Microfractures within the fault damage zone record the history of fault activity
Geosphere Science Sector, Civil Engineering Research Laboratory, Central Research Institute of Electric Power Industry (CRIEPI), Abiko, Japan
Corresponding author: K. Mizoguchi, Geosphere Science Sector, Civil Engineering Research Laboratory, Central Research Institute of Electric Power Industry (CRIEPI), Abiko 270-1194, Japan. (firstname.lastname@example.org)
 For faults without sedimentary covers, there is no robust method to obtain paleoseismic data that is crucial for the prediction of future damaging earthquake events. Sudden failure along faults during earthquakes induces off-fault damage surrounding such basement faults. We showed that the Quaternary-active fault has the damage zone characterized by a fracture density that decays exponentially with distance from the fault for both healed and open microfractures. In contrast, Quaternary-inactive faults contain only healed microfracture damage zone. Cross-cutting relationships between microfractures and a minimum healing temperature of ~100°C suggest that healed microfractures formed before, and at deeper levels, than did unhealed microfractures. The damage zone defined by open microfractures reflects the recent fault movement during exhumation, associated with erosion and regional uplift, from the maximum depth at which microfractures may remain unhealed. Microfracture analysis can therefore be used to examine the history of basement fault activity.
 Most earthquakes are caused by sudden failure along faults that may or may not be covered by layers of sediment. For faults that underlie sediments of known age, trenching of the fault scarp is commonly done to obtain abundant knowledge of paleoseismic records [e.g., McCalpin, 2009]. However, seismological investigations of faults found where basement rocks are exposed at the surface without sedimentary covers, as in mountainous regions, remain problematic, although several methods have been proposed [Kanaori et al., 1980; Ikeya et al., 1982; Kanaori, 1983; Fukuchi et al., 1986; Kralik et al., 1987; Zwingmann and Mancktelow, 2004].
 Brittle faulting along faults in basement rocks often results in the development of fault cores, composed of crushed rock fragments and fine-grained matrix (fault gouge, fault breccia, and cataclasite), and the surrounding fault damage zone that preserves the original host rock texture, but that is highly fractured from the microscale (~µm, microfracture) to the macroscale (~m, macrofracture) [e.g., Sibson, 1977; Wallace and Morris, 1986; Chester and Logan, 1986; Caine et al., 1996]. Previous studies of damage zone structure have focused on fault mechanics, examining damage zone formation and development in association with fault growth, and include studies of Coulomb failure in a homogeneous stress field, stress concentration at migrating fault tips, slip accumulation on a wavy fault plane, and linkage of isolated faults and dynamic stress at earthquake rupture tips [Vermilye and Scholz, 1998; Wilson et al., 2003; Rice et al., 2005; Childs et al., 2009; Mitchell and Faulkner, 2009; Savage and Brodsky, 2011]. The amount of damage (i.e., microfracture and macrofracture density) in a damage zone typically increases with proximity to the fault core. The damage zone thickness, defined by the distance from the fault at which fracture density decreases to the background level of the undamaged host rock, shows an increase with increasing net displacement along the fault [Mitchell and Faulkner, 2009; Savage and Brodsky, 2011]. Microfractures that form in the damage zone are generally classified as healed, sealed, or open fractures. The cross-cutting relationships between fractures can establish the sequence and timing of fracture formation [Wilson et al., 2003; Mitchell and Faulkner, 2009]. Here we performed a fracture analysis on different microfracture types within a damage zone. Each microfracture type formed during a different period of fault movement, indicating that the fractures in the damage zone record the history of fault activity.
2 Study Area
 We examined the spatial distribution of various types of microfractures within the damage zone surrounding a Quaternary-active fault in Tottori prefecture, southwest Japan [Sasaki et al., 2011]. The fault is located east of the epicenter of the 1943 Tottori earthquake (Moment magnitude Mw = 7.0, Average slip = 2.5 m, Recurrence interval = 4000–8000 years) (Figure 1) [Kanamori, 1972; Kaneda and Okada, 2002]. The fault cuts the Cretaceous Kyushozan granite and offsets the sedimentary layer containing the Daisen-Kurayoshi tephra (DKP, ca. 55 ka) [Machida and Arai, 1979], indicating that it has been active during the late Quaternary period [Sasaki et al., 2011].
 The basement fault zone consists of a fault core (about 75 cm wide) surrounded by a damage zone. The fault core is composed of purple-colored clayey gouge and granitic fault breccia. The contact between the gouge and breccia, which strikes N79°W and dips 87°N, is sharp and planar and is thought to be a fault plane. The fault gouge does not retain the structure of the host rock, and has a matrix-supported fabric, with a fine-grained clay-rich matrix surrounding angular to subangular grains of less than 0.5 mm, resulting from intense fragmentation and comminution during the localization of brittle deformation within the gouge zone and diagenetic alteration to clay minerals after deformation. Riedel shears developed in the gouge indicate right-lateral motion. The damage zone is composed of fractured Kyushozan granite in which macro- and microfractures are well developed, and the host rock textures are preserved. The Kyushozan granite is a fine-grained, porphyritic biotite-granite with >5 mm phenocrysts and whose major constituent minerals are quartz, alkali-feldspar, plagioclase, and biotite.
3 Microfracture Analysis
 For microfracture analysis, we took oriented samples for microfracture analysis, at 10 m to 180 m from the fault core within the damage zone which are well exposed in a quarry. Then, thin sections of the samples are made perpendicular to the fault plane and parallel to the slip direction, whose plane is suitable to observe fault-related microfractures [Engelder, 1974; Vermilye and Scholz, 1998; Wilson et al., 2003]. More than 40 quartz grains per sample were used in measurements of microfracture density. Quartz is suitable for estimating the damage that a rock sample has sustained because it has no weak crystallographic plane, such as cleavage, in contrast to feldspar and mica. Also, quartz is resistant to chemical weathering, which is commonly intense in fault zones. We marked each thin section with a square grid at 3 mm intervals. One grain from each grid square was selected for analysis to reduce operator sampling bias. We counted the number of microfractures that intersected a line drawn from one edge of the selected quartz grain, through its center point, to the other edge of the grain. The linear microfracture density (called “fracture density”) for each sample was calculated as the total number of microfractures intersecting the line, divided by the total length of the line.
 In the samples, we observed tensile microfracutres that consist primarily of intragranular or grain boundary fractures with minor transgranular fractures. We divided the microfractures into three types: (1) healed microfractures (HF) that contain aligned fluid inclusions (fluid inclusion planes); (2) sealed microfractures (SF) cemented with secondary minerals (iron oxides were observed in the fault studied); and (3) open microfracture (OF) that represent an opening between opposing fracture walls and lack cementation (Figure 2). Microscopic cross-cutting relationships between microfractures show that healed and sealed fractures are always cut by open fractures (Figure 2d). This indicates that open fractures are younger than other fractures. The measured microfractures occur in the following proportions: 81.5% HF, 1.2% SF, and 17.3% OF. As SF was recognized in only a few samples, spatial distribution of fracture density is only discussed for the other two microfracture types.
 Figure 3 shows the spatial variation in fracture density within the damage zone for HF and OF. The OF fracture density decreases from 3.5 fractures/mm to 1 fractures/mm with increasing distance from the fault core. While HF measurements do not show the same relationship between fracture density and perpendicular distance from the fault core, there is an increase in the fracture density from 7 fractures/mm to 15 fractures/mm, with increasing proximity to a subsidiary fault that contains a white clayey fault gouge of 3–5 mm in thickness and is located ~110 m from the main fault. The dashed lines in Figure 2 indicate best-fit log-linear relations obtained by a least squares fit with the exponential decay equation, ρ = ρ0 ∙ eα ∙ D, where ρ (fractures/mm) is microfracture density; D (m) is distance from the main and subsidiary faults for open and healed microfractures, respectively; ρ0 (fractures/mm) is fracture density at D = 0; and α (m−1) is the decay constant. Values of ρ0, α, and R-squared are 3.5 ± 0.4, −0.0040 ± 0.0013, and 0.610 for open microfractures, respectively; and 16.5 ± 1.4, −0.0088 ± 0.0020, and 0.708 for healed microfractures, respectively. Since these R-squared values are similar to those of the previous studies that have investigated the relation between spatial distribution of fractures associated tectonic faulting and distance from faults [Mitchell and Faulkner, 2009; 2012], it can be interpreted that there exists an exponential decay relation between microfracture density and perpendicular distance from the fault we studied and the density distribution results from fault activity.
4 Discussion and Conclusion
 The results from our microfracture analysis show that the damage zone of the Quaternary-active fault is characterized by an exponential increase in OF fracture density, as the fault is approached. Previous studies have identified fault damage zones with HF along faults in crystalline and sedimentary rocks [Vermilye and Scholz, 1998; Wilson et al., 2003; Faulkner et al., 2006]; however, these studies did not measure the density of individual microfractures for all types of microfractures observed in samples. A previous study examined the Caleta Coloso Fault in northern Chile, which has been inactive during the Quaternary period [Delouis et al., 1998], and measured fracture density on each microfracture type [Mitchell and Faulkner, 2009]. The results showed that OF fracture density is independent of distance from the fault core, whereas HF fracture density showed an exponential decay with distance from the fault. A comparison shows that faults with different histories of fault activity have different spatial distribution characteristics of microfractures in the fault damage zone.
 One important question to arise from this analysis is as follows: Why does Quaternary fault movement contribute to the presence of a damage zone associated with OF? Based on cross-cutting relationships between microfracture types, we suggest that HF is older than OF. HF observed in the damage zone was previously OF before becoming a healed plane of secondary fluid inclusions due to the local-scale diffusive transport of the constitutive atoms in the crystallizing minerals [Smith and Evans, 1984]. Experimental and field studies have shown that microfracture healing likely occurs at ambient temperatures above ~100°C [Smith and Evans, 1984; Laubach, 1989]. The minimum depth of healing in the present study, Dh, is estimated to be at least 1 km, assuming that the geothermal gradient is 50–100°C/km in the study area [Tanaka et al., 1999]; OF would have formed at a shallower depth.
 Here we propose a fault-damage model that considers the depth conditions for HF and OF formation, and the process of fault zone exhumation (Figure 4). Cumulative slip along faults in the brittle crust produces damage zones where HF and OF form at depths that are greater than and less than Dh, respectively (Figure 4a). The fault zone is brought to the surface during regional uplift and erosion. HF observed in the damage zone studies here suggested to be exhumed from a depth greater than Dh. During exhumation, if the fault continues to slip, the damage zone is overprinted by OF, as shown by an exponential increase in fracture density with proximity to the fault core (Figure 4b). When the fault is inactive during exhumation, OF formation is independent of distance from the fault core and is due to stress relief associated with near-surface weathering (Figure 4c). Our data indicate that the fault-related OF observed in this study are the result of fault movement since ~6.6 Ma, including during the Quaternary period. This constraint on timing was obtained using a Dh of 1 km and a constant uplift rate of 0.15 mm/yr, as indicated by the height of a marine terrace in the study area that formed during the last interglacial period [Koike and Machida, 2001].
 We have shown that when the damage zone OF fracture density increases exponentially with increasing proximity to the fault core, this documents recent faulting related to exhumation from a depth greater than the depth of the HF/OF boundary. This approach extends the reliable evaluation of basement fault activity to settings in which the seismic history is difficult to reconstruct using conventional paleoseismic trenching techniques. Microfracture analysis, based on our fault-damage model, may contribute to greatly improving probability estimates of future destructive earthquake events along faults, which is important in planning nearby civil engineering projects.
 We are grateful to T. Nagata, for their support for collecting rock samples in the field. Also, we would like to thank K. Okuzawa and M. Watanabe for making thin sections. The reviews by Tomoyuki Ohtani and an anonymous reviewer greatly helped us to improve the manuscript.
 The Editor thanks Tomoyuki Ohtani and an anonymous reviewer for their assistance in evaluating this paper.