Loading a 3-m granite slab containing a saw-cut simulated fault, we generated slip events that spontaneously nucleate, propagate, and arrest before reaching the ends of the sample. This work shows that slow (0.07 mm/s slip speeds) and fast (100 mm/s) contained slip events can occur on the same fault patch. We also present the systematic changes in radiated seismic waves both in time and frequency domain. The slow earthquakes are 100 ms in duration and radiate tremor-like signals superposed onto a low-frequency component of their ground motion. They are often preceded by slow slip (creep) and their seismic radiation has an ω −1 spectral shape, similar to slow earthquakes observed in nature. The fastest events have slip velocity, stress drop, and apparent stress (0.2 m/s, 0.4 MPa, and 1.2 kPa, respectively) similar to those of typical M −2.5 earthquakes, with a single distinct corner frequency and ω −2 spectral falloff at high frequencies, well fit by the Brune earthquake source model. The gap between slow and fast is filled with intermediate events with source spectra depleted near the corner frequency. This work shows that a fault patch of length p with conditions favorable to rupture can radiate in vastly different ways, based on small changes in p h * ; where h * is a critical nucleation length scale. Such a mechanism can help explain atypical scaling observed for low-frequency earthquakes that compose tectonic tremor.Plain Language Summary Some faults slip slowly and silently, while others are locked and then slip spontaneously in an abrupt fashion. There is debate on whether slow and fast slip are distinct processes or lie on a continuum of fault slip modes in nature. Recent laboratory observations show that a single fault can slip in both modes and at some intermediate velocities, creating a spectrum of slow to fast earthquakes. We have conducted laboratory earthquake experiments where we force a fault cut in a 3-m-long granite rock to slip under pressure. The experiments show that by giving a rupture more distance to accelerate before it runs in to unfavorable fault conditions and dies, it can emit high-frequency energy more efficiently and exhibit resemblance to natural regular earthquakes. On the other hand, if a rupture barely accelerates before stopping, it will emit weak tremors. This proposed mechanism and the seismic consequences highlighted in this study may explain the puzzling behavior of deep tectonic tremor sometimes radiated from slow earthquakes.A family of slow earthquakes have been observed in the past 20 years with the deployment of continuous GPS recording systems and high-sensitivity borehole seismometers and strain meters. Independent Key Points:• We generate contained earthquake-like slip events on a 3 m dry, homogeneous granite fault that do not rupture through the sample ends • We create a spectrum of slow to fast events, ranging from M −3.2 events with 50 kPa stress drop to M −2.5 quakes with 0.4 MPa stress drop • Slow events produce tremor-like seismic radiation and have an ω −1 spe...

No abstract

The interpretation of precursory seismicity can depend on a critical nucleation length scale h*, yet h* is largely unconstrained in the seismogenic crust. To estimate h* and associated earthquake nucleation processes at 2–7 km depths in Oklahoma, we studied seismic activity occurring prior to nine M 2.5–3.0 earthquakes that are aftershocks of the 3 September 2016 M 5.8 Pawnee, Oklahoma, earthquake. Four of the nine M 2.5–3.0 aftershocks studied did not have detectable seismicity within a 2 km radius of their hypocenters in the preceding 16 hr time windows. For the other five events, which did exhibit foreshock sequences, we estimated the static stress changes associated with each event of each sequence based on precise earthquake relocations and magnitude estimates. By carefully examining the spatiotemporal characteristics, we found all five of these M 2.5–3.0 aftershocks, and 70% of our studied events were plausibly triggered via static stress transfer from nearby earthquakes occurring hours to seconds earlier, consistent with the cascade nucleation model and a small h* in this region. The smallest earthquakes we could quantitatively study were M −1.5 events, which likely have 1–2 m rupture dimensions. The existence of these small events also supports a small nucleation length scale h*≤1  m, consistent with laboratory estimates. However, our observations cannot rule out more complicated earthquake initiation processes involving interactions between foreshocks and slow slip. Questions also remain as to whether aftershocks initiate differently from more isolated earthquakes.