had numerous geologically recent rupture episodes at 1717, 1620, and 1430 A.D., estimated to have moment magnitudes of 7.9 ± 0.3, 7.6 ± 0.3, and 7.9 ± 0.4, respectively . This, coupled with recent seismicity patterns, geodetic strain records, and regional low-velocity zones indicative of high pore fluid pressures deep in the fault zone, suggests that the fault is locked above 6-12 km depth and is in the late stage of the earthquake cycle [Eberhart-Phillips, 1995;Sutherland et al., 2007;De Pascale and Langridge, 2012;Lamb and Smith, 2013]. The global significance of the Alpine Fault in the study of major fault zones and crustal processes is considerable; the opportunity to study a locked continental plate boundary fault thought to be late in the earthquake cycle will yield insight into the prerupture conditions of a major fault and the nature of the interseismic physicochemical processes that have taken place.
The Deep Fault Drilling Project (DFDP) was undertaken to understand better the processes that lead to major earthquakes by drilling into and observing a major continental fault during its buildup to a large seismic event. The first phase of the Deep Fault Drilling Project (DFDP-1) comprised two pilot holes (DFDP-1A and DFDP-1B) drilled into an active thrust segment of the AFZ at Gaunt Creek, Westland, in January 2011 ( Figure 1). The boreholes reached depths of 100.6 and 151.4 m, respectively, sampling a range of fault lithologies including ultramylonites, cataclasites, and gouges. All samples used in this study were collected from core recovered from DFDP-1A and 1B. The principal slip zone (PSZ) of the AFZ was reached at 90.74 m depth in DFDP-1A and 128.44 m in DFDP-1B, marking the boundary between the Australian and Pacific Plates, with a second PSZ within footwall rocks recorded in DFDP-1B at 144 m. The PSZ in both boreholes is composed of clay-rich gouge ( Figure 1) .
Drilling revealed the presence of a~30 m thick alteration zone overprinting both the fault core and damage zone. This "alteration zone" was defined based on a distinctive light green color in core and in outcrop produced by elevated concentrations of alteration minerals (e.g., phyllosillicates and carbonates) above regional levels Townend et al., 2013;Toy et al., 2015;Boulton et al., 2017]. The degree of alteration increases with proximity to the PSZ within the fractured ultramylonites and cataclasites of the damage zone, implying that extensive fluid-rock interactions have occurred in the hanging wall of the Alpine Fault.

Sample Selection and Preparation
For this study, samples were collected from both the DFDP-1A and 1B drill cores. The drilled core diameter was 85 mm (PQ wireline bit size) and supplied for sample preparation as 200-300 mm length intact core rounds and half rounds. These core sections were axially oriented to the borehole axis, however, due to core rotation during retrieval the azimuthal orientation, could not be preserved [Sutherland et al., 2011;Toy et al., 2015]. The lithological character of the recovered core has been described in detail by Toy et al. [2015]. The core (and consequently the fault zone) was classified into a number of lithological units, based on the original protolith and level of damage/alteration. These unit classifications are outlined in Figure 2. Seven hanging wall cataclasites (Units 3 and 4), two footwall cataclasites (Units 6a and 6b), and the PSZ gouge were sampled in this study, representing a section across the plate boundary (Unit 5). For comparison, lithologies classified as potential protoliths for the Alpine Fault cataclasites were collected from the field and tested. These include an Alpine Schist, a foliated quartzofeldspathic schist, and a quartzofeldspathic ultramylonite all collected from the fault's hanging wall exposures in the Southern Alps foothills close to the drill site. For more details on these lithologies, see Cooper [1972] and Little et al. [2015], respectively. Most hanging wall cataclasites used in this study were sourced from the alteration zone (see Sutherland et al., 2012), a zone of heavily altered cataclasites rich in carbonate precipitation and authigenic phyllosilicates (illite/muscovite and chlorite). Figure 2 depicts selected optical micrographs of the tested material and a schematic outlining core sampling intervals.
Due to the friable nature of many of the test specimens, traditional methods of sample preparation such as water-lubricated coring, sawing, and grinding proved catastrophic, causing washout of the finer grain-size fractions and often total disaggregation. We therefore developed new, reliable preparation techniques to produce systematically intact samples suitable for experimental investigation. The aim was to produce cubic samples that could be used for measurements of anisotropy of both permeability and ultrasonic wave velocity.
Initially, the core was wrapped in thin plastic wrap followed by the application of three to four layers of ModRoc ® (gypsum plaster bandages). The original core orientation was recorded throughout the wrapping process, and once fully hardened, it was scribed onto the final plaster covering. Cutting the material was accomplished through the use of an annular saw (Revolution XT Ring Saw; Figure 3). The configuration of the annular blade means that only the actively cutting portion of the blade is in contact with the sample. This reduces grain plucking and disaggregation. Some cooling and lubricating fluid (water in this case) is required during cutting; this was applied by gently pressing a moist sponge to the blade edge.
After each cut, the exposed face was again covered by plastic wrap and ModRoc ® following the above procedure and repeated until a~15 × 15 × 15 mm cube was produced. The final cube is orientated to the vertical borehole axis, with two orientations orthogonal to the core cut according to any microstructural fabric if present, i.e., foliations, calcite veins, and fractures. A final step, if the specimen proved sufficiently robust, involved hand grinding the cube surfaces with dry, fine emery paper to provide a smooth flat surface to maximize acoustic wave and fluid transmission from the sample assembly to the sample cube. Once preparation was complete, the specimen, porous steel spacers, and solid steel spacers were jacketed in heat-shrink polyolefin in order to secure the sample. This was then jacketed in PVC tubing, which forms a physical barrier between the sample and its pore fluid, and the silicon oil confining medium ( Figure 4).
Once measurements were concluded, the experimental specimens were prepared for microstructural analysis, with their orientations preserved, using standard petrographic thin section techniques if the samples were sufficiently robust. Friable material was prepared by using "water-free" polishing techniques by Paul Hands of Hand On Thin Sections Ltd. As the DFDP core is not horizontally oriented, the relative orientation of the observed microstructures cannot be conclusively related to the fault orientation.

Experimental Procedure
The triaxial experimental apparatus used to undertake permeability and seismic velocity measurements is illustrated schematically in Figure 4 [Faulkner and Armitage, 2013]. During each experiment, the sample assembly was loaded into the pressure vessel, confining pressure (P c ) of 10 MPa was applied using a low-viscosity silicon oil confining medium, and finally, the specimen was allowed to saturate with deionized water pore fluid, which was allowed to equilibrate at a pore fluid pressure (P f ) of 5 MPa and an effective confining pressure (P e ) of 5 MPa, assuming a simple effective pressure law where the poroelastic constant α is equal to unity [Rice and Cleary, 1976]. Confining pressure and pore fluid pressure are servo-controlled to a resolution of ≤0.1 MPa. The seismic sample assembly with piezoelectric ceramics and pore fluid supply pipes set into the upstream and downstream platens. (c) A cube-shaped specimen set between porous steel spacers (both square and circle) and a circle to square adapter to allow coupling with the circular assembly platens.

10.1002/2017JB014355
Since each sample was recovered and tested again 3 times in three orthogonal orientations (x, y, and z, see Figure 2i), no measurements were recorded during the first pressure cycle each sample experienced. A pressure cycle consisted of increasing the confining pressure through the range 10, 20, 35, 45, 60, 70, 85, 95, and 110 MPa, while holding the P f at 5 MPa. These effective pressures (P eff = 5,15,30,40,55,65,80,90,and 105 MPa) are approximately equivalent to pressure conditions spanning the near surface to~5-6 km depth. This method means that, during an experiment, the samples were overconsolidated with respect to the near-surface conditions, removing the effects of sample preparation, storage, transport, and the effect of unloading during retrieval between tests on the same sample, making comparison with borehole measurements difficult. Another consideration to be made is the evolution of the fault rocks during the exhumation of the Alpine Fault. As mentioned above, the Alpine Fault is a mature fault that has undergone extensive and cyclical comminution and geochemical alteration. Consequently, the specimens tested are likely to be substantially different structurally and compositionally to fault rocks at depths of~5-6 km. Nonetheless, the results provide insight and an approximation of the permeability and seismic velocity of the Alpine Fault as it is exhumed from depth.
Piezoelectric ceramic transducers (piezoelectric lead zirconate titanate 5H, fundamental frequency at 1.5 MHz), with P wave and orthogonal S wave polarization directions, are housed in each opposing platen and are used to produce and record ultrasonic P and S waves through the samples. The ultrasonic velocities are measured using the through-transmission method, utilizing the travel times and specimen length [Birch, 1960] after correction for the travel time through the loading platens. Elastic wave speeds were picked and velocities calculated over an effective pressure range of 55-105 MPa for S waves (V s ) and 15-105 MPa for P waves (V p ).
Poor coupling between the loading platens and the test specimen at low effective pressures (<55 MPa for V s and <15 MPa for V p ) made reliable picking of arrival times difficult, and so these data are not reported.
Permeability measurements were performed using the pulse transient technique [Brace et al., 1968], involving the introduction of small pressure difference of <1 MPa to one end of the specimen and observing the pressure decay through the sample with time. One effective pressure cycle consisted of measuring the permeability, V p and V s at each effective pressure as the confining pressure is increased, with the unloading effective pressure path covering the same increments during measurement. A typical pressure cycle lasted between 3 and 7 days owing to the low permeability of some samples. Once one pressure cycle was completed for a given cube orientation, the specimen was removed from its jacket and reoriented and the above experimental procedure was repeated.
In a typical experiment, permeability and seismic velocities do not fully recover after deconfinement from the initial up-pressure leg, with upleg and downleg variations typically an order of magnitude for permeability and between 2 and 30% for seismic velocities; see Figure S1 in the supporting information for an illustration of this hysteresis. Subsequent variations after repeated pressure cycles become progressively smaller. Similar procedures and observations are reported by Bernabe [1987], Faulkner and Rutter [2000], and Morrow et al.
[2014]. The drying and unloading of samples at room conditions were minimized by conducting experiments for each sample orientation in immediate succession or by sealing the sample in heat shrink or plastic film where this rapidity was not possible.
As these samples have undergone compaction and consolidation up to pressures of 105 MPa, the measurements do not correlate with borehole wireline and slug tests but rather reflect greater depths on the Alpine Fault. We refer to Carpenter et al. [2014] for experimental results performed at in situ borehole conditions.

Permeability Measurements
Permeability results are plotted as log permeability against effective pressure and show a systematic decrease with increasing P eff , generally decreasing by larger increments over the lower P eff range, 5-40 MPa, presumably as handling/drilling-induced fractures close. This initial closure can account for a drop in permeability of up to 2 orders of magnitude (for example, the permeability trace for 1A-83.2 in Figure 5c); however, this effect more commonly manifests itself as a drop of 1 order of magnitude. Once above~30 MPa P eff , the rate of permeability reduction with effective pressure, in most cases, lessens with increasing pressure.
Permeability variation and an indication of the permeability anisotropy throughout the boreholes are illustrated in Figure 6, showing the range of permeabilities of each orientation with variation in P eff . With increasing proximity to the PSZ, permeability decreases. This trend is particularly noticeable in Figure 6a within the hanging wall alteration zone. This same feature is not as easily observed in Figure 6d, within the DFDP-1B core, as sparser sampling density makes it more difficult to highlight the trend, particularly close to the PSZ. Within the DFDP-1B alteration zone, there is a decrease in permeability with depth, and the lowest permeability value of 1.25 × 10 À21 m 2 was measured immediately above the PSZ. With increased distance from the PSZ, into the footwall cataclasites (e.g., 1B-142), permeability increases by up to 2 orders of magnitude from~10 À20 to~10 À18 m 2 , respectively (see Figure 6d). Permeability measurements for the "protolith" lithologies could not be performed as their permeability was too low for the resolution of the experimental apparatus, at~10 À22 m 2 . At permeabilities lower than~10 À21 m 2 , flow rates are very small and pores are inferred to be approaching molecular size. The permeability anisotropy exhibited by borehole specimens appears to have little correlation to orientation; this is conceivably due to the highly variable microstructures present in these fractured or comminuted lithologies. Direct correlation of permeability with specific microstructural causes is difficult to make, except in the case of 1A-90.67, the clay-rich PSZ gouge. This specimen exhibits the lowest permeability anisotropy at 40%, potentially due to its finer grained and more homogenous structure in comparison to other fault core lithologies.

Seismic Velocity Measurements
Elastic wave transmission through the specimens shows a similar relation to increasing P eff as permeability, with a rapid increase in velocity over initial pressure increments (e.g., 5 to 40 MPa) stabilizing once open fractures and porosity are reduced at high P eff (~40-50 MPa). Both V p and V s show similar trends across the fault; the material exhibiting the highest V p and V s is on the periphery of the pervasive alteration zone, farthest from the PSZ (1B-108.2 and 1B-120.1), with V p of~5700 m/s and V s of~4000 m/s (see Figures 6b-6f).
Both V p and V s decrease with increased proximity to the PSZ with a substantial reduction in velocity within the PSZ at 1A-90.67. Across the respective pressure ranges, hanging wall V p values span 4500-5900 m/s with V s values between 2400 and 4000 m/s. Material from the footwall exhibits lower velocities than the hanging wall with V p values ranging from 4400 to 5600 m/s and V s values from 2200 to 3500 m/s. PSZ material exhibits the lowest V p from 3900 to 4200 m/s and V s at 1900 to 2000 m/s. Figures 6g and 6h show seismic velocity measurements for the "protolith" lithologies, with the Alpine Schist V p from 4800 to 6700 m/s and V s at 3400 to 4250 m/s, while the ultramylonite had V p of 5600 to 6170 m/s and V s 3600 to 4030 m/s. Performing measurements parallel to the down core axis and in two further mutually orthogonal orientations allowed the seismic velocity anisotropy present in these rocks to be investigated and approximated. As only three orientations were measured in this study and these directions are not necessarily oriented parallel to the principal axes of the dynamic elasticity tensor, our measurements provide only a minimum constraint on anisotropy. Figure 7 illustrates the magnitude of this anisotropy with depth at 105 MPa, with the anisotropy calculated using Equation (1) following Mainprice and Humbert [1994], where V n max and V n min are the maximum and minimum seismic velocities at a given P eff increment with the subscript n denoting the wave type, while V n is the averaged velocity (0.5 (V n max + V n min) and AV n is the value of anisotropy in percent.
The degree of anisotropy present in the test specimens appears far greater for permeability than for seismic wave velocity. Presumably, this is largely due to the maintenance of interconnected voids and fractures for Each row denotes the orientation of the specimen. D-down is equivalent to the down core axis, while X-down and Y-down represent two horizontal directions in the borehole; note that the X and Y directions do not correlate between specimens due to the inability to horizontally orient samples from the drill core.
Journal of Geophysical Research: Solid Earth 10.1002/2017JB014355 Figure 6. (a-i) Permeability and seismic velocity ranges for material from DFDP-1A and DFDP-1B with depth, including field derived protolith lithologies (Figures 6g and 6h). Each tested specimen is represented by a bar color coded to orientation (see Figure 6i); note that the X and Y directions do not correlate between specimens due to the inability to horizontally orient samples from the drill core. fluid mobility, which will be greatly affected by increasing P eff , while the propagation of elastic waves is not limited by fracture/void connectivity in the same way. Similar observations were made by Fortin et al. [2011] when hydrostatically loading fractured basalt. We must also state that the anisotropies measured in this study, due to the arbitrary horizontal directions of the test specimens, are a minimum value of transverse anisotropy.
In most specimens, most notably in 1A-83.2 and 1B-142.2, V s1 anisotropy is much greater than that found for V p , potentially due to attenuation as an S wave pulse will be more sensitive to scattering of wave energy via impingement on an inhomogeneity, potentially a pore, fracture, foliation, or lithological interface [Sato et al., 2012]. This may explain the 33 and 17% difference in the anisotropy in 1A-83.2 and 1B-142.2, respectively. 1A-83.2 contains a penetrative foliation that may allow for easier transmission of P waves, while the S waves are more inhibited by the aforementioned factors, particularly pores and fractures. However, 1B-142.2 has no obvious mineral or fracture alignments, implying that the heterogeneity in clast size, shape, and orientation may also result in a greater variation of V s with orientation than with V p . Saturation of the fractures present may also be an important influence on seismic wave velocity, particularly in regard to the anisotropy observed. In this study we can assume that the fractures are fully saturated, explaining the greater anisotropy seen in Vs than in Vp; if anisotropy was greater in Vp, this would imply that fractures remained dry throughout the experiments [Guéguen and Sarout, 2011]. Figure 2 shows the mm-to cm-scale microstructures that govern the variation of physical properties throughout the Alpine Fault damage zone in both the hanging wall and footwall. Gouge-filled fractures are common throughout the alteration zone, generally manifesting as low-displacement shears filled with unconsolidated gouge material of the same composition as adjacent wall rock with little authigenic clay, appearing as angular to subangular shaped clasts ranging from sub-μm to 100 μm diameter (Figure 8c). Occasionally, elongated clasts show a preferred orientation parallel to the vein walls ( Figure 8e). These fractures are typified by their lack of carbonate cementation or phyllosilicate alteration and assumingly belong to a relatively recent episode of deformation. Fracture density through the Alpine Fault's damage/alteration zone was quantified by Williams et al. [2016] using CT scans of the DFDP-1 core material, reporting that damage zone structures (including open, healed, and semihealed fractures) do not increase in density with proximity to the PSZ, remaining largely uniform throughout the cored interval of the fault. The distribution of these fractures and the presence of authigenic clay minerals within them are the key controlling factors governing the permeability and seismic velocity variations of these rocks. NanoCT scans of specimens from this study are included in Videos S1 and S2 in the supporting information; these images highlight the presence of throughgoing open fractures with little evidence of shear, particularly visible in 1B-120.1 and fractures that contain gouge material visible in 1B-133.3. Cataclasites in the hanging wall are generated from an amphibolite-facies quartzofeldspathic or metabasic mylonitic protolith and are typified by a pervasive carbonate infill, as both cement and vein precipitation, composing of~5% of the bulk mineralogy. Veining is evidently multigenerational, exhibiting cross-cutting relationships, varying orientations (Figure 2), and evidence of reactivation shown by crack-seal structures. Foliation, when present, is typified by alternating phyllosilicates or opaque oxides anastomosing around sub-mm relict protolith quartz fragments.

Specimen Microstructural Variation
In contrast, the footwall cataclasites contain far less carbonate veins and, as a result, are far more incohesive and generally exhibit a lighter grey-white color owing to their prehnite-pumpellyite to greenschistfacies granitic protolith . Foliation is similarly variable in pervasiveness, from distinct as in 1B-133.3 (Figure 2e) with obvious alternating quartz/feldspar and phyllosilicate fabric, to completely absent as in 1B-142 (Figure 2f). Foliation is seen to be highly variable in orientation on the sub-cm scale.
In contrast with both cataclasites, the PSZ ultracomminuted gouge has a finer grain size (<10 μm) and consists of two main gouge types that occur in discrete layers~30-50 mm thick: a brown smectitic gouge and a blue chloritic/micaceous gouge, described in detail by Boulton et al. [2014], the latter of which was tested in this study.
Overall, the mineralogy across the fault, in both the hanging wall and footwall, remains largely constant. X-ray diffraction results from Boulton et al. [2017] show that calcite, precipitated as cements and veins, vary in proportion more than any other mineral phase, comprising~5% of the modal mineralogy in the hanging wall, decreasing sharply across the PSZ to~1.5% in the footwall. Phyllosilicates such as chlorite and

Discussion
The temporal and spatial variations of permeability and seismic velocity within fault zones are thought to be largely influenced by the variation of fracture density with respect to the fault zone [Faulkner et al., 2003;Faulkner, 2008, 2012]. This picture is complicated when lithological variation occurs with the fault zone [e.g., Wibberley and Shimamoto, 2002] and also when mineralization occurs within both the damage zone and fault core. Mineralization can occur in two ways: as authigenic mineral growth that produces clay mineral formation and as precipitation that may, respectively, heal and seal the fault [Madariaga et al., 1998;Tenthorey and Cox, 2006]. In this section, we consider how mineralogical and microstructural characteristics influence the fault zone physical properties, how the Alpine Fault Zone compares with other mature fault zones, and finally, how mineralogy and microstructure influence earthquake processes such as thermal pressurization and the propagation of seismic waves.

Microstructural Controls on Physical Properties
Other than a few discrete features such as the PSZ and pseudotachylyte veins, the bulk mineralogy remains largely consistent across the fault zone Boulton et al., 2017]. Hence, the bulk mineralogy does not appear to be a major factor controlling the observed variations in physical properties. Consequently, the fracture density might be considered to be the primary influence on the physical properties at the scale of the experimental measurements. As fractures make up proportionally far less of the bulk rock than the matrix, the small variations of mineralogy associated with fractures, primarily carbonate and phyllosilicate fill, are masked by the bulk rock mineralogy in regard to seismic velocities. However, small variations in the overall mineralogy, due to fracture infill, could significantly modify the fluid transport properties of the fractures and hence those of the bulk rock. Williams et al. [2016], using combined CT and SEM analyses of the DFDP core, to a fracture aperture resolution of >1 mm, discovered that, within the alteration zone, the majority of the fault damage is cemented with varying carbonate, clay, and gouge materials (72% being fully sealed and an additional 24% remaining partially sealed). Outside of the alteration zone (<50 m away from the PSZ), fault damage may persist between periods of seismicity, as evidenced by the high in situ permeability measured during wireline slug tests and the lack of intact core recovery outside of the alteration zone due to disaggregation . Fracture porosity is likely to be the main control of fluid flow in these damage zone ultramylonites and cataclasites, as the intact protolith or cemented regions have very low permeabilities, below the capabilities of measurement of the experimental apparatus detailed above, at <10 À22 m 2 as measured in the protolith lithologies in this study. The length scales of the fractures within a rock mass have great influence on permeability and seismic wave velocities; this scale effect is detailed in Brace et al. [1968] and Nara et al. [2011] regarding permeability and Budiansky and O'connell [1976] regarding seismic velocities. Samples prepared for this study were actively chosen away from the macroscopic fractures described by Williams et al. [2016] in order to avoid sample failure during preparation. With this in mind, the measurements reported here represent the intact matrix properties of the fault lithologies, giving the minimum permeability and maximum seismic velocity values of the Alpine Fault due to the fine-scale sampling bias inherent in this study, where heterogeneity of microstructure is a key characteristic of all fault rocks analyzed. In the following section, the nature of the fracture networks and their fillings present in the experimental samples are explored.
Open fractures within core material are predominantly opened along preexisting gouge-filled or carbonatefilled fractures, as visible in Figures 8c and 8e. These are presumably stress-relief fractures, opened during core retrieval, handling and resting prior to testing and microstructural analysis preparation. This is supported through the lack of displacement on the fracture or mobilization of the fracture fill, indicating that the fractures have opened up by dilation in "dry" conditions. As illustrated in section 3.1, there is a significant initial drop in permeability and increase in seismic velocity after an initial pressure interval between 5 and 40 MPa, after which trends in the data remain largely constant, marking the closure of open fractures and voids and allowing a more representative measurement for the specimen to be taken [Nara et al., 2011].
Immediately after a seismic event, it is possible that the permeability within the fault damage zone is enhanced by 3-8 orders of magnitude [Miller et al., 2004;Mitchell and Faulkner, 2008;Gomila et al., 2016]. This transiently enhanced fluid mobility coupled with reduced grain size via cataclasis can result in a period of heightened fluid-rock interaction. Warr and Cox [2001] determined two generations of phyllosillicates on the Alpine Fault, hydrous chloritization at~320°C, followed by the generation of swelling clays at shallower depths and temperatures of~120°C. The dominant clay phases within the Alpine fault hanging wall are white mica (illite/muscovite) and chlorite [Boulton et al., 2012[Boulton et al., , 2017Schleicher et al., 2015]. Schleicher et al. [2015] found that enhanced levels of authigenic illite and smectite are present in the PSZ in contrast to higher levels of metamorphic illite/muscovite in surrounding cataclasites. White mica and chlorite fill a large proportion of fractures in the hanging wall, while smectite is found solely in the brown PSZ fault gouges, generated from the low-temperature alteration of illite/muscovite and chlorite, which is further altered into kaolinite and goethite [Craw, 1984;Boulton et al., 2012Boulton et al., , 2017Schleicher et al., 2015], with the combination of ultracomminuted grains and swelling clays on the PSZ reducing permeability to~10 À20 to 10 À21 m 2 , becoming a hydraulic seal.

10.1002/2017JB014355
The nature of carbonate precipitation is variable, occurring as veins both parallel and cross-cutting foliation, bounding gouge-filled fractures, disseminated cement in gouge-filled fractures and fragmented clasts within gouge and cataclasite. Hanging wall cataclasites contain a pervasive carbonate infill and, as a result, are more cohesive than the carbonate-poor footwall cataclasites, resulting in the increase in permeability with distance from the PSZ into the footwall (Figure 6d).The presence of calcite networks indicates periods of damageenhanced fluid transport focused within the hanging wall cataclasites and ultramylonites. Mineralization effectively "seals" the newly formed postseismic fracture networks, resulting in an increase in strength and stiffness relative to its immediate postrupture state and lowering of permeability of the fault zone lithologies [Lund Snee et al., 2014]. Similar processes are inferred to take place in geothermal fields, with mineral precipitation reducing permeability potentially on the time scale of days for mm-scale fractures in high temperature settings, especially if considerable fluid pressure fluctuation occurs [Griffiths et al., 2016;McNamara et al., 2016].
A main control on the anisotropy of physical properties is the presence or lack thereof of a tectonic foliation [Okaya et al., 1995;Okaya and Christensen, 2002]. For instance, the poorly developed fabric in the PSZ gouge exhibits the least anisotropy, while distinctly foliated specimens such as 1A-83.2 exhibit a strong anisotropy, with higher seismic wave velocity and permeability parallel to foliation and slower values perpendicular to the foliation (Figures 6 and 7). However, the nature of the cataclastic fabrics within the alteration/damage zone is spatially variable, with pods of intensely foliated material (1B-133.3) located within wide swath of unfoliated material (1B-142) or vice versa (Figures 2e and 2f and 8e and 8f). With this in mind, the experimental results from this study place a lower limit on matrix permeability and an upper limit on matrix seismic wave velocity, respectively, through the Alpine Fault. Overall, both permeability and seismic velocity anisotropy show a general trend to decreasing toward the principal slip zones, in both DFDP-1A and 1B.

Correlation With Other Studies
Figures 6b and 6e show the DFDP-1 wireline logs as a black trace in comparison to the results presented here. As each test specimen was precompacted to 105 MPa P eff , it is not possible to make a direct comparison to measurements made at pressure conditions within the DFDP-1 borehole. Furthermore, the sampling length scale for the P wave velocity wireline tool (a single source, dual receiver logging sonde as documented in Townend et al. [2013]) used to collect V p data is 304.8 mm, approximately 20 times the length of the experimental specimens used in this study. The selection of samples and subsequent sample preparation means that measurements were made on coherent core material that was perhaps more intact than the in situ material. Large, relative to specimen size, throughgoing fractures may not have been sampled in the experiments, and hence, measurements provide a minimum value of permeability and a maximum value of seismic velocity. However, the trends of the data show a reasonable comparison, indicating that the measurements made here may be representative of the permeability and seismic velocity at greater depths, equivalent to the effective pressures used in the experiments. Slug tests performed in the distal damage zone of DFDP-1 recorded in situ permeability of 10 À14 m 2 , indicating that permeability may drop by up to 5 orders of magnitude at low effective pressures toward the fault core . Carpenter et al. [2014], illustrated in Figures 6b and 6e, presented similar laboratory measurements of DFDP-1 fault rock permeability and seismic velocity, but at a lower P eff range (2.5-65 MPa) with no preconsolidation. The permeability data presented in this study correlate with Carpenter et al.'s [2014] data, with measurement ranges overlapping at similar depths in the core. In terms of seismic velocity, the data presented here are higher than those in Carpenter et al. [2014]. This is most likely due to preconsolidation of samples in the method used here, and therefore, the values for seismic velocity are more representative of those at greater depth than the data of Carpenter et al. [2014]. Other factors that might produce differences are the variation in specimen dimensions, frequency, and natural specimen variation.
The overall variations in seismic velocities with depth in this study and that of Carpenter et al. [2014] correlate well with each other (Figures 6b and 6e). The PSZ exhibits lower velocities than the surrounding cataclasites, and the higher velocities witnessed in the hanging wall cataclasites in contrast to the footwall are indicative of a greater degree of carbonate cementation and cohesiveness and potentially asymmetrical damage imparted upon the fault zone during seismicity [Townend et al., 2013].

10.1002/2017JB014355
Physical property measurements of the Alpine Fault Zone are consistent with a number of mature, geochemically altered fault zones worldwide. Similarities with studies performed on the Nojima Fault Zone [Mizoguchi et al., 2008], Median Tectonic Line (MTL) [Wibberley and Shimamoto, 2002], the San Andreas Fault [Morrow et al., 2014], and the Carboneras Fault [Faulkner et al., 2003] highlight the microstructural complexity within a fault damage zone, with numerous episodes of brittle rupture damage and subsequent fluid-rock interactions. Laboratory gas permeability measurements of the MTL fault core show values between 10 À13 and 10 À21 m 2 (P eff = 50-180 MPa) that are attributed to enhanced grain size reduction with increasing proximity to the PSZ [Wibberley and Shimamoto, 2002]. Material reclaimed from depth on the actively creeping segment of the San Andreas Fault shows a similar range of permeabilities of 10 À18 to 10 À23 m 2 (P eff = 10-120 MPa), with gouge material exhibiting lower permeabilities than bounding cataclasites [Morrow et al., 2014]. The gouge zones of the Carboneras Fault show permeability values in the range of 10 À17 to 10 À22 m 2 , with permeability anisotropy of up to 3 orders of magnitude [Faulkner and Rutter, 2000;Faulkner et al., 2003].
The permeability variations of up to 6 orders of magnitude in previous studies seem typical of major, mature fault zones and are related directly to degree of comminution and associated authigenic mineral generation and cementation. Both experimental and in situ seismic velocity measurements have revealed that faults, and their associated damage zones are regions of low seismic velocities [Mooney and Ginzburg, 1986;Li et al., 2004;Jeppson et al., 2010;Jeppson and Tobin, 2015], which concur with velocity measurements in this study.

Implications for Fault Zone Properties During the Seismic Cycle
The hydraulic properties of fault rock have implications for transient variations in pore fluid behavior during seismic slip. Coseismic thermal effects may promote slip propagation via processes such as thermal pressurization [Rempel and Rice, 2006;Rice, 2006]. The low permeability of the core of the Alpine Fault and alteration zone in the hanging wall may favor thermal pressurization during earthquake rupture [Segall and Rice, 1995;Wibberley, 2002;Mizoguchi et al., 2008;Leclère et al., 2015]. Evidence for the development of overpressures within the damage/alteration zone includes gouge-filled fractures emanating from, and propagating through, the PSZ. These features indicate that dilatancy-driven pore depressurization may also stabilize slip and arrest rupture [Segall and Rice, 1995;Samuelson et al., 2009].
Geochemical studies of vein mineralization [Menzies et al., 2014] and fluids from hot springs from the Alpine Fault hanging wall [Barnes et al., 1978] show that much of the fluid circulating at <2 km depth are meteorically derived. Fluid inclusion measurements indicate that water infiltration is at least 2 orders of magnitude greater than metamorphic water production [Menzies et al., 2016]. In its low permeability prerupture state the Alpine fault appears to act as a barrier to fluid flow, isolating fluid geochemical regimes on a regional scale. Similarities can be drawn with low fluid 3 He/ 4 He ratios in San Andreas Fault fluids at seismogenic depths, indicating a dominantly meteorically derived source. As the overall permeability structure across the creeping sections of the San Andreas is low, it has been suggested that the proportion of mantle-derived fluids will increase within more permeable faults at distance from the active San Andreas strands within the North American Plate [Wiersberg and Erzinger, 2007].
The strong, penetrative tectonic and metamorphic foliations present in the protolith host-rock bounding the Alpine Fault, the hanging wall Alpine Schist and mylonite sequence lithologies, have a strong regional seismic anisotropy [Christensen and Okaya, 2007]. This anisotropy extends to depths of <90 km into the deep lithosphere and is broadly distributed, according to shear wave splitting studies [Karalliyadda and Savage, 2013]. Fault Zone Guided Waves (FZGWs) have been observed on the Alpine Fault, in which seismic waves are channeled and propagated along the fault zone [Eccles et al., 2015]. These FZGWs occur due to material contrast between the fault zone material and the intact country rock, causing seismic waves to focus along low-velocity zones. Our observations of fault zone velocities in comparison to the intact country rock are consistent with the presence of these FZGWs. These types of seismic wave allow the fault to be imaged and some of the properties of the fault at depth, such as fault zone width, to be inferred [Li et al., 1997;Ben-Zion, 1998]. The use of FZGWs is critically dependent on the velocity contrast between the fault and the country rock and the orientation of any anisotropic fabric in the country rock. If the slow direction of any anisotropy is oriented perpendicular to the low-velocity fault zone, seismic energy may escape the low-velocity zone, making the fault difficult to recognize through FZGWs [Kelly et al., 2017]. The seismic velocities and anisotropies reported in this study support previous regional observations on the behavior Journal of Geophysical Research: Solid Earth 10.1002/2017JB014355 of seismic waves on the Alpine Fault, where the fault core material has lower seismic velocities than the fastest direction in the country rock hosting it and the heterogeneity of microstructures within the fault zone serves to homogenize any anisotropy [Eccles et al., 2015].

Conclusion
Our suite of experimental measurements shows a zone of low permeability and low seismic velocity material bounding the PSZ of the Central Alpine Fault, under conditions approximating depths of 5-6 km.
Measurements performed on protolith lithologies show lower permeabilities and higher seismic velocities than the fault-damaged rock owing to their intact, unfractured nature. Low permeabilities at low experimental pressures are explained by the presence of fine grained alteration products and carbonate precipitation within the fracture networks of the fault hanging wall. The PSZ, being composed of ultra-fine grained alteration products, acts as a barrier to cross-fault fluid flow postseismicity. This is in agreement with other petrophysical studies on other mature faults subjected to extensive comminution and fluid-rock alteration [Wibberley and Shimamoto, 2002;Faulkner et al., 2003;Boulton et al., 2012;Townend et al., 2013;Carpenter et al., 2014;Morrow et al., 2014].
The evolution of fault rock microstructure with seismic shear and subsequent interseismic fluid-rock interaction has a great effect on the hydraulic and elastic properties of fault zones. This evolution is complex and cyclical, potentially promoting rupture via dynamic coseismic processes such as thermal pressurization and dilatancy hardening. During interseismic periods, fluid flow is likely concentrated upon pervasive, unconsolidated gouge-filled fracture networks that are progressively mineralized/altered, reducing along-fault communication of pore fluid. With fluid focusing along these fracture networks, fault rock fabrics have little influence on permeability anisotropy. In addition, as the cataclastic microstructures are so variable on a mm-cm scale, their overall effect within the fault zone would be to homogenize seismic velocities, limiting seismic anisotropy. The main microstructural controls on fault rock physical properties are throughgoing fractures, both gouge-filled and mineralized, which have a great effect on seismic wave speeds and fluid flow at the experimental specimen scale and likely to act as preferential fluid flow paths and planes of discontinuity in the Alpine Fault Zone as a whole.