We conducted permeability measurements on rocks sampled from an exhumed ancient subduction zone fault in the Cretaceous Shimanto accretionary complex of Japan. The permeability under seismogenic environment conditions shows a heterogeneous structure across the fault zone. Permeability of the sandstone in turbidite sequence is the lowest, whereas the highly sheared zone has the largest permeability. A combination of permeability contrast between these two zones and fluid pressure source, including tectonically driven burial and dehydration, could result in a concentration of pore fluids along the fault, which causes Deep Seismic Reflections during underplating stage as delineated in seismic reflection studies. Following a rise of temperature up to 250°C, the permeability of all the sampled rock types became smaller than at room temperature, and also gradually decreased with increasing hold time.
 Investigating the fluid transport properties of subduction zone faults is crucial for understanding the shear strength and slip-stability, or instability, of subduction zone faults. Models invoking fluid pressure controlled fault weakening and slip instability require data on the permeability structure of the fault zone [e.g., Mase and Smith, 1987; Moore and Saffer, 2001]. Brittle fault zones are lithologically heterogeneous and structurally anisotropic, and may act as fluid conduits or barriers that enhance or impede fluid flow [e.g., Caine et al., 1996]. Permeability data for fault gouge from the San Andreas fault, the Carboneras fault in Southern Spain, and the Nojima fault in Japan, show that the permeability of the protolith becomes significantly enhanced in a damaged zone but decreases into the core zone of fault gouge [e.g., Faulkner and Rutter, 2001]. Previous studies, however, did not focus on the permeability structure around the subduction zone fault. Despite the influence of pore pressure on a wide range of subduction-zone fault processes [e.g., Saffer and Bekins, 1998], few previous studies have evaluated the permeability structure around the subduction-zone fault placed in a well-defined structural context using experimental methods. In this paper, we aim to determine the entire permeability structure and fluid flow properties around an ancient subduction zone fault in the Cretaceous Shimanto accretionary complex of Japan. We have conducted a series of permeability measurements under conditions appropriate to the seismogenic environment using rocks sampled from the outcrop of the exhumed fault.
2. Geological Setting of the Fault Zone and Sampling
 The Cretaceous Shimanto accretionary complex in Shikoku, Japan, is composed of sandstone dominant coherent unit (turbidite sequence) and deformed zone of mélange unit. The Okitsu Mélange includes scaly black shales enclosing disrupted pillow basalts, pelagic to hemipelagic red shales, and sandstone lenses. The Okitsu Mélange displays a systematic fabric indicating a top-to-the-south, sinistral-reverse sense of shear, which may have been related to underthrusting [Taira et al., 1980]. Distributed layers of basalt extend over more than several kilometers, parallel to the general trend of foliation of the Okitsu Mélange (Figure 1). They are significant markers that delineate geological structures that are characterized by duplex structures at a large scale. The pseudotachylyte-bearing fault is a roof thrust (studied area) above underthrusted Okitsu Mélange, which is overlain by an offscraped package of coherent unit (Nonokawa Formation) [Ikesawa et al., 2003]. The age difference between the Coniacian to Maastrichtian black shale and Cenomanian-Turonian pelagic red shale in the mélange (<30 m.y.) indicates subduction of a young oceanic plate in the Late Cretaceous period [Taira et al., 1980; Sakaguchi, 1996]. Thermal conditions around this area, which have been measured using vitrinite reflectance, show that the maximum past temperature of the Nonokawa Formation and the Okitsu Mélange ranged up to 230 ± 30°C and 270 ± 30°C, respectively. Further, the average thermal gradient is estimated to have been 50°C/km [Sakaguchi, 1996]. These results indicate that the burial depth of studied area was 4–5 km, and the fault zone would have been in the seismogenic zone during the underthrusting. It is also worth noting that the young crust and high thermal gradient in the ancient subduction system make it a good analog for the modern Nankai trench.
Figure 2a shows a detailed geological section of the studied area, which has an excellent exposure. White stars in Figure 2a denote the locations of samples used in the experiments: Black shale from the Okitsu Mélange, altered basalt, a foliated rock from the highly sheared zone, clastic sandstone, black shale and sandstone from the Nonokawa Formation. The rock in the highly sheared zone has a well-developed cataclastic S-C fabric with slickenlines on the C surfaces. Pseudotachylytes and ultracataclasites are observed along the C surfaces as dark veins less than few millimeters in thickness [Ikesawa et al., 2003]. Black shale in the Okitsu Mélange also has foliation surfaces; however, quartz (or calcite) veins are complexly developed and crosscut the foliation surfaces. The altered basalt includes mineral vein networks composed of quartz, calcite, ankerite, and chlorite. Since it has been suggested that the vein minerals around the fault zone have a close relationship with hydrothermal alteration or dynamic fault slip, we consider that the vein minerals in the sampled rocks were not formed as secondary minerals precipitated near the surface after exposure of the zone. The average porosity and grain density of each rock type are listed in Table 1. Note that the porosity of most rock types is less than 1% except for the highly sheared zone, some of which have high porosities. This indicates a heterogeneous distribution of porosity in the highly sheared zone. Cylindrical test specimens (length = 40 mm, diameter = 20 mm) were cored to an accuracy of within 0.02 mm. To drive away air trapped in interstitial pores of the rock samples, the samples were vacuum saturated with distilled water.
Table 1. List of Porosity, Grain Density and Pressure-Sensitivity of Permeability (−Δlogk/ΔPceff) for Each Rock Typea
Par.: Parallel to foliation, Ver.: Vertical to foliation.
Sandstone (coherent unit)
Black shale (coherent unit)
Highly sheared zone
Par., High porosity
Par., Low porosity
Black shale (mélange)
3. Experimental Methods
 All experiments measuring permeability were conducted in a triaxial pressure apparatus at the Earthquake Research Institute, University of Tokyo [e.g., Kato et al., 2003]. We upgraded the apparatus for measuring permeability. The cylindrical rock specimens were jacketed by silver sleeves. To ensure that there was no flow between the silver sleeve and the surface of the specimen, we conducted a test with a dummy specimen of stainless steel. We applied a pore pressure difference (3 MPa) between the top and bottom faces of the stainless steel specimen, and found that no pressure leak occurred during a period of half a day. The confining pressure and pore pressure were raised in steps to the desired test values, and servo-controlled to be constant. Temperature was raised at a constant rate of 3°C/s to a pre-set value. The flow of water through the specimen was driven by a fluid intensifier, and each permeability measurement was conducted after the displacement of fluid intensifier control reached a steady state.
 A transient pulse test [Brace et al., 1968] and an oscillating pore pressure technique [Fisher and Paterson, 1992] were applied to measure permeability in this study. We have demonstrated that these two different methods provide almost the same permeability values for the range 10−16 ∼ 10−20 m2, within experimental error. It is, thus, concluded that the permeability measured using the apparatus does not significantly depend on which method is applied. For the transient pulse test, our apparatus had two reservoirs with volumes of 60 cm3 and 90 cm3, which were evaluated at the pore pressure of 105 MPa by neglecting the elastic distortion of the system. The value of compressibility β was determined by measuring the decrease in the volume of fluid against the pore pressure step at room temperature and 250°C. It was clearly demonstrated that β at 250°C was only up to 1% larger than that at room temperature, which means that the effect of temperature on β was negligible. This is because a large amount of water was exposed under room temperatures, whereas a small amount of water was warmed by the internal heater in pressure vessel. The average compressibility of water at elevated temperature, thus, was almost the same as that at room temperature. A pore pressure differential of 2 MPa was imposed between the inlet and the outlet of the specimen as a pulse or with a sinusoidal oscillation of the amplitude. For the oscillating pore pressure method, the volume of the downstream pore fluid system (Vr) was estimated to be about 7500 mm3. The storage capacity of the downstream reservoir (Bd) was calculated from the relation Bd ∼ Vr × β. The uncertainty in the absolute permeability is around ±40 %. The lower limit of accuracy for our measuring system is about 5 × 10−21 m2 at room temperature. Note that the lower limit at 250°C is 7 × 10−22 m2, due to the decrease in viscosity of water.
 First, we investigated the dependence of permeability on the effective confining pressure Pceff at room temperature. Subsequently, the specimen was heated up to 250°C, then held under the conditions (Pc, Pp, T) = (140 MPa, 105 MPa, 250°C), simulating a depth of 4–5 km from the surface of the accretionary prism under suprahydrostatic pore water pressures, to measure the permeability evolution.
4. Results and Discussion
 The pressure-sensitivity of permeability, which is defined as −Δlogk/ΔPceff, was evaluated for each rock type at Pceff ranging from 15 MPa to 35 MPa as listed in Table 1. These data show that permeability decreased about one order of magnitude for a 50 MPa effective pressure increase for all rocks except for those in the highly sheared zone parallel to the foliation, which had a fifty percent greater sensitivity to pressure change. This indicates that the relative relation of permeability between each rock type is not strongly dependent on Pceff.
Figure 2b shows the permeability structure tested at room temperature across the fault zone. Most of permeability values were evaluated at the conditions (Pc, Pp) = (140 MPa, 115 MPa). One experiment using rock from the highly sheared zone was conducted at Pc = 50 MPa with the same Pceff of 25 MPa. Permeability tested at the conditions (Pc, Pp) = (140 MPa, 125 MPa) almost coincides with that at (Pc, Pp) = (35 MPa, 20 MPa), which demonstrates that the permeability tested in this study clearly obeys the effective stress law. Thus, Pceff is significant for permeability. In Figure 2b, permeability has a heterogeneous structure across the fault zone. Note that the permeability of the sandstone in Nonokawa Formation is the lowest (10−20 ∼ 10−19 m2). In contrast, permeability of the highly sheared zone with high porosity has the largest values, ranging from 10−15 to 10−16 m2. Low porosity rock from the highly sheared zone has a lower permeability than that with high porosity, and the foliation does not affect the permeability significantly more than the porosity does it. Although the black shale of the mélange seems to be anisotropic, the anisotropy is not strong because the foliation surfaces are curved, and quartz (or calcite) veins are complexly distributed. The black shale in the mélange, clastic sandstone and altered basalt in the fault zone have almost the same permeability values, and the black shale in Nonokawa Formation has slightly larger permeability than them.
 Permeability evolution was investigated through holding experiments conducted without differential axial stress. Permeability variation tested at the conditions (Pc, Pp, T) = (140 MPa, 105 MPa, 30°C) showed a slight decrease in permeability against hold time. Figure 3 plots the permeability evolution following a temperature increase, measured under the conditions (Pc, Pp, T) = (140 MPa, 105 MPa, 250°C). The permeability at room temperature, just before the increase in temperature was applied, is shown for each sample at time = 0 in this figure. Note that the permeability of all the rock types became lower at 250°C than at room temperature. Furthermore, it was found that the permeability at 250°C continuously decreased with hold time for all the rock types, and that the reduction rate decreased gradually with hold time. The reduction rate of permeability did not significantly depend on rock type, which suggests that the relative contrasts in permeability values around the fault zone would be held even at elevated temperatures, at least for short hold times (∼10 hours). This observation is of significant importance for investigating the fluid flow properties around faults, since most previous studies using natural rocks measured the permeability values at room temperature only, and incorporated them into a model for fluid flow. The decrease in permeability found at elevated temperature may be induced by the thermally activated processes such as mineral dissolution-precipitation sealing [e.g., Moore et al., 1994] or dislocation creep within grains [Zhang et al., 1994].
 One may question why the sandstone in coherent unit of Nonokawa Formation has a low permeability. Microscopic observations of the thin sections show that grains in the sandstone of coherent unit are poorly sorted. In short, very fine grains closely fill the spaces between large grains (∼1 mm). It is, thus, suggested that this closely packed structure of grains leads to a suppression of water flow in the spaces between the large grains, and that the permeability of the sandstone is low as a result.
 On the basis of seismic reflection surveys in the Nankai Trough, Park et al.  delineated reflections with negative polarities (Deep Strong Reflectors; DSRs) beneath the Nankai accretionary prism, 20–60 km landward of the frontal thrust. The DSRs are located deeper than the negative polarity décollement near the frontal thrust. Park et al.  interpreted the DSRs to indicate elevated fluid pressures. The fault zone in the present study is consistent with the duplex-model, and corresponds to the area where the décollement near the frontal thrust steps down (Figure 4). The present results show the permeability contrast between the sandstone in coherent unit and the highly sheared zone. A combination of the permeability contrast and the fluid pressure source, including tectonically driven burial and dehydration, would lead to a concentration of pore fluids along the roof thrust, which causes DSRs during underplating stage.
 Care should be taken as permeability values estimated in the laboratory show the minimum value at field scales, because permeability has a scale-dependency. Thus, the fracture systems not sampled at the core-scale may be dominant hydrologic features in situ conditions. We need to investigate the scale-dependency of fluid pathways in future study. One may argue that the precipitation of vein materials marks paleo-fluid flow pathways. Since they are now sealed, the present measurements do not evaluate the effect of them on permeability at in situ conditions. However, the intergranular permeability measured in this study would be an adequate analog for in situ conditions, if small-scale fractures were important in the real system.
 We conducted permeability measurements on rocks sampled around an exhumed ancient subduction zone fault. The permeability shows a heterogeneous structure across the fault zone. Permeability of the sandstone in turbidite sequence is the lowest, whereas the highly sheared zone has the largest permeability. Following a rise of temperature up to 250°C, the permeability of all the sampled rock types became smaller than at room temperature, and also gradually decreased with increasing hold time.
 We are very grateful to two anonymous reviewers for their constructive comments and helpful suggestions, which led to substantial improvement in the original manuscript.