Findings

Introduction

Subduction zone earthquakes exhibit spatial variability in seismic behavior along megathrusts, ranging from stable sliding or episodic slow slip events (SSEs) to large earthquakes (Ide et al. 2007). This variation likely reflects spatial variations in frictional properties along the seismogenic portion of plate-boundary megathrusts (e.g., Bilek and Lay 1998; Scholz 1998; Bilek 2007; Wang and He 2008; Rushing and Lay 2012).

An increasing number of experimental studies have found that the frictional properties of subduction-zone material vary depending on mineralogy, such as clay content, emphasizing the expected strong material control on frictional properties along seismogenic subduction zone thrusts (Vrolijk 1990; Brown et al. 2003; Kopf and Brown 2003; Ikari et al. 2009a; Ujiie and Tsutsumi 2010; Tsutsumi et al. 2011; Ikari et al. 2013; Saito et al. 2013; Takahashi et al. 2013; Ujiie et al. 2013). Given the diversity in subduction-zone stratigraphy (Underwood 2007), it is essential to characterize the material inputs to subduction zones with respect to their intrinsic frictional properties if fault behavior within seismogenic subduction zones is to be better understood (Vrolijk 1990; Underwood 2007). However, available experimental data have thus far been limited mostly to clayey subduction-zone materials.

In this study, to explore variations in the intrinsic frictional properties of subduction zone sediment inputs that may be manifested in earthquake behavior along an erosional subduction zone interface, we examined the frictional properties of such sediments on the Cocos plate offshore the Osa Peninsula, Costa Rica. Our results reveal that variations in material inputs should produce variations in frictional properties along faults in the shallow Costa Rica subduction zone, providing an important constraint on faulting along the Costa Rica subduction zone interface.

Geological setting

Along the Middle America Trench, offshore the western margin of Costa Rica, the oceanic Cocos plate is subducting beneath the Caribbean plate at a convergence rate of 70 to 90 mm/year, forming the southern end of the Middle America Trench (Protti et al. 1994; DeMets, 2001; Figure 1a). The Cocos plate subducting beneath Costa Rica was formed at two different spreading centers, the East Pacific Rise (EPR) and the Cocos-Nazca spreading center (CNS), and has been significantly influenced by Galapagos hotspot volcanism (von Huene et al. 2000; Barckhausen et al. 2001). The largest feature formed by the passage of the Cocos plate over the Galapagos hotspot is the 2.5-km-high Cocos Ridge (Figure 1).

Figure 1
figure 1

Topographic and bathymetric map of the Costa Rica area and Hole U1381A summary. (a) Topographic and bathymetric map of the Costa Rica area (Marine Geoscience Data System, MGDS; http://www.marine-geo.org; Ryan et al. 2009). The tectonic setting and major elements on the map are after Barckhausen et al. (2001). EPR, East Pacific Rise crust; CNS, Cocos-Nazca Spreading Center crust. Seismic event locations (stars) are from Arroyo et al. (2009). (b) Summary of the lithostratigraphy (after Expedition 334 Scientists 2012). Red arrows show the depth intervals of the core samples from which the experimental samples were taken. Sample 7R-2 W shows both white and black parts: the white part (7R-2 W white) can be classified in the silicic to calcareous ooze group, and the black part (7R-2 W black) in the tephra group.

Full size image

The location of shallow earthquakes along the plate interface varies along strike: earthquakes occur shallower and closer to the trench in the CNS crust than in the EPR crust (Protti et al. 1994; Newman et al. 2002). In June 2002, a Mw 6.4 earthquake occurred at a depth as shallow as ~9 km along the Costa Rica margin offshore the Osa Peninsula (Arroyo et al. 2009). The subduction zone beneath the Caribbean plate is also characterized by strong variations in fault-slip behavior. Recently, the network of GPS stations around the Nicoya Peninsula has detected SSEs accompanied by seismic tremors in the Costa Rica subduction zone, not only at depths of 25 to 30 km near the down-dip limit of the seismogenic zone but also at depths of ~6 km near the up-dip limit (Outerbridge et al. 2010; Jiang et al. 2012). However, SSEs are not observed around the Osa Peninsula.

Integrated Ocean Drilling Program (IODP) Expedition 334, which also constituted a part of the Costa Rica Seismogenesis Project (CRISP), was established to understand the processes that control the nucleation and seismic rupture of large earthquakes along the Costa Rica erosive margin offshore the Osa Peninsula (Vannucchi et al. 2012). The primary objective of Expedition 334 was to both sample and quantify the sediment inputs to the seismogenic zone of an erosive subduction margin. In this expedition, Site U1381, which is located 50 km offshore the Osa Peninsula (Figure 1), is a key site for characterizing the material inputs to the subduction zone. The samples analyzed in the present study were recovered from this site.

Experimental samples

All the samples tested in this study were collected from the sediments on the Cocos plate offshore the Osa Peninsula, at Site U1381, during IODP Expedition 334 (Figure 1a). The cores recovered from this site can be divided into three lithostratigraphic units (Figure 1b; Expedition 334 Scientists, 2012). The 95.57-m-thick cover sequence is divided into two units on the basis of lithology: Unit I is composed of clay to silty clay and Unit II is composed of silicic to calcareous ooze (Expedition 334 Scientists 2012). Underlying these sediments, Unit III consists mainly of pillow basalt (Expedition 334 Scientists 2012). A total of 29 tephra horizons are observed within the cover sediments (3 horizons in Unit I and 26 horizons in Unit II), with the tephra layers in Unit II being composed mostly of volcanic glass (Expedition 334 Scientists 2012). This study examined the frictional properties of four clay to silty clay samples (Unit I, Figure 2a), six silicic to calcareous ooze samples (Unit II, Figure 2c), and two tephra samples (tephras in Unit II, Figure 2b).

Figure 2
figure 2

XRD analysis and smear slide observation of the experimental samples. XRD patterns of the tested samples and representative smear slide pictures of the samples. (a) Clay and silty clay. (b) Volcanic glass. (c) Silicic to calcareous ooze. Q, quartz; Pl, plagioclase; C, calcite; H, halite; Py, pyrite; clay, clay minerals.

Full size image

Biogenic components are observed in most of the samples. However, Unit II is distinguished from Unit I by its abundant biogenic components (Expedition 334 Scientists 2012). The clay to silty clay samples contain foraminifers, calcareous nannofossils, a few sponge spicules, and minor amounts of small silicic shells of organisms such as diatoms (Figure 2a). The silicic to calcareous ooze samples consist mainly of radiolarians, diatoms, silicoflagellates, sponge spicules, foraminifers, and calcareous nannofossils (Figure 2c). The tephra samples also include foraminifers and calcareous nannofossils.

Figure 2 shows X-ray diffraction (XRD) patterns obtained from the samples. The clay to silty clay samples consist of quartz, plagioclase, pyrite, phyllosilicate, and biogenic components. In contrast, the silicic to calcareous ooze samples of Unit II are almost free of clay minerals and are instead composed mainly of calcite and possibly amorphous silica. The existence of amorphous silica is indicated by a broad peak detected at 2θ = 20° to 30°. The tephra samples also show similar broad peaks that indicate an abundance of glass.

Experimental procedure

Friction experiments were conducted on the samples using a rotary-shear, intermediate- to high-velocity friction-testing machine. The collected samples were dried at 60°C for 24 h and disaggregated gently using a pestle. For each experiment, 0.5 g of the sample was placed between a pair of gabbro cylinders, each measuring 24.98 mm in diameter. Distilled water was added to the experimental gouge layer to saturate it. The experimental fault was surrounded by a Teflon® (DuPont, Wilmington, DE, USA) ring to avoid gouge expulsion during experiments. The gouge was axially pre-compacted under the test conditions (5 MPa) for 30 min.

In the friction experiments, all the pre-compacted samples were subjected to an initial slip velocity of 0.28 mm/s until ~170 mm of displacement (three rotations of the sample cylinder), during which time the friction value either increased or decreased to attain a roughly steady-state value. After this pre-sliding, a sequence of velocity stepping by a factor of 10 was imposed to examine the velocity dependence of friction for loading velocities of 0.0028 to 2.8 mm/s and displacements of >250 mm.

Velocity dependence was measured directly from plots of the data (e.g., Figure 3a,b,c,d) as the frictional constitutive parameter (a - b) = (Δμss/ΔlnV) for a step change in sliding velocity using the first several hundred microns of displacement after the stepped velocity changes. As illustrated in Figure 3a,b,c,d, the response of friction to velocity steps is superimposed on longer-term gradual changes in friction, which occur over displacements of sometimes more than a few millimeters. These displacements are far larger than the typical critical weakening displacements reported for various gouge materials (e.g., Marone et al. 1990; Marone 1998). We interpret that such long-term behavior would likely be ascribed to the fluctuation in friction caused by the sample rotation, rather than to a transient frictional behavior modeled by the rate- and state-dependent friction laws (e.g., Dieterich 1979; Ruina 1983).

Figure 3
figure 3

Velocity dependence of friction measured by velocity-step tests. Representative results of the velocity-step tests. (a), (b) Sample U1381A-6R-2 W from Unit I (clay to silty clay) shows positive velocity dependence at V = 0.028 to 2.8 mm/s. (c), (d) Sample U1381A-8R-1 W from Unit II (silicic to calcareous ooze) shows negative velocity dependence at V = 0.028 to 0.28 mm/s and positive velocity dependence at V = 0.28 to 2.8 mm/s. The inset figures show an enlarged plot for the duration of the velocity step. (e) Values of frictional velocity dependence of the tested samples for the range of sliding velocities V = 0.0028 to 2.8 mm/s.

Full size image

Experimental results

The frictional behavior of the samples at a constant slip velocity V of 0.28 mm/s differs between the samples of different lithostratigraphic units (Figure 4). For the clay to silty clay samples (Unit I), the friction coefficient μ is ~0.4 upon the onset of sliding, then gradually decreases over the initial 40 mm of sliding to attain a roughly constant value of ~0.15 (Figure 4b). In contrast, μ of the silicic to calcareous ooze samples is initially 0.4 to 0.6, then gradually increases to attain a roughly constant value of 0.6 to 0.8 (Figure 4c). The friction coefficient μ of the tephra samples is roughly constant within the range 0.6 to 0.8 over the entire displacement of the pre-sliding.

Figure 4
figure 4

Difference in shear strength between Unit I and Unit II sediments. (a) Friction coefficient values (both initial peak and steady-state values) at constant velocity (0.28 mm/s) plotted against the depth intervals (mbsf) of the recovered core samples. (b), (c) Evolution of the coefficient of friction with shear displacement for two representative samples: U1381A-6R-2 W (Unit I) and U1381A-8R-1 W (Unit II).

Full size image

Figure 4a shows the initial peak and steady-state μ values measured for all the samples. The values of friction coefficient μ of the sediments at the onset of sliding at V = 0.28 mm/s, except for the tephra samples, are roughly in the same range of 0.4 to 0.6. In contrast, the steady-state μ values display clear differences between samples from different lithostratigraphic units. It is notable that the values of μ of the silicic to calcareous ooze samples decrease with greater sampling depth.

The samples tested in this study can be divided into two groups with regard to the velocity dependence of steady-state friction. Two representative results are shown in Figure 3a,c, in which V was changed in a stepwise manner between 0.028 and 0.28 mm/s. Sample U1381A-6R-2 W (Figure 3a), representative of the clay to silty clay samples of Unit I, shows consistently positive velocity dependence at these velocities, whereas sample U1381A-8R-1 W (Figure 3c), representative of the silicic to calcareous ooze samples of Unit II, shows mostly negative velocity dependence. The critical slip distances Dc at these velocities also differ between the two samples: the value of sample U1381A-6R-2 W is ~0.1 mm, whereas that of sample U1381A-8R-1 W is ~0.05 mm. At higher velocities of >0.28 mm/s, the velocity dependence of the clay to silty clay is still positive (Figure 3b), whereas that of the silicic to calcareous ooze is either neutral or positive (Figure 3d).

Values of (a - b) are plotted as a function of up-step velocities in Figure 3e. Values of (a - b) of the clay to silty clay samples are low (<0.005) at lower velocities but increase to ~0.01 at the highest velocities. The absolute values of (a - b) of the silicic to calcareous ooze samples are mostly low (<0.005) at the lowest velocities of 0.0028 to 0.028 mm/s. At intermediate velocities of 0.028 to 0.28 mm/s, absolute values of (a - b) become smaller and approach zero, implying that the silicic to calcareous ooze samples exhibit an almost neutral velocity dependence of friction at these velocities (Figure 3e).

Similar contrasting frictional properties of subduction zone material have recently been reported by Ikari et al. (2013) from an experimental study of the sediments entering the Costa Rica subduction zone offshore the Nicoya Peninsula. In the Nicoya Peninsula region, calcareous nannofossil chalk is frictionally strong (μ = 0.71 to 0.88) and shows velocity-weakening behavior, whereas the hemipelagic clayey sediment is weak (μ = 0.22 to 0.35) and exhibits mostly velocity-strengthening behavior.

Discussion and conclusions

The observed weakness of the clay to silty clay sediments suggests that the frictional strength of faults is controlled mainly by the content of clay minerals, as has been demonstrated in many experimental studies (e.g., Summers and Byerlee 1977; Morrow et al. 1992, 2000; Brown et al. 2003; Kopf and Brown 2003; Moore and Lockner 2004; Ikari et al. 2009a,[b]; Tembe et al. 2010; Saito et al. 2013; Takahashi et al. 2013). In contrast, the steady-state μ values of the silicic to calcareous ooze sediments from Unit II are high, measuring ~0.8 and are comparable to the values reported for crustal rock samples (Byerlee 1978). The strength of the silicic to calcareous ooze samples probably arises because they are almost free of weak clay minerals.

The initial peak and steady-state μ values of the silicic to calcareous ooze samples decrease systematically with increasing depth (Figure 4a). No systematic change in physical properties (e.g., P-wave velocity, porosity, and shear strength) that would correlate to the variation in μ values has been reported for the Unit II cores (Expedition 334 Scientists 2012). However, it is interesting to note that some geochemical constituents of the pore fluid of the cores, such as the contents of Ca and SO4, change monotonically with depth within Unit II (Expedition 334 Scientists 2012). This suggests the possible influence of fluid chemistry on the frictional properties of the fault material.

In the toe region of subduction zones, décollements are likely to initiate within, or shift their positions to, zones of weaker stratigraphic intervals (Underwood 2007). Given the low frictional coefficient values observed for the clay to silty clay samples of Unit I, the décollement at the CRISP transect offshore the Osa Peninsula likely initiated in Unit I and was initially very weak. It should be noted that our velocity-step tests reveal that clay to silty clay exhibits only positive frictional velocity dependence (Figure 3). The observed velocity-strengthening behavior and the very low steady-state μ values of the Unit I sediments suggest that the faults in the shallow portion of the Costa Rica subduction zone are stable and thus behave as creeping segments.

In contrast, the silicic to calcareous ooze samples from Unit II, which are characterized by higher μ values, exhibit mostly velocity-weakening behavior for slip velocities of 0.0028 to 0.28 mm/s (Figure 3e). Velocity-weakening behavior implies potentially unstable fault motion (Scholz 1998). In addition, the lower Dc values of silicic to calcareous ooze samples may mean that faults that develop in Unit II are more unstable. The difference in Dc values between Unit I and Unit II may be caused by different clay mineral contents. The June 2002 Mw 6.4 earthquake implies the existence of velocity-weakening material in this focal depth region. Fault slip that nucleated along a fault in Unit II would be a likely candidate for the source of the shallow earthquake event.

Variation in the distribution and type of earthquakes along the Costa Rica subduction zone, such as the shallow earthquake that occurred at Osa Peninsula (the 2002 Mw 6.4 event, Figure 1a) and the SSEs at Nicoya Peninsula, is suggestive of variations in fault zone material along the plate interface. The simple stratification signature of the material inputs (Figure 1b) may help us to better understand the processes leading to a heterogeneous distribution of fault material along the subduction channel. For instance, the topography of the seafloor may contribute to generating a heterogeneous distribution of materials along the plate boundary (Ikari et al. 2013). Considering the possible association of seismicity with the dewatering reaction along the plate interface (e.g., Spinelli et al. 2006; Audet and Schwartz 2013), the stabilizing effect of pore fluid pressure in the vicinity of velocity-weakening faults must be taken into account. High pore fluid pressure may produce a conditionally stable region along a velocity-weakening fault with a small value of (a - b) by reducing the effective normal stress (Scholz 1998). A velocity-weakening fault in Unit II would be stable under quasistatic loading but may become unstable under sufficiently strong dynamic loading. Velocity-strengthening behavior, as observed for the Unit II samples at higher velocities of 0.28 to 2.8 mm/s (Figure 3e), may also act to stabilize the propagation of earthquake ruptures, and may influence slip patterns along faults or the magnitudes of earthquakes. As suggested by Saito et al. (2013), such velocity-strengthening behavior, in addition to variation in pore fluid pressure, is important for understanding the behavior of faults in subduction zones.