Laboratory Generated M -6 Earthquakes

Abstract

We consider whether mm-scale earthquake-like seismic events generated in laboratory experiments are consistent with our understanding of the physics of larger earthquakes. This work focuses on a population of 48 very small shocks that are foreshocks and aftershocks of stick–slip events occurring on a 2.0 m by 0.4 m simulated strike-slip fault cut through a large granite sample. Unlike the larger stick–slip events that rupture the entirety of the simulated fault, the small foreshocks and aftershocks are contained events whose properties are controlled by the rigidity of the surrounding granite blocks rather than characteristics of the experimental apparatus. The large size of the experimental apparatus, high fidelity sensors, rigorous treatment of wave propagation effects, and in situ system calibration separates this study from traditional acoustic emission analyses and allows these sources to be studied with as much rigor as larger natural earthquakes. The tiny events have short (3–6 μs) rise times and are well modeled by simple double couple focal mechanisms that are consistent with left-lateral slip occurring on a mm-scale patch of the precut fault surface. The repeatability of the experiments indicates that they are the result of frictional processes on the simulated fault surface rather than grain crushing or fracture of fresh rock. Our waveform analysis shows no significant differences (other than size) between the M -7 to M -5.5 earthquakes reported here and larger natural earthquakes. Their source characteristics such as stress drop (1–10 MPa) appear to be entirely consistent with earthquake scaling laws derived for larger earthquakes.

Appendix

Our estimates of seismic moment and stress drop are dependent on the absolute accuracy of the Green’s functions, so we deemed it worthwhile to verify them by means of finite element modeling. We computed solutions to a 3D finite element model using the software PyLith (Aagaard et al. 2013). Explicit dynamic finite element models useful for the computation of elastodynamic Green’s functions become unstable when the number of elements per wavelength decreases to below about 10. The element size in our model is 3 mm, so the maximum frequency for S waves (v _s = 2,700 m/s) that our model can accommodate is about 90 kHz. Figure 9 shows synthetic seismograms obtained from the finite element model (thick grey lines) compared to synthetic seismograms obtained from the generalized ray theory code (black dashed lines) for the source and sensor geometries and orientations shown in Fig. 3. These synthetic seismograms are calculated exactly the same way as those shown in Fig. 3 except the instrument response function is not included and the width t₀ of a pulse-shaped moment rate function $\dot{m}(t)$ is set to 15 μs. This relatively wide pulse width (15 μs, corner frequency ~67 kHz) is required to keep the majority of the wave energy below our 90 kHz band limit, as set by the stability of the finite element model. The two solutions agree during the first part of the signals when only direct P and S waves and near field terms are present, and they diverge at later times due to (1) errors associated with later arriving reflections in the generalized ray theory code, and (2) reflections off the sides of finite element model which are not accounted for in the infinite slab geometry of the generalized ray solution.

Fig. 9

Comparison between synthetic seismograms obtained from our finite element model (thick grey lines) and synthetic seismograms obtained from the generalized ray theory code (black dashed lines), for the same source location as the LabEQ shown in Fig. 3 but different source duration. These synthetics include source and wave propagation components but do not include sensor distortions. The source term is a left-lateral strike-slip double couple focal mechanism with pulse-shaped moment rate function $\dot{m}(t)$ that is 15 μs wide and 85 kNm/s tall. Positive displacement is in the direction outward from the sample