Hironori Kawakata

Three-dimensional observations of faulting process in Westerly granite under uniaxial and triaxial conditions by X-ray CT scan

Abstract Observations of spatial fault development in granite undergoing compression provide new insights into the process of faulting. Dry intact Westerly granite samples were loaded under a confining pressure of 100 MPa (triaxial conditions) and 5 MPa (∼ uniaxial conditions), and the progress of faulting was controlled by maintaining the increment of circumferential displacement at a constant rate, which apparently stiffened the machine. The samples were unloaded after they experienced some degree of stress drop and were successfully recovered before faulting progressed further. A conventional medical X-ray CT scanning system was used to image the sample interiors. Three-dimensional fault systems were detected with sequential X-ray CT images. It was found that three-dimensional reconstruction by X-ray CT images yields not only three-dimensional images of the fault system, but also provides fault cross-section images with much less artificial noise (artifacts) than does direct X-ray CT imaging. Three-dimensional images show that a fault system that developed under uniaxial conditions is much more complicated than a fault system produced by triaxial conditions. In addition, the fault plane produced under uniaxial conditions is inclined at a lower angle to the maximum compressive axis than under triaxial conditions. Comparing X-ray CT images, we show that a fault nucleates locally on the sample surface just after peak stress, then develops into the final fault plane in the residual stress stage of the complete stress–strain relationship under triaxial conditions.

The observations of faulting in westerly granite under triaxial compression by X-ray CT scan

Abstract Observations of the spatial development of cracks in granite undergoing triaxial compression provide new insights into the proces of fault formation. Dry intact Westerly granite samples were loaded under confining pressure of 100 MPa, and the progress of fault formation was controlled by maintaining the increment of circumferential displacement at a constant rate. The samples were unloaded when they experienced stress drops and were successfully recovered without severe crushing. A stereo microscope was used to observe the spatial pattern of cracks on the sample surfaces, and an X-ray CT scanning system was used to non-destructively image the sample interiors. Microscope observations show that several fault-like linear cracks developed on the surfaces of all the samples and have an angle of about 30 degrees to the maximum compression axis. This suggests that fault nucleation began just after peak stress. The X-ray CT images show that a fault-like linear crack grew inward toward the sample interior as axial stress decreased. It is concluded that the fault nucleated locally on the sample surface just after peak stress, and developed into an apparent fault by the plateau stage of the complete stress-strain relationship.

Observation of numerous aftershocks of an Mw 1.9 earthquake with an AE network installed in a deep gold mine in South Africa

This is the first report from the JAGUARS (JApanese-German Underground Acoustic Emission Research in South Africa) project, the overall aim of which is to observe ultra-small fracturing in a more or less natural environment. We installed a local (∼40-m span) network of eight acoustic emission (AE) sensors, which have the capability to observe up to 200 kHz at a depth of 3.3 km in a South African gold mine. Our specific objective was to monitor a 30-m thick dyke that remains as a dip pillar against active mining ∼90 m above our network. An Mw 1.9 earthquake whose hypocenter was ∼30 m above the network occurred in the dyke. Although the mineowned geophone (4.5 Hz) network detected only five earthquakes in the surrounding 200×200×150-m3 volume within the first 150 h following the main shock, our AE network detected more than 20,000 earthquakes in the same period. More than 13,000 of these formed a distinct planar cluster (∼100×80 m2) on which the main shock hypocenter lay, suggesting that this cluster delineates the main shock rupture plane. Most of the aftershocks were presumably very small, probably as low as M ∼ −4. The aftershock cluster dipped ∼60°. This is consistent with normal faulting under a nearly vertical compression field, as indicated by nearly horizontal breakouts found in a borehole crossing the rupture plane.

Scale dependence of rock friction at high work rate

Determination of the frictional properties of rocks is crucial for an understanding of earthquake mechanics, because most earthquakes are caused by frictional sliding along faults. Prior studies using rotary shear apparatus revealed a marked decrease in frictional strength, which can cause a large stress drop and strong shaking, with increasing slip rate and increasing work rate. (The mechanical work rate per unit area equals the product of the shear stress and the slip rate.) However, those important findings were obtained in experiments using rock specimens with dimensions of only several centimetres, which are much smaller than the dimensions of a natural fault (of the order of 1,000 metres). Here we use a large-scale biaxial friction apparatus with metre-sized rock specimens to investigate scale-dependent rock friction. The experiments show that rock friction in metre-sized rock specimens starts to decrease at a work rate that is one order of magnitude smaller than that in centimetre-sized rock specimens. Mechanical, visual and material observations suggest that slip-evolved stress heterogeneity on the fault accounts for the difference. On the basis of these observations, we propose that stress-concentrated areas exist in which frictional slip produces more wear materials (gouge) than in areas outside, resulting in further stress concentrations at these areas. Shear stress on the fault is primarily sustained by stress-concentrated areas that undergo a high work rate, so those areas should weaken rapidly and cause the macroscopic frictional strength to decrease abruptly. To verify this idea, we conducted numerical simulations assuming that local friction follows the frictional properties observed on centimetre-sized rock specimens. The simulations reproduced the macroscopic frictional properties observed on the metre-sized rock specimens. Given that localized stress concentrations commonly occur naturally, our results suggest that a natural fault may lose its strength faster than would be expected from the properties estimated from centimetre-sized rock samples.

Magnitude −7 level earthquakes: A new lower limit of self-similarity in seismic scaling relationships

Microfractures occurring in a rock sample that are called acoustic emission (AE) events show some similar features to earthquakes. However, it remains to be shown whether or not AE equate to ultramicroearthquakes. In this study, we show the existence of magnitude −7 level earthquakes based on seismological analyses of AE source parameters. Advances in multichannel, broadband, high-speed continuous recording of AE under seismogenic pressure conditions has facilitated increasingly robust measurement. Source parameters of AE show that AE events satisfy the same scaling relationship as natural earthquakes in which seismic moment is inversely proportional to the cube of corner frequency. This result suggests that both millimeter scale fractures and natural earthquakes of kilometer scale ruptures are highly similar as physical processes. Hence, AE events can be interpreted as ultramicroearthquakes having a magnitude of about −7. These results demonstrate that laboratory observation is an effective approach in studying natural earthquake generation process.

Nucleation process of an M2 earthquake in a deep gold mine in South Africa inferred from on-fault foreshock activity

Using a network of sensitive high-frequency acoustic emission sensors, we observed foreshock activity of an Mw 2.2 earthquake (main shock) in a deep gold mine in South Africa. Foreshock activity, which selectively occurred on a part of the rupture plane of the forthcoming main shock, lasted for at least 6 months until the main shock. Rock samples recovered from the main shock source region showed evidence of ancient hydrothermal alteration on the main shock rupture plane, suggesting that the foreshock activity occurred on a preexisting weakness. The foreshocks during 3 months before the main shock were concentrated in three clusters (F1–F3), which we interpret as representing localized preslip at multiple sites. While the location of mining area, the source of stress perturbations, changed with time, the locations of foreshock clusters did not change, suggesting that the preslip patches were controlled by strength heterogeneity rather than stress distribution. Activity over the entire foreshock area was generally constant, but the largest cluster (F2) showed accelerated activity starting at least 7 days before the main shock, while mining stress did not increase in this period. The main shock initiated at a point close to F1, away from F2. All the six foreshocks during the final 41 h occurred in F1 and F2 and in-between. These suggest that in the last stage of the preparation process of the main shock, preslip patches interacted with each other through the stress concentration ahead of the expanding preslip patch (F2), which should be the only driving force of the preparation process under the constant external load.

Quasi‐static slip patch growth to 20 m on a geological fault inferred from acoustic emissions in a South African gold mine

Three months of acoustic emission (AE) monitoring in a South African gold mine down to Mw −5 revealed a newly emergent planar cluster of 7557 events −3.9 ≤ Mw ≤ −1.8 (typical rupture radius of 6–70 cm) that expanded with time to reach a size of 20 m on a preexisting geological fault near an active mining front 1 km beneath the ground. It had a sharply defined, planar configuration, with hypocenters aggregated within a thickness of only several decimeters. We infer that the zone defines an aseismic slip patch on the fault, wherein the individual AEs represent failures of very small asperities being loaded by the aseismic slip. Additional support for the interpretation was obtained by analyzing composite focal mechanisms and repeating events. The patch expansion over 2 months was likely quasistatic because all individual AEs ruptured much smaller areas than the cluster size at the corresponding time. The b values dropped gradually from 2.6 to 1.4, consistent with a significant increase in shear stress expected of the mining style. Another cluster with similar characteristics emerged later on a neighboring part of the same fault and grew to a 10 m extent in the last weeks of the study period. The quasi-static expansion of inferred localized slow-slip patches to sizes of 10–20 m suggests that the critical crack length on natural faults can be at least as large, much exceeding the decimeter range derived from laboratory stick-slip experiments on saw-cut rocks.

Broadband P waves transmitting through fracturing Westerly granite before and after the peak stress under a triaxial compressive condition

We analyzed temporal changes in the velocity and amplitude of P waves transmitted through a granite sample during a triaxial compression test, with the goal of monitoring the fault formation process associated with open and shear cracking. We used newly developed transducer assemblies for the broadband recording, and we continued to record transmitting waves even after the peak stress occurred. For transmitting P waves with paths parallel to the maximum compressive axis, we found that both the first wave amplitude and the velocity decreased after dilatancy started, and they kept decreasing even after the peak stress. In addition, the large nonlinear decrease in amplitude was associated with a rapid decrease in differential stress, whereas the rate of decrease in velocity remained almost constant. Thus, before the rapid decrease of differential stress, when both the amplitude and the velocity gradually decreased, open cracking was indicated to be dominant. Thereafter, shear cracking was indicated to become dominant in synchronization with the rapid decrease in differential stress. It is suggested that a main fault started to grow around the sample surface and then progressed into the sample interior; this corresponded to the rapid stress decrease. This fault acts as a strong scatterer for P waves that are parallel to the maximum compressive axis.

Gross structure of a fault during its formation process in Westerly granite

Abstract We investigated the gross structure of a fault in three dimensions during its formation process. The observed sample was Westerly granite recovered from compression after peak stress but before dynamic rupture. The sample was infused with a fluorescent resin, mechanically stripped, then observed with a stereo-microscope under UV light. The main fault plane was found to be macroscopically elliptical, and the fault trace on each polished surface was identical to that detected by X-ray CT scanning. Evidence suggests that the fault plane widened its cracked zone as it grew. The gross structures of the fault system, such as a step-over structure and sub-faults parallel to the main fault, were also detected. We conclude that the step-over structure was formed during the fault formation process.

Temporal changes in the Q of broadband P waves transmitting through a fracturing westerly granite sample under triaxial compressive conditions

Abstract We analyzed temporal changes in the quality factor ( Q ) of P waves transmitting through a fracturing Westerly granite sample under triaxial compressive conditions, using the rise time of the initial wave. Broadband recording was performed throughout, even after the differential stress reached the peak strength. We calibrated a transducer assembly and corrected the recorded waveforms. Using theoretical waveforms, we derived a new equation relating Q P , the rise time for the velocity waveform, and the travel time. Q P remained at ∼23 before the start of dilatancy and decreased to ∼17 after the peak stress was reached. A large nonlinear decrease in Q P was not clearly observed, while the first wave amplitude clearly decreased rapidly after the occurrence of the peak stress.