Pulse-like Rupture Research Articles

A proxy implementation of thermal pressurization for earthquake cycle modelling on rate-and-state faults

SUMMARY The reduction of effective normal stress during earthquake slip due to thermal pressurization of fault zone pore fluids is a significant fault weakening mechanism. Explicit incorporation of this process into frictional fault models involves solving the diffusion equations for fluid pressure and temperature outside the fault at each time step, which significantly increases the computational complexity. Here, we propose a proxy for thermal pressurization implemented through a modification of the rate-and-state friction law. This approach is designed to emulate the fault weakening and the relationship between breakdown energy and slip resulting from thermal pressurization and is appropriate for fully dynamic simulations of multiple earthquake cycles. It preserves the computational efficiency of conventional rate-and-state friction models, which in turn can enable systematic studies to advance our understanding of the effects of fault weakening on earthquake mechanics. In 2.5-D simulations of pulse-like ruptures on faults with finite seismogenic width, based on our thermal pressurization proxy, we find that the spatial distribution of slip velocity near the rupture front is consistent with the conventional square-root singularity, despite continued slip-weakening within the pulse, once the rupture has propagated a distance larger than the rupture width. An unconventional singularity appears only at shorter rupture distances. We further derive and verify numerically a theoretical estimate of the breakdown energy dissipated by our implementation of thermal pressurization. These results support the use of fracture mechanics theory to understand the propagation and arrest of very large earthquakes.

Pulse-like ruptures, seismic swarms, and tremorgenic slow-slip events with thermally activated friction

The evolution of frictional resistance on a fault affects the characteristics of seismic ruptures. A wide range of rupture styles, from slow-slip events to fast earthquakes, can be explained under the isothermal rate- and state-dependent friction framework. However, laboratory experiments indicate that friction also depends on temperature, with a largely unknown impact on rupture patterns. Here, we explore how thermally activated friction affects rupture behavior in quasi-dynamic models of seismic cycles with a single velocity-weakening, temperature-strengthening asperity, whereby frictional healing occurs behind the rupture front due to shear heating. A transition from crack-like to pulse-like rupture propagation with self-healing fronts occurs as the temperature strengthening effect increases, spontaneously inducing steady, decaying, or growing pulses. With increasing activation energy, the cycle turns into earthquake swarms and tremorgenic slow-slip events, both characterized by strong interactions between slow and fast ruptures. The temperature sensitivity of friction may contribute to the natural complexity of the seismic phenomenon, potentially explaining a much wider spectrum of rupture behaviors and recurrence patterns.

Rupture styles linked to recurrence patterns in seismic cycles with a compliant fault zone

Seismic cycles emerge in a broad range of rupture styles, from slow-slip events to pulse-like earthquake sequences. Meanwhile, large earthquakes in paleoseismic and instrumental catalogues exhibit various recurrence patterns going from periodic to chaotic cycles with characteristic or dissimilar ruptures. The potential connection between these observations is still poorly known. Here, we investigate the link between rupture styles and recurrence patterns in quasi-dynamic models of seismicity in two-dimensional faults embedded in a compliant zone, exploring a wide range of frictional and fault zone properties. The recurrence patterns evolve from purely periodic to multiple-periodic time- and slip-predictable cycles, to chaotic sequences of super-cycles with full and partial ruptures with an increasing number of aftershocks. This transition is accompanied by changes of rupture styles from crack-like to pulse-like ruptures. These behaviors can be obtained either by a more compliant fault zone or a reduced characteristic weakening distance of friction. The effects of the compliant zone and other physical characteristics can be conflated into a single non-dimensional number, such that seismic cycles with similar behaviors can be obtained with or without a compliant fault zone in quasi-dynamic simulations. The connection between recurrence patterns and rupture styles implies that the paleoseismic record can bring useful constraints to rupture characteristics and fault zone properties.

Minimal model for the onset of slip pulses in frictional rupture.

We present a minimal one-dimensional continuum model for the transition from cracklike to pulselike propagation of frictional rupture. In its nondimensional form, the model depends on only two free parameters: the nondimensional prestress and an elasticity ratio that accounts for the finite height of the system. The model predicts stable slip pulse solutions for slip boundary conditions, and unstable slip pulse solutions for stress boundary conditions. The results demonstrate that a mechanism based solely on elastic relaxation and redistribution of initial prestress can cause pulselike rupture, without any particular rate or slip dependences of dynamic friction. This means that pulselike propagation along frictional interfaces is likely a generic feature that can occur in systems of finite thickness over a wide range of friction constitutive laws.

Laboratory earthquakes along faults with a low velocity zone: Directionality and pulse-like ruptures

Low velocity zone (LVZ) is a common structural feature in geological faults. To examine the effect of LVZ on faulting, a laboratory fault model considering the LVZ is constructed and loaded to different levels. Subsequently ruptures are triggered on one of its contact boundaries with the host rock. The bilateral rupture propagation along this laboratory fault that separates different materials shows distinct directionality in both rupture speeds and rupture modes. The rupture in the positive direction with respect to the fault is always crack-like growing at the Generalized Rayleigh (GR) wave speed. However, in the negative direction both stable and unstable slip pulses are observed. Higher load and thinner LVZ facilitate the unstable slip pulses. The trailing tip for both types of pulses propagates at the GR wave speed, the leading tip of the stable pulse propagates with the GR wave speed while that of the unstable pulse features a supershear speed.

Propagation of large earthquakes as self-healing pulses or mild cracks.

Observations suggest that mature faults host large earthquakes at much lower levels of stress than their expected static strength1-11. Potential explanations are that the faults are quasi-statically strong but experience considerable weakening during earthquakes, or that the faults are persistently weak, for example, because of fluid overpressure. Here we use numerical modelling to examine these competing theories for simulated earthquake ruptures that satisfy the wellknown observations of 1-10megapascal stress drops and limited heat production. In that regime, quasi-statically strong but dynamically weak faults mainly host relatively sharp, self-healing pulse-like ruptures, with only a small portion of the fault slipping at a given time, whereas persistently weak faults host milder ruptures with more spread-out slip, which are called crack-like ruptures. We find that the sharper self-healing pulses, which exhibit larger dynamic stress changes compared to their static stress changes, result in much larger radiated energy than that inferred teleseismically for megathrust events12. By contrast, milder crack-like ruptures on persistently weak faults, which produce comparable static and dynamic stress changes, are consistent with the seismological observations. The larger radiated energy of self-healing pulses is similar to the limited regional inferences available for crustal strike-slip faults. Our findings suggest that either large earthquakes rarely propagate as self-healing pulses, with potential differences between tectonic settings, or their radiated energy is substantially underestimated, raising questions about earthquake physics and the expected shaking from large earthquakes.

Crack to pulse transition and magnitude statistics during earthquake cycles on a self-similar rough fault

Faults in nature demonstrate fluctuations from planarity at most length scales that are relevant for earthquake dynamics. These fluctuations may influence all stages of the seismic cycle; earthquake nucleation, propagation, arrest, and inter-seismic behavior. Here I show quasi-dynamic plane-strain simulations of earthquake cycles on a self-similar and finite 10 km long rough fault with amplitude-to-wavelength ratio α=0.01. The minimum roughness wavelength, λmin, and nucleation length scales are well resolved and much smaller than the fault length. Stress relaxation and fault loading is implemented using a variation of the backslip approach, which allows for efficient simulations of multiple cycles without stresses becoming unrealistically large. I explore varying λmin for the same stochastically generated realization of a rough fractal fault. Decreasing λmin causes the minimum and maximum earthquakes sizes to decrease. Thus the fault seismicity is characterized by smaller and more numerous earthquakes, on the other hand, increasing the λmin results in fewer and larger events. However, in all cases, the inferred b-value is constant and the same as for a reference no-roughness simulation (α=0). I identify a new mechanism for generating pulse-like ruptures. Seismic events are initially crack-like, but at a critical length scale, they continue to propagate as pulses, locking in an approximately fixed amount of slip. I investigate this transition using simple arguments and derive a characteristic pulse length, Lc=λmin/(4π4α2) and slip distance, δc based on roughness drag. I hypothesize that the ratio λmin/α2 can be roughly estimated from kinematic rupture models. Furthermore, I suggest that when the fault size is much larger than Lc, then most space-time characteristics of slip differ between a rough fault and a corresponding planar fault.

Cascading and pulse-like ruptures during the 2019 Ridgecrest earthquakes in the Eastern California Shear Zone

On July 4 2019, a Mw 6.5 earthquake, followed 34 h later by a Mw 7.1 event, struck Searles Valley, California. These events are part of a long-lived cluster of historical earthquakes along the Eastern California Shear Zone (ECSZ) which started in 1872 and are associated with temporarily elevated strain rates. We find that the Mw 6.5 event initiated on a right-lateral NW striking fault and then ruptured a left-lateral fault to the surface. This event triggered right-lateral slip during the Mw 7.1 earthquake. It started as a bilateral, crack-like rupture on a segment brought closer to failure by the Mw 6.5 event. The rupture evolved to pulse-like as it propagated at a relatively slow velocity (2 km/s) along a segment that was unloaded by the Mw 6.5 event. It stopped abruptly at the Coso volcanic area and at the Garlock Fault and brought some neighbouring faults closer to failure.

Similar scaling laws for earthquakes and Cascadia slow-slip events.

Faults can slip not onlyepisodically during earthquakes but also during transient aseismic slip events1-5, often called slow-slip events. Previous studies based on observations compiled from various tectonic settings6-8 have suggested that the moment of slow-slip events is proportional to their duration, instead of following the duration-cubed scaling found for earthquakes9. This finding has spurred efforts to unravel the cause of the difference in scaling6,10-14. Thanks to a new catalogue of slow-slip events on the Cascadia megathrust based on the inversion of surface deformation measurements between 2007 and 201715, we find that a cubic moment-duration scaling law is more likely. Like regular earthquakes, slow-slip events also have a moment thatis proportional to A3/2, where A is the rupture area, and obey the Gutenberg-Richter relationship between frequency and magnitude. Finally, these slow-slip events show pulse-like ruptures similar to seismic ruptures. The scaling properties of slow-slip events are thus strikingly similar to those of regular earthquakes, suggesting that they are governed by similar dynamic properties.

Full-field Ultrahigh-speed Quantification of Dynamic Shear Ruptures Using Digital Image Correlation

Producing dynamic ruptures in the laboratory allows us to study fundamental characteristics of interface dynamics. Our laboratory earthquake experimental setup has been successfully used to reproduce a number of dynamic rupture phenomena, including supershear transition, bimaterial effect, and pulse-like rupture propagation. However, previous diagnostics, based on photoelasticity and laser velocimeters, were not able to quantify the full-field behavior of dynamic ruptures and, as a consequence, many key rupture features remained obscure. Here we report on our dynamic full-field measurements of displacement, velocities, strains and strain rates associated with the spontaneous propagation of shear ruptures in the laboratory earthquake setup. These measurements are obtained by combining ultrahigh-speed photography with the digital image correlation (DIC) method, enhanced to capture displacement discontinuities. Images of dynamic shear ruptures are taken at 1-2 million frames/s over several sizes of the field of view and analyzed with DIC to produce a sequence of evolving full-field maps. The imaging area size is selected to either capture the rupture features in the far field or to focus on near-field structures, at an enhanced spatial resolution. Simultaneous velocimeter measurements on selected experiments verify the accuracy of the DIC measurements. Owing to the increased ability of our measurements to resolve the characteristic field structures of shear ruptures, we have recently been able to observe rupture dynamics at an unprecedented level of detail, including the formation of pressure and shear shock fronts in viscoelastic materials and the evolution of dynamic friction.

Finite-fault source inversion using adjoint methods in 3-D heterogeneous media

Accounting for lateral heterogeneities in the 3-D velocity structure of the crust is known to improve earthquake source inversion, compared to results based on 1-D velocity models which are routinely assumed to derive finite-fault slip models. The conventional approach to include known 3-D heterogeneity in source inversion involves pre-computing 3-D Green’s functions, which requires a number of 3-D wave propagation simulations proportional to the number of stations or to the number of fault cells. The computational cost of such an approach is prohibitive for the dense data sets that could be provided by future earthquake observation systems. Here, we propose an adjoint-based optimization technique to invert for the spatio-temporal evolution of slip velocity. The approach does not require pre-computed Green’s functions. The adjoint method provides the gradient of the cost function, which is used to improve the model iteratively employing an iterative gradient-based minimization method. The adjoint approach is shown to be computationally more efficient than the conventional approach based on pre-computed Green’s functions in a broad range of situations. We consider data up to 1 Hz from a Haskell source scenario (a steady pulse-like rupture) on a vertical strike-slip fault embedded in an elastic 3-D heterogeneous velocity model. The velocity model comprises a uniform background and a 3-D stochastic perturbation with the von Karman correlation function. Source inversions based on the 3-D velocity model are performed for two different station configurations, a dense and a sparse network with 1 and 20 km station spacing, respectively. These reference inversions show that our inversion scheme adequately retrieves the rise time when the velocity model is exactly known, and illustrates how dense coverage improves the inference of peak-slip velocities. We investigate the effects of uncertainties in the velocity model by performing source inversions based on an incorrect, homogeneous velocity model. We find that, for velocity uncertainties that have standard deviation and correlation length typical of available 3-D crustal models, the inverted sources can be severely contaminated by spurious features even if the station density is high. When data from thousand or more receivers is used in source inversions in 3-D heterogeneous media, the computational cost of the method proposed in this work is at least two orders of magnitude lower than source inversion based on pre-computed Green’s functions.

Shear Fractures of Extreme Dynamics

Natural and laboratory observations show that shear ruptures (faults) can propagate with extreme dynamics (up to intersonic rupture velocities) through intact materials and along pre-existing faults with frictional and coherent (bonded) interfaces. The rupture propagation is accompanied by significant fault strength weakening in the rupture head. Although essential for understanding earthquakes, rock mechanics, tribology and fractures, the question of what physical processes determine how that weakening occurs is still unresolved. The general approach today to explain the fault weakening is based upon the strong velocity-weakening friction law according to which the fault strength drops rapidly with slip velocity. Different mechanisms of strength weakening caused by slip velocity have been proposed including thermal effect, high-frequency compressional waves, expansion of pore fluid, macroscopic melting and gel formation. This paper proposes that shear ruptures of extreme dynamics propagating in intact materials and in pre-existing frictional and coherent interfaces are governed by the same recently identified mechanism which is associated with an intensive microcracking process in the rupture tip observed for all types of extreme ruptures. The microcracking process creates, in certain conditions, a special fan-like microstructure shear resistance of which is extremely low (up to an order of magnitude less than the frictional strength). The fan-structure representing the rupture head provides strong interface weakening and causes high slip and rupture velocities. In contrast with the velocity-weakening dependency, this mechanism provides the opposite weakening-velocity effect. The fan-mechanism differs remarkably from all reported earlier mechanisms, and it can provide such important features observed in extreme ruptures as: extreme slip and rupture velocities, high slip velocity without heating, off-fault tensile cracking, transition from crack-like to pulse-like rupture mode at variation in loading conditions, dramatic embrittlement of hard rocks at highly confined compression, abnormally low transient strength of hard rocks at high confining stresses, etc. All these questions are discussed in the paper.

Resolution of Rise Time in Earthquake Slip Inversions: Effect of Station Spacing and Rupture Velocity

Earthquake finite‐source inversions provide us with a window into earthquake dynamics and physics. Unfortunately, rise time, an important source parameter that describes the local slip duration, is still quite poorly resolved. This may be at least partly due to sparsity of currently available seismic networks, which have average sensor spacing of a few tens of kilometers at best. However, next generation observation systems could increase the density of sensing by orders of magnitude. Here, we explore whether such dense networks would improve the resolution of the rise time in idealized scenarios. We consider steady‐state pulselike ruptures with spatially uniform slip, rise time, and rupture speed and either Haskell or Yoffe slip‐rate function on a vertical strike‐slip fault. Synthetic data for various network spacings are generated by forward wave propagation simulations, and then source inversions are carried out using that data. The inversions use a nonparametric linear inversion method that does not impose any restrictions on rupture complexity, rupture velocity, or rise time. We show that rupture velocity is an important factor in determining the rise‐time resolution. For sub‐Rayleigh rupture speeds, there is a characteristic length related to the decay of the wavefield away from the fault that depends on rupture speed and rise time such that only networks with smaller station spacings can adequately resolve the rise time. For supershear ruptures, the wavefield contains homogeneous S waves the decay of which is much slower, and an adequate resolution of the rise time can be achieved for all station spacings considered in this study (up to few tens of kilometers). Finally, we find that even if dense measurements come at the expense of large noise (e.g., 1 cm/s noise for space‐based optical systems), the conclusions on the performance of dense networks still hold.

Pulse-Like Rupture Induced by Three-Dimensional Fault Zone Flower Structures

Faults are often embedded in low-velocity fault zones (LVFZ) caused by material damage. Previous 2D dynamic rupture simulations (Huang and Ampuero, 2011; Huang et al., 2014) showed that if the wave velocity contrast between the LVFZ and the country rock is strong enough, ruptures can behave as pulses, i.e. with local slip duration (rise time) much shorter than whole rupture duration. Local slip arrest (healing) is generated by waves reflected from the LVFZ–country rock interface. This effect is robust against a wide range of fault zone widths, absence of frictional healing, variation of initial stress conditions, attenuation, and off-fault plasticity. These numerical studies covered two-dimensional problems with fault-parallel fault zone structures. Here, we extend previous work to 3D and geometries that are more typical of natural fault zones, including complexities such as flower structures with depth-dependent velocity and thickness, and limited fault zone depth extent. This investigation requires high resolution and flexible mesh generation, which are enabled here by the high-order accurate arbitrary high-order derivatives discontinuous Galerkin method with an unstructured tetrahedral element discretization (Peltieset al., 2012). We show that the healing mechanism induced by waves reflected in the LVFZ also operates efficiently in such three-dimensional fault zone structures and that, in addition, a new healing mechanism is induced by unloading waves generated when the rupture reaches the surface. The first mechanism leads to very short rise time controlled by the LVFZ width to wave speed ratio. The second mechanism leads to generally longer, depth-increasing rise times, is also conditioned by the existence of an LVFZ, and persists at some depth below the bottom of the LVFZ. Our simulations show that the generation of slip pulses by these two mechanisms is robust to the depth extent of the LVFZ and to the position of the hypocenter. The first healing mechanism is dominant for events with hypocenter inside the LVFZ. The second one is dominant if the hypocenter is deeper than a shallow LVFZ. These results suggest that the depth-dependence of rise time might help constrain the depth extent of the LVFZ. We also show that ruptures can spontaneously stop in flower-like LVFZs with uniform velocity reduction, but continue propagating as slip pulses if velocity reduction is depth-dependent.

Properties of inelastic yielding zones generated by in-plane dynamic ruptures—I. Model description and basic results

We discuss results associated with 2-D numerical simulations of in-plane dynamic ruptures on a fault governed by slip-weakening and rate-and-state friction laws with off-fault yielding. The onset of yielding is determined by a Mohr–Coulomb-type criterion whereas the subsequent inelastic response is described by a Duvaut-Lions-type viscoplastic rheology. The study attempts to identify key parameters and conditions that control the spatial distribution and the intensity variation of off-fault yielding zones, the local orientation of the expected microfractures, and scaling relations or correlations among different quantities that can be used to characterize the yielding zones. In this paper, we present example results for crack and pulse ruptures, along with calculations of energy partition and characteristics of the simulated off-fault yielding zones. A companion follow-up paper provides a comprehensive parameter-space study of various examined features. In agreement with previous studies, the location and shape of the off-fault yielding zones depend strongly on the angle of the background maximum compressive stress relative to the fault and the crack versus pulse mode of rupture. Following initial transients associated with nucleation of ruptures, the rate of various energy components (including off-fault dissipation) linearly increases with time for cracks, while approaching a constant level for pulse-like ruptures. The local angle to the fault of the expected microfractures is generally shallower and steeper than in the compressional and extensional quadrants, respectively. The scalar seismic potency density decays logarithmically with increasing fault normal distance, with decay slope and maximum value that are influenced by the operating stress field.

A new paradigm for simulating pulse-like ruptures: the pulse energy equation

We investigate the chaotic behaviour of slip pulses that propagate in a spring block slider model with velocity weakening friction by numerically solving a computationally intensive set of n coupled non-linear equations, where n is the number of blocks. We observe that the system evolves into a spatially heterogeneous pre-stress after the occurrence of a sufficient number of events. We observe that, although the spatiotemporal evolution of the amplitude of a slip pulse in a single event is surprisingly complex, the geometric description of the pulses is simple and self-similar with respect to the size of the pulse. This observation allows us to write an energy balance equation that describes the evolution of the pulse as it propagates through the known pre-stress. The equation predicts the evolution of individual ruptures and reduces the computational time dramatically. The long-time solution of the equation reveals its multiscale nature and its potential to match many of the long-time statistics of the original system, but with a much shorter computational time.

Scale Effects in Simple Models for the Dynamics of Faults

Although the evidence for complexity is overwhelming, the dynamics of faulting is still poorly understood. Whilst it has long been known that discreteness in numerical earthquake models produces complexity, the mathematical structure and form of this complexity has never been fully established. Using a simple 1D nonlinear fault model we show how complexity can arise in discrete models through the presence of nonlinear, scale-dependent (or mesh-dependent) terms. We show that scale-dependencies may be a significant factor in the generation of slip complexity and pulse-like rupture over multiple earthquake cycles. We demonstrate that the introduction of length scales in discrete earthquake models implies that both strongly weakening friction and scale-dependent processes may be necessary in generating the pulse-like rupture mode and earthquake complexity over multiple earthquake cycles.

Pulse-like ruptures induced by low-velocity fault zones

Low-velocity fault zones (LVFZs) are found in most mature faults. They are usually 100–400 m wide and have ~20%–60% wave velocity reductions relative to the country rock. To study the effect of LVFZs on earthquake rupture and the radiated wavefield, we conducted two-dimensional (2-D) simulations of dynamic rupture on faults that bisect a LVFZ, considering a range of velocity reductions and widths. Most earthquakes apparently have slip rise times much shorter than their overall rupture duration. A number of dynamic mechanisms for such pulse-like ruptures have been proposed, including frictional self-healing, fault strength heterogeneities, and bimaterial effects. We find that ruptures in LVFZs with strong enough wave velocity contrast behave as pulses. These pulses are generated by fault locking induced by waves reflected from the boundaries of the LVFZ. This mechanism of pulse generation is robust to variations of initial stress, smoothness of the LVFZ structure, rupture mode, and exclusion of frictional healing. Moreover, we find that LVFZs can generate complex rupture patterns. LVFZs with low-velocity reduction induce multiple rupture fronts involving coexisting pulses and cracks. LVFZs with certain widths can accelerate the transition to supershear rupture speed. These additional effects of LVFZs on dynamic rupture can contribute to the complexity of high-frequency ground motions.

Pulse-like and crack-like dynamic shear ruptures on frictional interfaces: experimental evidence, numerical modeling, and implications

Destructive large earthquakes occur as dynamic frictional ruptures along pre-existing interfaces (or faults) in the Earth’s crust. One of the important issues in earthquake dynamics is the local duration of relative displacement or slip. Seismic inversions show that earthquakes may propagate as self-healing pulse-like ruptures, with local slip duration being much shorter than the overall rupture duration. Yet many classical models produce crack-like ruptures, with local slip durations comparable to the overall rupture duration. We study rupture modes in an experimental set up designed to mimic a fault prestressed both in compression and in shear. Our experiments demonstrate systematic variation from crack-like to pulse-like rupture modes as nondimensional shear prestress is decreased. The results of our experiments are consistent with theories of ruptures on interfaces with velocity-weakening friction. To consider the possibility that slip-weakening friction can also result in such rupture mode transition in the presence of the dynamic nucleation procedure employed by the experimental setup, we conduct numerical simulations with linear slip-weakening friction. In the simulations, we use the parameter regimes that were shown in previous studies to reproduce supershear transition distances obtained in the same experimental setup. We find that simulations with linear slip-weakening friction are unable to reproduce pulse-like ruptures, even in the presence of the dynamic initiation procedure. Our experimental results and simulations imply that velocity-weakening friction plays an important role in dynamic behavior of shear ruptures and needs to be included in earthquake models.

Pulse-like and crack-like ruptures in experiments mimicking crustal earthquakes

Theoretical studies have shown that the issue of rupture modes has important implications for fault constitutive laws, stress conditions on faults, energy partition and heat generation during earthquakes, scaling laws, and spatiotemporal complexity of fault slip. Early theoretical models treated earthquakes as crack-like ruptures, but seismic inversions indicate that earthquake ruptures may propagate in a self-healing pulse-like mode. A number of explanations for the existence of slip pulses have been proposed and continue to be vigorously debated. This study presents experimental observations of spontaneous pulse-like ruptures in a homogeneous linear-elastic setting that mimics crustal earthquakes; reveals how different rupture modes are selected based on the level of fault prestress; demonstrates that both rupture modes can transition to supershear speeds; and advocates, based on comparison with theoretical studies, the importance of velocity-weakening friction for earthquake dynamics.