betaFeedback

WorksPhysics and Astronomy

Breakdown of the classical rupture theory and earthquake propagation in the "forbidden" super-Rayleigh range

Abstract

Earthquakes propagating faster than the shear wave-speed are commonly thought to undergo a super-shear transition upon which they discontinuously jump from the sub-Rayleigh regime to the super-shear one. The super-Rayleigh regime, i.e., the range of propagation speeds between the Rayleigh and shear wave-speeds, is regarded as "forbidden" by the two-dimensional classical rupture theory. Here, we revisit the assumptions underlying the classical theory and develop a rupture theory that takes into account the dependence of the fault strength (frictional resistance) on the slip rate. The theory quantitatively agrees with numerical simulations nearly up to the Rayleigh wave-speed. Yet, very close to the latter, two-dimensional rupture solutions change their character due to frictional rate nonlinearity and rupture continuously propagates through the "forbidden" super-Rayleigh range into the super-shear regime, without a sharp super-shear transition. These results demonstrate that frictional rate dependence, generically observed in experiments, can have profound implications for fast earthquake propagation.

Introduction

The propagation speed of earthquake rupture along a geological fault — a prominent example of a frictional interface formed by two contacting bodies — directly affects seismic energy radiation and consequently also the associated ground motion and seismic hazard [1]. It is now established that a non-negligible fraction of the large earthquakes documented over the last few decades were super-shear, i.e., propagated at speeds exceeding the shear wave-speed csc_{\mathrm{s}} [3] [2]. Such super-shear ruptures are accompanied by the generation of Mach wavefronts, which potentially give rise to strong shaking over distances away from the fault that are significantly larger than those of sub-shear earthquakes [4, 5, 6, 7, 8]. Understanding the emergence of super-shear rupture is therefore a basic problem in earthquake physics.

A central aspect of this problem stems from the two-dimensional (2D) classical rupture theory [9, 10, 11], which predicts that shear rupture along homogeneous faults cannot propagate at speeds between the Rayleigh wave-speed cRc_{\mathrm{R}} (the speed of surface waves) and csc_{\mathrm{s}}. Accepting this prediction, a major theoretical challenge has been to identify possible physical mechanisms to overcome the “forbidden” super-Rayleigh speeds range, allowing rupture to transition from the sub-Rayleigh regime to the super-shear one, i.e., to undergo a super-shear transition.

The earliest mechanism, and a prominent one, demonstrated that shear rupture along homogeneous 2D faults can nucleate secondary (‘daughter’) rupture ahead of the main (‘mother’) rupture edge [12, 13, 14, 15, 16] — depending on the propagation distance and the frictional stress conditions [17] —, which upon coalescence with the main rupture can give rise to a discontinuous jump from the sub-Rayleigh regime to the super-shear one. This discontinuous ‘mother-daughter’ super-shear transition can be realized in three dimensions (3D) as well [17]. Various other super-shear transition mechanisms have been explored, including fault heterogeneities in both 2D and 3D (e.g., asperities or barriers [18, 19, 20, 21, 22], fault steps overs [23, 24, 25], fault roughness [26], fault damage zone [27], geometric effects (e.g., free surface effects [28]), and genuine 3D effects [29]), among others. Super-shear rupture has also been documented in laboratory experiments, mainly on glassy polymers [16, 30, 31, 32, 33] but also on rocks [34, 35].

Here, we revisit the classical 2D rupture theory and its underlying assumptions, which lead to the prediction of a forbidden speeds range. One of these assumptions involves a residual frictional strength behind the rupture edge that is independent of the sliding rate, which is inconsistent with a broad range of observations that demonstrate the intrinsic and nonlinear rate dependence of frictional interfaces [36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51]. We develop a linearized rupture theory that relaxes this assumption and demonstrate its validity both well below the Rayleigh wave-speed cRc_{\mathrm{R}} and close to it using large-scale numerical simulations. Yet, the theory breaks down very close to cRc_{\mathrm{R}} due to frictional rate nonlinearity and numerical solutions reveal a continuous frictional rupture propagation through the forbidden super-Rayleigh speeds range into the super-shear regime, without a sharp super-shear transition. These findings demonstrate the possible limitations of the classical 2D rupture theory and highlight the importance of frictional rate dependence for fast earthquake propagation.

Generalized rupture-edge singular fields and energy balance

Our first goal is to explicitly spell out the underlying assumptions of the classical 2D rupture theory and to explore their implications, which will also lead to our central theoretical development. To that aim, consider a fault/frictional interface formed by two large linear elastic bodies under the application of a constant compressive (normal) stress. That is, it is assumed that the two identical bodies follow the isotropic and homogeneous Hooke’s law (1+ν)μ[∇u+(∇u)T]=σ−ν(Itr⁡σ−σ)(1+\nu)\mu[\nabla\mathbf{u}+(\nabla\mathbf{u})^{\mathsf{T}}]=\boldsymbol{\sigma}-\nu(I\operatorname{tr}\boldsymbol{\sigma}-\boldsymbol{\sigma}) [52], relating Cauchy’s stress tensor σ(x,y,t)\boldsymbol{\sigma}(x,y,t) to the displacement vector u(x,y,t)\mathbf{u}(x,y,t). Here, II is the identity tensor, ν\nu is Poisson’s ratio and μ\mu is the shear modulus. The coordinates are chosen such that the interface lies along the xx-axis, the direction normal to the interface is the yy-axis (the interface is located at y=0y=0) and tt is time. The Navier-Lamé linear elastodynamic equation [52] for u\mathbf{u} is obtained when Hooke’s law is used inside the momentum equation ∇⋅σ=ρu¨\nabla\cdot\boldsymbol{\sigma}=\rho\ddot{\mathbf{u}}, where ρ\rho is the mass density and a superposed dot denotes a partial time derivative.

Consider then, in addition to the above-mentioned normal stress, denoted by σ\sigma, the application of a far-field shear stress τd\tau_{\mathrm{d}} such that σxy(x,y=±∞,t)=τd\sigma_{xy}(x,y=\pm\infty,t)=\tau_{\mathrm{d}}, which drives an interfacial shear rupture. The location of the rupture edge features a general time dependence L(t)L(t) in the sub-shear range, i.e., 0≤cr(t)≤cs0\leq c_{\mathrm{r}}(t)\leq c_{\mathrm{s}}, where cr(t)≡L˙(t)c_{\mathrm{r}}(t)\equiv\dot{L}(t). The edge separates a non-sliding interfacial state ahead of it from a sliding state behind it. That is, the slip (tangential displacement discontinuity) δ(x,t)≡2ux(x,y=0,t)\delta(x,t)\equiv2u_x(x,y=0,t) satisfies δ(x≥L(t),t)=0\delta(x\geq L(t),t)=0 and δ(x<L(t),t)>0\delta(x<L(t),t)>0. The former implies that the fault/interface behaves elastically ahead of the rupture edge. Behind it, on the other hand, the shear stress is determined by the contact interaction between the sliding bodies, i.e., by the frictional strength τ\tau, such that σxy(x<L(t),y=0,t)=τ\sigma_{xy}(x<L(t),y=0,t)=\tau. The problem formulation is completed once the dependence of τ\tau on interfacial fields is explicitly indicated, i.e., when the interfacial constitutive relation is specified. We first proceed without specifying it.

A central assumption of the classical 2D rupture theory [10, 11] is that solutions to above-formulated problem, i.e., the resulting linear elastodynamic fields, admit a power-law series expansion in terms of a polar coordinate system (r,θ)(r,\theta) co-moving with the rupture edge (such that its origin is located at (x−L(t),y=0)(x-L(t),y=0) and θ=0\theta=0 is the propagation direction) [53]. Most importantly, it is assumed that the power-law series is dominated by a singular contribution to the displacement gradient proportional to rξr^\xi, with −1<ξ<0-1<\xi<0, sufficiently close to the edge [54]. That is, it is assumed that

∇u(r,θ,t)∼rξwith−1<ξ<0,(1)\nabla\mathbf{u}(r,\theta,t)\sim r^\xi\qquad\text{with}\qquad-1<\xi<0, \tag*{(1)}

asymptotically in the small rr limit.

Equation (1) implies that the leading-order contribution to the shear stress ahead of the rupture edge takes the form σxy(r,θ=0,t)=KII(ξ)rξ/2π\sigma_{xy}(r,\theta=0,t)=K_{\mathrm{II}}^{(\xi)}r^\xi/\sqrt{2\pi}, valid for any instantaneous, time-dependent propagation speed cr(t)c_{\mathrm{r}}(t) [10]. Here, KII(ξ)K_{\mathrm{II}}^{(\xi)} quantifies the intensity of the singularity (1/2π1/\sqrt{2\pi} is a convention), which cannot be obtained by an asymptotic analysis. In these terms, the slip rate v(r,t)=δ˙(r,t)v(r,t)=\dot{\delta}(r,t) and the shear stress behind the rupture edge take the form (see SI Appendix)

v(r,t)=KII(ξ)2πrξsin⁡(πξ)2crαsμR(cr)(αs2−1),(2)v(r,t)=\frac{K_{\mathrm{II}}^{(\xi)}}{\sqrt{2\pi}}r^\xi\sin(\pi\xi)\frac{2c_{\mathrm{r}}\alpha_{\mathrm{s}}}{\mu R(c_{\mathrm{r}})}(\alpha_{\mathrm{s}}^2-1), \tag*{(2)}
σxy(r,θ=±π,t)=KIII(ξ)2πrξcos⁡(ξπ).(3)\sigma_{xy}(r,\theta=\pm\pi,t)=\frac{K_{\mathrm{III}}^{(\xi)}}{\sqrt{2\pi}}r^\xi\cos(\xi\pi). \tag*{(3)}

Here, αs,d(cr)=1−cr2/cs,d2\alpha_{\mathrm{s},\mathrm{d}}(c_{\mathrm{r}})=\sqrt{1-c_{\mathrm{r}}^2/c_{\mathrm{s,d}}^2} and R(cr)=4αdαs−(1+αs2)2R(c_{\mathrm{r}})=4\alpha_{\mathrm{d}}\alpha_{\mathrm{s}}-(1+\alpha_{\mathrm{s}}^2)^2 are functions of cr(t)c_{\mathrm{r}}(t), where cdc_{\mathrm{d}} is the dilatational wave-speed. The properties of the Rayleigh function R(cr)R(c_{\mathrm{r}}), which vanishes at the Rayleigh wave-speed cRc_{\mathrm{R}}, will be discussed later.

Equations (2)–(3) can be used to quantify rupture-edge energy balance. In this context, it is important to note that the singular fields discussed above cannot lead to real divergences, i.e., they are regularized over a small lengthscale, the so-called cohesive zone of size ℓc\ell_{\mathrm{c}} [10, 11, 55, 56]. That is, the near-edge singular fields in fact correspond to ‘intermediate asymptotics’, valid on scales rr much smaller than the macroscopic scales characterizing the problem, but larger than the cohesive zone, i.e., r≥ℓcr\geq\ell_{\mathrm{c}}. This implies, for example, that the slip rate continuously transitions from the no-sliding state (v=0v=0) ahead of the rupture edge to the sliding state (v>0v>0) behind it, where it peaks at r∼ℓcr\sim\ell_{\mathrm{c}} and then follows Eq. (2) at intermediate scales.

At these intermediate scales, the shear stress is expected to follow Eq. (3) such that we can relate the frictional strength τ\tau to the slip δ\delta. To proceed, we note that partial time derivatives in the problem can be related to spatial gradients according to ∂t=cr∂r\partial_t=c_{\mathrm{r}}\partial_r, valid when cr(t)c_{\mathrm{r}}(t) does not vary rapidly. Consequently, we have v(r,t)∼cr∂δ(r,t)/∂rv(r,t)\sim c_{\mathrm{r}}\partial\delta(r,t)/\partial r, which upon using Eqs. (2)–(3) and the boundary condition σxy(r,θ=±π,t)=τ\sigma_{xy}(r,\theta=\pm\pi,t)=\tau, leads to τ(δ)∼τδξ1+ξ\tau(\delta)\sim\tau\delta^{\frac{\xi}{1+\xi}}. The latter is valid for r>ℓcr>\ell_{\mathrm{c}}, i.e., for δ>δc\delta>\delta_{\mathrm{c}}, where δc\delta_{\mathrm{c}} is the slip accumulated when the rupture edge propagated a distance ℓc\ell_{\mathrm{c}}. To quantify the dissipation associated with frictional slip behind the edge, we define the “breakdown energy” Gf(δ)≡∫0δ[τ(δ′)−τ(δ)] dδ′G_{\mathrm{f}}(\delta)\equiv\int_0^\delta[\tau(\delta')-\tau(\delta)]\,\mathrm{d}\delta' [49, 55, 57, 58, 59, 60], from which we obtain (see SI Appendix)

Gf(δ)=Gc(δδc)ξG(ξ)withξG(ξ)=1+2ξ1+ξ,(4)G_{\mathrm{f}}(\delta)=G_{\mathrm{c}}\left(\frac{\delta}{\delta_{\mathrm{c}}}\right)^{\xi_{\mathrm{G}}(\xi)} \qquad\text{with}\qquad \xi_{\mathrm{G}}(\xi)=\frac{1+2\xi}{1+\xi}, \tag*{(4)}

for δ>δc\delta>\delta_{\mathrm{c}}, where Gc≡Gf(δc)G_{\mathrm{c}}\equiv G_{\mathrm{f}}(\delta_{\mathrm{c}}) is the edge-localized dissipation.

Finally, GcG_{\mathrm{c}}—that quantifies the near-edge dissipation over the cohesive zone scale ℓc\ell_{\mathrm{c}}—is balanced by the flux of elastic energy into the edge region, which is quantified by the energy release rate GG [10]. The latter can be computed using the above-derived singular fields, leading to

G=1−ν2μAII(cr,ξ)[KII(ξ)]2ℓc2ξ+1=Gc,(5)G=\frac{1-\nu}{2\mu}A_{\mathrm{II}}(c_{\mathrm{r}},\xi)\left[K_{\mathrm{II}}^{(\xi)}\right]^2\ell_{\mathrm{c}}^{2\xi+1}=G_{\mathrm{c}}, \tag*{(5)}

where AII(cr,ξ)A_{\mathrm{II}}(c_{\mathrm{r}},\xi) is given in the SI Appendix. Equation (5) constitutes rupture-edge energy balance.

Equations (1)–(5) are valid for any ξ\xi in the range −1<ξ<0-1<\xi<0. It is crucial to stress that Eqs. (1)–(5) directly follow from the assumptions of the classical 2D rupture theory, i.e., bulk linear elastodynamics and a power-law series solution dominated by a singular contribution, with no additional assumptions, prior to the selection of the singularity order ξ\xi. The latter and its implications next are discussed next. The friction law involves strongly non-equilibrium, nonlinear, and dissipative physics and may vary from fault to fault. Our goal here is not to consider a specific and very detailed friction law, but rather to identify generic and minimal properties of the frictional strength and explore their implications.

The selection of the singularity order and the forbidden speeds range

The singularity order ξ\xi is selected by the boundary condition σxy(r,θ=±π,t)=τ\sigma_{xy}(r,\theta=\pm\pi,t)=\tau once the dependence of the frictional strength τ\tau on other interfacial fields — i.e., the interfacial/fault constitutive relation (the “friction law”) — is specified.

A very large body of experimental evidence — accumulated over several decades — showed that the frictional strength depends on the slip rate vv and on the state of the fault [36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51], expressed as τ(v,…)\tau(v,\ldots). Here, the ellipsis stands for interfacial state fields (e.g., roughness-induced contact area, gouge layer, fluid pore pressure, temperature etc.) that follow their own evolution equations. Generically, τ(v,…)\tau(v,\ldots) is nonlinear. To make theoretical progress, we focus on scales behind the rupture edge larger than ℓc\ell_{\mathrm{c}} (i.e., on δ>δc\delta>\delta_{\mathrm{c}}), where the frictional strength is expected to be predominantly slaved to the slip rate, i.e., to be given by τ(v)\tau(v). We then expand the latter as [61, 62]

τ(v)≃τ0+ηeffv+O(v2),(6)\tau(v)\simeq\tau_0+\eta_{\mathrm{eff}}v+\mathcal{O}(v^2), \tag*{(6)}

over some range of slip rates, where ηeff\eta_{\mathrm{eff}} is an effective viscous-friction coefficient.

Equation (6) allows to discuss another crucial assumption of the classical rupture theory. The latter assumes no rate dependence, implying that τ(v)≃τ0\tau(v) \simeq\tau_0, where τ0\tau_0 is a constant residual stress. Consequently, the boundary condition σxy(r,θ=±π,t)=τ0\sigma_{xy}(r,\theta=\pm\pi,t)=\tau_0 implies that the singular contribution to the shear stress behind the rupture edge in Eq. (3) must vanish, i.e., that ξ=−1/2\xi=-1/2, which is nothing but the conventional square root singularity of the classical theory [10]. Substituting ξ=−1/2\xi=-1/2 in Eqs. (1)–(5), all classical results are recovered [10]. Most importantly, ξG(ξ=−1/2)=0\xi_{\mathrm{G}}(\xi=-1/2)=0 in Eq. (4), which implies Gf(δ)=GcG_{\mathrm{f}}(\delta)=G_{\mathrm{c}} for δ>δc\delta>\delta_{\mathrm{c}}, KII≡KII(−1/2)K_{\mathrm{II}}\equiv K_{\mathrm{II}}^{(-1/2)} is the classical stress-intensity-factor, the dependence of GG on ℓc\ell_{\mathrm{c}} in Eq. (5) is eliminated and AII(cr,ξ=−1/2)=(cr/cs)2αs/[(1−ν)R(cr)]A_{\mathrm{II}}(c_{\mathrm{r}},\xi=-1/2)=(c_{\mathrm{r}}/c_{\mathrm{s}})^{2}\alpha_{\mathrm{s}}/[(1-\nu)R(c_{\mathrm{r}})] therein. The latter is the origin of the “forbidden” super-Rayleigh speeds range in the classical theory, as explained next.

The Rayleigh function R(cr)=4αdαs−(1+αs2)2R(c_{\mathrm{r}})=4\alpha_{\mathrm{d}}\alpha_{\mathrm{s}}-(1+\alpha_{\mathrm{s}}^{2})^{2} is positive in the sub-Rayleigh regime, 0≤cr<cR0\leq c_{\mathrm{r}}<c_{\mathrm{R}}, vanishes at the Rayleigh wave-speed, cr=cRc_{\mathrm{r}}=c_{\mathrm{R}}, and becomes negative in the super-Rayleigh regime, cR<cr<csc_{\mathrm{R}}<c_{\mathrm{r}}<c_{\mathrm{s}}. Consequently, AII(cr,ξ=−1/2)∼1/R(cr)A_{\mathrm{II}}(c_{\mathrm{r}},\xi=-1/2)\sim1/R(c_{\mathrm{r}}) diverges as cr→cRc_{\mathrm{r}}\to c_{\mathrm{R}} and becomes negative in cR<cr<csc_{\mathrm{R}}<c_{\mathrm{r}}<c_{\mathrm{s}}, implying that the rupture-edge energy balance G=GcG=G_{\mathrm{c}} cannot be satisfied (note that Gc>0G_{\mathrm{c}}>0 for all speeds). Moreover, the slip rate in Eq. (2) with ξ=−1/2\xi=-1/2 is predicted to feature the same divergence and sign change, which appear to violate physical expectation. Consequently, the classical theory excludes rupture propagation at super-Rayleigh speeds [10].

The assumption about the lack of frictional rate dependence, i.e., τ(v)≃τ0\tau(v)\simeq\tau_0, is clearly inconsistent with a broad range of observations [63]. Our next goal is to relax this assumption and thoroughly explore its implications for frictional rupture (earthquake) propagations, especially at high speeds. Therefore, we use the leading-order rate dependent term in Eq. (6), which preserves linearity and allows analytical progress, inside the frictional boundary condition σxy(r,θ=±π,t)=τ(v)\sigma_{xy}(r,\theta=\pm\pi,t)=\tau(v). Together with Eqs. (2)–(3), we obtain

cot⁡(πξ)≃−ηeffcsμ2(cr/cs)3αsR(cr)(7)\cot(\pi\xi)\simeq-\frac{\eta_{\mathrm{eff}}c_{\mathrm{s}}}{\mu}\frac{2(c_{\mathrm{r}}/c_{\mathrm{s}})^{3}\alpha_{\mathrm{s}}}{R(c_{\mathrm{r}})} \tag*{(7)}

which selects the singularity order ξ\xi. It is important to note that Eq. (7), whose properties will be further discussed below, predicts that ξ\xi is non-universal, depending on both the rupture speed crc_{\mathrm{r}} and the friction law, through ηeff(cr)\eta_{\mathrm{eff}}(c_{\mathrm{r}}).

Equations (1)–(7) constitute the unconventional singularities theory of frictional rupture. It builds on, and expands, the recent developments of [61, 62]. Unconventional singularity orders, ξ≠−1/2\xi\ne-1/2, have emerged in the framework of various rate-dependent constitutive laws, including visco-friction [61, 62, 64, 65, 66], thermal pressurization [49], flash heating [50], as well as in hydraulic fracture [67, 68, 69]. Moreover, a non-constant breakdown energy Gf(δ)G_{\mathrm{f}}(\delta) for δ>δc\delta>\delta_{\mathrm{c}}, i.e., Eq. (4) with ξG(ξ)>0\xi_{\mathrm{G}}(\xi)>0, has been extensively discussed in recent literature, e.g., [49, 50, 61, 63, 70, 71, 72, 73], though not in the context of super-Rayleigh rupture propagation and the super-shear transition. Equations (1)–(7), some of which have already received recent experimental support [71, 72], have not been systematically and quantitatively tested so far against frictional rate-dependent elastodynamic simulations of in-plane shear rupture, which is our next goal.

Testing the unconventional singularities theory

To test the theory, to be abbreviated hereafter as UST, we performed large-scale 2D boundary integral method simulations employing a nonlinear, rate-and-state friction law [74]. The details of the friction law are given in the SI Appendix, where its single most important property in the present context is that its steady sliding frictional strength τss(v)\tau_{\mathrm{ss}}(v) depends nonlinearly on the slip rate vv and features an N shape, observed in numerous experiments [47, 48]. The N-shape gives rise to rate-strengthening branch of this N-shaped curve, above its local minimum at (vmin⁡,τmin⁡)(v_{\min},\tau_{\min}), is expected to be relevant at the tail of a propagating earthquake, potentially making Eq. (6) with ηeff>0\eta_{\mathrm{eff}}>0 relevant.

We generate rupture by the earthquake nucleation procedure explained in SI Appendix under various background stress (prestress) levels τd\tau_{\mathrm{d}}. For a fixed earthquake nucleation procedure, friction parameters and fault size, cr(t)c_{\mathrm{r}}(t) is mainly controlled by τd\tau_{\mathrm{d}}. We tracked it, along with the spatiotemporal evolution of the slip rate v(x,t)v(x,t) and of the fault strength τ(x,t)\tau(x,t) fields. These computations allow to quantitatively test Eqs. (1)–(7). It is crucial to stress that the latter offer very stringent predictions; for example, ξ\xi can be obtained from Gf(δ)G_{\mathrm{f}}(\delta) of Eq. (4) and independently from Eq. (7), once ηeff\eta_{\mathrm{eff}} is extracted.

In Figure 1, we present a very comprehensive test of Eqs. (1)–(7) for a slowly accelerating rupture propagating well below cRc_{\mathrm{R}}, here cr(t)/cR≃0.27c_{\mathrm{r}}(t)/c_{\mathrm{R}}\simeq0.27. The results demonstrate (see figure caption for details) that the theory is in very good quantitative agreement with all the numerical observations with an unconventional singularity order ξ≃−0.36\xi\simeq-0.36, independently and self-consistently obtained from both extracting ξG(ξ)\xi_{\mathrm{G}}(\xi) in Eq. (4) and from Eq. (7), involving the measured ηeff\eta_{\mathrm{eff}}. The results also imply that the classical 2D rupture theory with ξ=−1/2\xi=-1/2 fails to quantitatively account for the numerical data.

Image not extracted

Figure 1. Testing the unconventional singularity theory well below cRc_{\mathrm{R}}. A frictional rupture propagating at cr(t)/cR≃0.27c_{\mathrm{r}}(t)/c_{\mathrm{R}}\simeq0.27, obtained by a 2D boundary integral method computation, is used to test Eqs. (1)–(7). (a) τ(δ)\tau(\delta) at a fixed position on the fault/interface during the passage of the rupture. The physical meaning of the vertical lines, here and elsewhere in the figure, will be explained below. Note that τ\tau continues to weaken with slip δ\delta even far beyond the large strength drop. (b) The corresponding breakdown energy Gf(δ)G_{\mathrm{f}}(\delta) (see text for definition). The superposed dashed line corresponds to a power-law predicted in Eq. (4), valid from δc\delta_{\mathrm{c}} (vertical solid line) over a range Δ\Delta (up to the vertical dashed line). The double-logarithmic slope is ξG\xi_G, from which ξ≃−0.36\xi\simeq-0.36 is obtained using Eq. (4). (c) τ(r)\tau(r) behind the rupture edge, cf. the prediction in Eq. (3), where r=∣xp−x∣r=|x_{\mathrm{p}}-x| and xpx_{\mathrm{p}} is the peak shear stress position. The cohesive zone of size ℓc\ell_{\mathrm{c}}, appearing in Eq. (5), is the distance between xpx_{\mathrm{p}} and the vertical solid line (defined above). (d) The corresponding slip rate field v(r)v(r), cf. the prediction in Eq. (2). The superposed dashed line is discussed in panel (f). (e) τ(v)\tau(v), where the superposed dashed line is a linear fit between the vertical lines, corresponding to Eq. (6). It results in ηeff\eta_{\mathrm{eff}}, which upon solving Eq. (7), yields ξ≃−0.36\xi\simeq-0.36, in excellent agreement with the value independently obtained in panel (b). (f) The same as panel (d), but on a double-logarithmic scale. The superposed dashed line is the prediction in Eq. (2) with ξ\xi predicted in panels (b) and (e). The same line is superposed in panel (d), both also determining KII(ξ)K_{\mathrm{II}}^{(\xi)}, the generalized stress-intensity-factor. (g) KII(ξ)K_{\mathrm{II}}^{(\xi)} vs. tt and (inset) the corresponding cr(t)c_{\mathrm{r}}(t). (h) Testing Eq. (5), where the RHS Gc(t)G_{\mathrm{c}}(t) (diamonds) is obtained using Eq. (4) as in panel (b) according to Gc=Gf(δc)G_{\mathrm{c}}=G_{\mathrm{f}}(\delta_{\mathrm{c}}), and the LHS (circles) using the physical parameters extracted above. See text and SI Appendix for additional details and extracted parameter values.

Next, we consider significantly larger propagation speeds. Specifically, we increased the background stress τd\tau_{\mathrm{d}} and focus on a time snapshot for which cr(t)/cR≃0.93c_{\mathrm{r}}(t)/c_{\mathrm{R}}\simeq0.93, i.e., relatively close to cRc_{\mathrm{R}}. The analysis is presented in Fig. 2 and demonstrates yet again good quantitative agreement between the UST and the numerical observations with ξ=−(0.34±0.03)\xi=-(0.34\pm0.03), where the small uncertainty reflects a small difference in the predictions of Eq. (4) and Eq. (7). These results further highlight the breakdown of the classical 2D rupture theory with ξ=−1/2\xi=-1/2 in our numerical data and the success of the UST up to high propagation speeds. They also raise the following question: what does the UST predict for rupture propagation very close to cRc_{\mathrm{R}} and in the super-Rayleigh range?

Image not extracted

Figure 2. Testing the unconventional singularities theory close to cRc_{\mathrm{R}}. Similarly to Figure 1, but for cr(t)≃0.93cRc_{\mathrm{r}}(t) \simeq0.93c_{\mathrm{R}}. (a) As in Figure 1d. (inset) As in Figure 1e. (b) As in Figure 1b. The obtained unconventional singularity order is ξ=−(0.34±0.03)\xi=-(0.34\pm0.03), where the upper value corresponds to panel (b) and the lower to the inset of panel (a), see text and SI Appendix for details and all extracted parameter values.

Approaching cRc_{\mathrm{R}} and super-Rayleigh propagation

To start addressing the central question posed above, we first analyze the prediction for ξ\xi in Eq. (7) and its crc_{\mathrm{r}} dependence as cRc_{\mathrm{R}} is approached. Since R(cr→cR)→0R(c_{\mathrm{r}}\to c_{\mathrm{R}})\to0, the right-hand-side (RHS) of Eq. (7) diverges in this limit, implying that the left-hand-side (LHS) cot⁡[πξ(cr)]∼1/sin⁡[πξ(cr)]\cot[\pi\xi(c_{\mathrm{r}})]\sim1/\sin[\pi\xi(c_{\mathrm{r}})] also diverges, while sin⁡[πξ(cr)]/R(cr)\sin[\pi\xi(c_{\mathrm{r}})]/R(c_{\mathrm{r}}) remains finite. Consequently, Eq. (7) predicts that ξ→0\xi\to0 for cr→cRc_{\mathrm{r}}\to c_{\mathrm{R}}, and that the slip rate in Eq. (2)—that is proportional to sin⁡[πξ(cr)]/R(cr)\sin[\pi\xi(c_{\mathrm{r}})]/R(c_{\mathrm{r}})—does not experience any divergence in this limit, in sharp contrast to the corresponding prediction of the classical theory with ξ=−1/2\xi=-1/2. These predictions assume that ηeff(cr)\eta_{\mathrm{eff}}(c_{\mathrm{r}}) reveals no special behavior in the vicinity of cRc_{\mathrm{R}}, which is expected on general grounds and verified (SI Appendix).

A vanishing singularity order, ξ→0\xi\to0, amounts to rupture propagation without stress amplification. This can be excluded on physical grounds since a stress amplification relative to the prestress in the close vicinity of the rupture edge is necessary for interfacial failure and for the propagation of frictional rupture. Consequently, the UST appears to signal its own breakdown in the cr→cRc_{\mathrm{r}}\to c_{\mathrm{R}} limit. Since Eq. (7) is equivalent to the frictional boundary condition σxy(r,θ=±π,t)=τ(v)\sigma_{xy}(r,\theta=\pm\pi,t)=\tau(v) under the linearization of τ(v)\tau(v) in Eq. (6), this breakdown suggests that nonlinearity in τ(v)\tau(v) becomes increasingly important in the cr→cRc_{\mathrm{r}}\to c_{\mathrm{R}} limit. In the presence of such nonlinearity, the power-law series assumption of the classical rupture theory cannot hold and ξ\xi can no longer be selected.

A direct and transparent way to probe the emergence of frictional rate nonlinearity in τ(v)\tau(v) in the cr→cRc_{\mathrm{r}}\to c_{\mathrm{R}} limit is through the behavior of Gf(δ)G_{\mathrm{f}}(\delta) for δ>δc\delta>\delta_{\mathrm{c}}. On the one hand, the UST in Eq. (4) predicts that Gf(δ)G_{\mathrm{f}}(\delta) varies as a power-law with ξG(ξ→0)→1\xi_{\mathrm{G}}(\xi\to0)\to1 in this limit. On the other hand, concave considerations suggest that Gf(δ)G_{\mathrm{f}}(\delta) will not reveal a clean power-law with ξG→1\xi_{\mathrm{G}}\to1. To quantitatively distinguish between the two, we quantify the power-law scaling range of Gf(δ)G_{\mathrm{f}}(\delta), denoted by Δ\Delta in Figure 1b and Fig. 2b, where the emergence of frictional rate nonlinearity in τ(v)\tau(v) would be manifested in a significant reduction in Δ\Delta.

In Figure 3a, we plot Δ(cr)/δc\Delta(c_{\mathrm{r}})/\delta_{\mathrm{c}} that is quantified exactly as in Figure 1b and Fig. 2b, i.e., by estimating the δ\delta range over which Gf(δ)G_{\mathrm{f}}(\delta) follows a straight line for δ>δc\delta>\delta_{\mathrm{c}} in a double-logarithmic representation (see SI Appendix). It is observed that Δ(cr)/δc\Delta(c_{\mathrm{r}})/\delta_{\mathrm{c}} slowly decreases in the range 0.27cR<cr<0.93cR0.27c_{\mathrm{R}}<c_{\mathrm{r}}<0.93c_{\mathrm{R}} (with some variation within each dataset, see caption), where we already established the validity of the UST, but rather strongly decreases for 0.93cR<cr<cR0.93c_{\mathrm{R}}<c_{\mathrm{r}}<c_{\mathrm{R}} (shaded region).

Image not extracted

Figure 3. The breakdown of the unconventional singularities theory approaching cRc_{\mathrm{R}}. (a) Δ(cr)/δc\Delta(c_{\mathrm{r}})/\delta_{\mathrm{c}}, quantified as in Figure 1b and 2b. Each color/symbol corresponds to a different value of the prestress τd\tau_{\mathrm{d}}, increasing from left to right, see SI Appendix. The shaded region starts at cr≃0.93cRc_{\mathrm{r}}\simeq0.93c_{\mathrm{R}}, corresponding to the rupture that is analyzed in Fig. 2 and shown therein to be quantitatively consistent with the UST, see text. It is observed that Δ(cr)\Delta(c_{\mathrm{r}}) slowly decreases in the range 0.27cR<cr<0.93cR0.27c_{\mathrm{R}}<c_{\mathrm{r}}<0.93c_{\mathrm{R}} (with some variation within each τd\tau_{\mathrm{d}} dataset), but significantly decreases for cr>0.93cRc_{\mathrm{r}}>0.93c_{\mathrm{R}}, corresponding to the emergence of significant frictional rate nonlinearity as cr→cRc_{\mathrm{r}}\to c_{\mathrm{R}}. It marks the breakdown of the UST as cRc_{\mathrm{R}} is approached. (b) The power-law scaling exponent ξG(cr)\xi_G(c_{\mathrm{r}}) corresponding to panel (a). Values in the shaded region (0.93cR<cr<cR0.93c_{\mathrm{R}}<c_{\mathrm{r}}<c_{\mathrm{R}}) are reported, despite the narrow scaling range in the shaded region shown in panel (a), to highlight the qualitative difference compared to ξG(cr→cR)→1\xi_G(c_{\mathrm{r}}\to c_{\mathrm{R}})\to1 (predicted by Eqs. (4) and (7)). (c) The ratio of ξˉ\bar{\xi}, predicted by Eq. (7), and ξ\xi, obtained by inverting ξG(ξ)\xi_G(\xi) of Eq. (4), vs. cr<0.93cRc_{\mathrm{r}}<0.93c_{\mathrm{R}}. ξˉ/ξ\bar{\xi}/\xi is close to unity across the entire range considered, with a small decline as cr→0.93cRc_{\mathrm{r}}\to0.93c_{\mathrm{R}}, see text.

While the values of Δ(cr)/δc\Delta(c_{\mathrm{r}})/\delta_{\mathrm{c}} in the shaded region become too small for a power-law to be meaningfully defined, we nevertheless plot ξG(cr)\xi_{\mathrm{G}}(c_{\mathrm{r}}) (the slope of the line in the double-logarithmic representation of Gf(δ)G_{\mathrm{f}}(\delta) that is used to determine Δ(cr)\Delta(c_{\mathrm{r}}) in Figure 3b. Recall, that Eq. (7) together with Eq. (4) predict ξG(cr→cR)→1\xi_{\mathrm{G}}(c_{\mathrm{r}}\to c_{\mathrm{R}})\to1. Yet, Figure 3b reveals no sign of such a behavior as cRc_{\mathrm{R}} is approached. Taken together, Figure 3a–Figure 3b strongly indicate that the UST fails close to cRc_{\mathrm{R}}, most likely due to frictional rate nonlinearity. To reiterate and highlight the success of the theory not very close to cRc_{\mathrm{R}}, already demonstrated in Figure 1–2, we denote by ξ(η)\xi^{(\eta)} the prediction in Eq. (7)—to distinguish it from ξ\xi that is obtained by inverting ξG\xi_{\mathrm{G}} of Eq. (4)—and plot their ratio in Figure 3c for 0.27cR<cr<0.93cR0.27c_{\mathrm{R}}<c_{\mathrm{r}}<0.93c_{\mathrm{R}}. It is observed that ξ(η)/ξ\xi^{(\eta)}/\xi is close to unity across the entire range considered, with a small decline as cr≃0.93cRc_{\mathrm{r}}\simeq0.93c_{\mathrm{R}} is approached, already alluded to above in relation to Fig. 2.

The results in Figure 3 clearly indicate that the UST is valid over a broad range of rupture propagation speeds, but fails as cr→cRc_{\mathrm{r}}\to c_{\mathrm{R}} due to an increasing importance of frictional rate nonlinearity. The most remarkable implication of these results is that there is no reason to expect the Rayleigh function R(cr)R(c_{\mathrm{r}}), and consequently its zero at cr=cRc_{\mathrm{r}}=c_{\mathrm{R}}, to appear in the nonlinear theory close

@page 5/17 starts_mid=1 ends_mid=1 numerical data and the success of the UST up to high propagation speeds. They also raise the following question: what does the UST predict for rupture propagation very close to cRc_{\mathrm{R}} and in the super-Rayleigh range?

Image not extracted

Figure 2 (PDF p. 5). Testing the unconventional singularities theory close to cRc_{\mathrm{R}}. Similarly to Figure 1, but for cr(t)≃0.93cRc_{\mathrm{r}}(t) \simeq0.93c_{\mathrm{R}}. (a) As in Figure 1d. (inset) As in Figure 1e. (b) As in Figure 1b. The obtained unconventional singularity order is ξ=−(0.34±0.03)\xi=-(0.34\pm0.03), where the upper value corresponds to panel (b) and the lower to the inset of panel (a), see text and SI Appendix for details and all extracted parameter values.

Approaching cRc_{\mathrm{R}} and super-Rayleigh propagation

To start addressing the central question posed above, we first analyze the prediction for ξ\xi in Eq. (7) and its crc_{\mathrm{r}} dependence as cRc_{\mathrm{R}} is approached. Since R(cr→cR)→0R(c_{\mathrm{r}}\to c_{\mathrm{R}})\to0, the right-hand-side (RHS) of Eq. (7) diverges in this limit, implying that the left-hand-side (LHS) cot⁡[πξ(cr)]∼1/sin⁡[πξ(cr)]\cot[\pi\xi(c_{\mathrm{r}})]\sim1/\sin[\pi\xi(c_{\mathrm{r}})] also diverges, while sin⁡[πξ(cr)]/R(cr)\sin[\pi\xi(c_{\mathrm{r}})]/R(c_{\mathrm{r}}) remains finite. Consequently, Eq. (7) predicts that ξ→0\xi\to0 for cr→cRc_{\mathrm{r}}\to c_{\mathrm{R}}, and that the slip rate in Eq. (2)—that is proportional to sin⁡[πξ(cr)]/R(cr)\sin[\pi\xi(c_{\mathrm{r}})]/R(c_{\mathrm{r}})—does not experience any divergence in this limit, in sharp contrast to the corresponding prediction of the classical theory with ξ=−1/2\xi=-1/2. These predictions assume that ηeff(cr)\eta_{\mathrm{eff}}(c_{\mathrm{r}}) reveals no special behavior in the vicinity of cRc_{\mathrm{R}}, which is expected on general grounds and verified (SI Appendix).

A vanishing singularity order, ξ→0\xi\to0, amounts to rupture propagation without stress amplification. This can be excluded on physical grounds since a stress amplification relative to the prestress in the close vicinity of the rupture edge is necessary for interfacial failure and for the propagation of frictional rupture. Consequently, the UST appears to signal its own breakdown in the cr→cRc_{\mathrm{r}}\to c_{\mathrm{R}} limit. Since Eq. (7) is equivalent to the frictional boundary condition σxy(r,θ=±π,t)=τ(v)\sigma_{xy}(r,\theta=\pm\pi,t)=\tau(v) under the linearization of τ(v)\tau(v) in Eq. (6), this breakdown suggests that nonlinearity in τ(v)\tau(v) becomes increasingly important in the cr→cRc_{\mathrm{r}}\to c_{\mathrm{R}} limit. In the presence of such nonlinearity, the power-law series assumption of the classical rupture theory cannot hold and ξ\xi can no longer be selected.

A direct and transparent way to probe the emergence of frictional rate nonlinearity in τ(v)\tau(v) in the cr→cRc_{\mathrm{r}}\to c_{\mathrm{R}} limit is through the behavior of Gf(δ)G_{\mathrm{f}}(\delta) for δ>δc\delta>\delta_{\mathrm{c}}. On the one hand, the UST in Eq. (4) predicts that Gf(δ)G_{\mathrm{f}}(\delta) varies as a power-law with ξG(ξ→0)→1\xi_{\mathrm{G}}(\xi\to0)\to1 in this limit. On the other hand, concave considerations suggest that Gf(δ)G_{\mathrm{f}}(\delta) will not reveal a clean power-law with ξG→1\xi_{\mathrm{G}}\to1. To quantitatively distinguish between the two, we quantify the power-law scaling range of Gf(δ)G_{\mathrm{f}}(\delta), denoted by Δ\Delta in Figure 1b and Fig. 2b, where the emergence of frictional rate nonlinearity in τ(v)\tau(v) would be manifested in a significant reduction in Δ\Delta.

In Figure 3a, we plot Δ(cr)/δc\Delta(c_{\mathrm{r}})/\delta_{\mathrm{c}} that is quantified exactly as in Figure 1b and Fig. 2b, i.e., by estimating the δ\delta range over which Gf(δ)G_{\mathrm{f}}(\delta) follows a straight line for δ>δc\delta>\delta_{\mathrm{c}} in a double-logarithmic representation (see SI Appendix). It is observed that Δ(cr)/δc\Delta(c_{\mathrm{r}})/\delta_{\mathrm{c}} slowly decreases in the range 0.27cR<cr<0.93cR0.27c_{\mathrm{R}}<c_{\mathrm{r}}<0.93c_{\mathrm{R}} (with some variation within each dataset, see caption), where we already established the validity of the UST, but rather strongly decreases for 0.93cR<cr<cR0.93c_{\mathrm{R}}<c_{\mathrm{r}}<c_{\mathrm{R}} (shaded region).

While the values of Δ(cr)/δc\Delta(c_{\mathrm{r}})/\delta_{\mathrm{c}} in the shaded region become too small for a power-law to be meaningfully defined, we nevertheless plot ξG(cr)\xi_{\mathrm{G}}(c_{\mathrm{r}}) (the slope of the line in the double-logarithmic representation of Gf(δ)G_{\mathrm{f}}(\delta) that is used to determine Δ(cr)\Delta(c_{\mathrm{r}}) in Figure 3b. Recall, that Eq. (7) together with Eq. (4) predict ξG(cr→cR)→1\xi_{\mathrm{G}}(c_{\mathrm{r}}\to c_{\mathrm{R}})\to1. Yet, Figure 3b reveals no sign of such a behavior as cRc_{\mathrm{R}} is approached. Taken together, Figure 3a–Figure 3b strongly indicate that the UST fails close to cRc_{\mathrm{R}}, most likely due to frictional rate nonlinearity. To reiterate and highlight the success of the theory not very close to cRc_{\mathrm{R}}, already demonstrated in Figure 1–2, we denote by ξ(η)\xi^{(\eta)} the prediction in Eq. (7)—to distinguish it from ξ\xi that is obtained by inverting ξG\xi_{\mathrm{G}} of Eq. (4)—and plot their ratio in Figure 3c for 0.27cR<cr<0.93cR0.27c_{\mathrm{R}}<c_{\mathrm{r}}<0.93c_{\mathrm{R}}. It is observed that ξ(η)/ξ\xi^{(\eta)}/\xi is close to unity across the entire range considered, with a small decline as cr≃0.93cRc_{\mathrm{r}}\simeq0.93c_{\mathrm{R}} is approached, already alluded to above in relation to Fig. 2.

The results in Figure 3 clearly indicate that the UST is valid over a broad range of rupture propagation speeds, but fails as cr→cRc_{\mathrm{r}}\to c_{\mathrm{R}} due to an increasing importance of frictional rate nonlinearity. The most remarkable implication of these results is that there is no reason to expect the Rayleigh function R(cr)R(c_{\mathrm{r}}), and consequently its zero at cr=cRc_{\mathrm{r}}=c_{\mathrm{R}}, to appear in the nonlinear theory close to the cRc_{\mathrm{R}}. In other words, the Rayleigh wave-speed cRc_{\mathrm{R}} is not expected to be a ‘special propagation speed’ anymore and likewise one expects that G>0G>0 in the super-Rayleigh range of speeds cR<cr<cSc_{\mathrm{R}}<c_{\mathrm{r}}<c_{\mathrm{S}}. If so, frictional rupture propagation in the “forbidden” super-Rayleigh range should be possible. Indeed, this is demonstrated in Figure 4, where rupture smoothly and continuously propagates into the “forbidden” super-Rayleigh range for various prestress levels τd\tau_{\mathrm{d}} (the inset presents v(r)v(r) as cRc_{\mathrm{R}} is crossed for one τd\tau_{\mathrm{d}} value, see caption). This is a central finding of this study.

Image not extracted

Figure 4. Super-Rayleigh and super-shear frictional rupture propagation. cr(t)c_{\mathrm{r}}(t) obtained for different prestress levels τd\tau_{\mathrm{d}} (increasing from the bottom curve to the top one), see SI Appendix, revealing smooth and continuous frictional rupture propagation into and inside the super-Rayleigh range. The top curve reveals a continuous super-shear crossover (note that cS=1.074cRc_{\mathrm{S}}=1.074c_{\mathrm{R}} for the used elastic moduli). (inset) v(r)v(r) for the three crc_{\mathrm{r}} values marked on the second curve from the bottom, see legend. The smallest crc_{\mathrm{r}} is somewhat below cRc_{\mathrm{R}} (the same as in Fig. 2a), the middle one is essentially at cRc_{\mathrm{R}} and the largest features cr>cRc_{\mathrm{r}}>c_{\mathrm{R}}. Note that there is no divergence and change in the sign of v(r)v(r) upon crossing cRc_{\mathrm{R}}, in contrast to the classical theory prediction in Eq. (2), once ξ=−1/2\xi=-1/2 is substituted therein.

A continuous super-shear crossover

The results discussed so far demonstrated that the generic rate dependence of the frictional strength gives rise to the breakdown of the classical 2D rupture theory and that the linear approximation to τ(v)\tau(v) in Eq. (6) opens the way to the development of the UST. The latter theory has been comprehensively shown to be in quantitative agreement with extensive numerical data up to rupture propagation speeds close to cRc_{\mathrm{R}}. Yet, very close to cRc_{\mathrm{R}}, the UST breaks down in itself, most likely due to nonlinearity in τ(v)\tau(v). Consequently, frictional rupture crosses cRc_{\mathrm{R}} and propagates inside the “forbidden” super-Rayleigh range.

An important corollary of this major observation is that there should be no discontinuous/sharp super-shear transition in our numerical simulations, i.e., that rupture also smoothly and continuously propagates into the super-shear regime. This expectation is explicitly supported by the cr(t)c_{\mathrm{r}}(t) curve for the largest τd\tau_{\mathrm{d}} value in Figure 4, yet again demonstrating the potentially profound implications of frictional rate dependence for fast earthquake propagation. Continuous rupture propagation through the super-Rayleigh range into the super-shear one has been documented in experiments [33] and numerical simulations [75, 76, 77, 78, 79], where it has been termed the “direct transition”, though it has not been fully understood theoretically.

Summary and conclusion

The overarching theme of this work is to revisit the assumptions underlying the classical 2D theory of frictional rupture, with a focus on the roles played by a nonlinear rate-dependence frictional strength τ(v)\tau(v), which is experimentally established. While we consider a broad range of rupture propagation speeds, special attention is given to high propagation speeds, where the classical theory predicts the existence of a “forbidden” super-Rayleigh range, which raises the fundamental question of how rupture can transition into the super-shear range. We explain that the classical theory is based on the leading-order approximation of τ(v)\tau(v), simply a rate-independent constant (the residual stress τ0\tau_0), combined with the assumption that the linear elastodynamic bulk equations admit a power-law series solution, with a singular contribution. The latter is the famous ξ=−1/2\xi=-1/2 singularity of the classical theory.

By relaxing the first assumption, i.e., by considering a perturbative expansion of τ(v)\tau(v) up to the linear term in vv, we develop the UST, which offers analytic predictions for frictional rupture with a non-universal singularity order ξ≠−1/2\xi\ne-1/2. Using extensive numerical simulations, we demonstrated that the UST is in quantitative agreement with numerical observations over a broad range of propagation speed with ξ≠−1/2\xi\ne-1/2, including close to the Rayleigh wave-speed cRc_{\mathrm{R}}. These results demonstrate the breakdown the classical 2D theory of frictional rupture due to the rate-dependence of frictional strength τ(v)\tau(v), approximated to linear order in the slip rate vv. Recent experiments also support the UST in the sub-Rayleigh range [71, 72, 73].

Yet, as cRc_{\mathrm{R}} is approached, the range of validity of the linear approximation to τ(v)\tau(v) shrinks and the UST breaks down in itself due to nonlinearity in τ(v)\tau(v), which most likely invalidates the power-law series solution assumption, i.e., a singular exponent ξ\xi can no longer be selected. As a consequence, frictional rupture smoothly and continuously propagates into and inside the so-called “forbidden” super-Rayleigh range. This finding has an immediate and important implication: in contrast to prediction of the classical theory, frictional rupture can also smoothly and continuously propagate into the super-shear range, i.e., surpassing the shear wave-speed csc_{\mathrm{s}} with no discontinuous transition, as we explicitly demonstrated. A similar behavior has been documented in a laboratory experiment [33].

Our results about the nature of the emergence of super-shear frictional rupture are not necessarily universal, i.e., some aspects of them might depend on properties of τ(v,…)\tau(v,\ldots), where the ellipsis stands for interfacial state fields. Future work should elucidate what properties of the interfacial constitutive law give rise to qualitative deviations from the predictions of the classical 2D rupture theory. On the other hand, our results show that the classical 2D rupture theory — and specifically the existence of a “forbidden” super-Rayleigh range — are not universal. Future work should consider applications of the developed theory and its predictions to geophysical field and laboratory observations, e.g., in relation to recent observations [33, 80, 81], as well as extensions to 3D.

Author contributions — E.B. conceived and designed research, developed the theory and wrote the paper; A.P. performed all numerical simulations, the comparison to the theory and generated all figures; F.B. contributed to the theory development and provided support to the simulations; all authors discussed the results and contributed to the writing.

Acknowledgements— We thank Efim Brener and Alina Shafir for useful discussions. E.B. acknowledges support from the Minerva Foundation (with funding from the Federal German Ministry for Education and Research) and the Harold Perlman Family. F.B. acknowledges support from the Research Council of Norway through its Researcher Project for Young Talents program (UN-LOC, project 345008). The computations were carried out at Weizmann Institute of Science, using the Faculty of Chemistry’s high-performance computing facility CHEMFARM, which is supported in part by the Ben May Center for Chemical Theory and Computation.

The goal of this document is to provide technical details about the theoretical and numerical results presented in the manuscript.

The unconventional singularities theory (UST)

The development of the theory starts, as explained in the manuscript, from the linear momentum balance equation under plane-strain conditions

ρu¨−∇⋅σ=0(8)\rho\ddot{\mathbf{u}}-\nabla\cdot\boldsymbol{\sigma}=0 \tag*{(8)}

valid for each of the bodies forming the fault. Here, ρ\rho being the mass density, σ\boldsymbol{\sigma} the stress tensor, u(x,y,t)\mathbf{u}(x,y,t) is the 2D displacement vector and each superposed dot denotes a partial time derivative. Based on the Helmholtz decomposition theorem, any vector field can be expressed as the sum of two contributions derived from a scalar potential ϕ\phi and a vectorial potential Ψ\boldsymbol{\Psi}, respectively, according to

u=∇ϕ+∇×Ψ.(9)\mathbf{u}=\nabla\phi+\nabla\times\boldsymbol{\Psi}. \tag*{(9)}

Using Hooke’s law for isotropic elastic bodies, the Helmholtz decomposition in Eq. (9) can be used to rewrite the momentum balance equation as a set of two wave equations

cd2∇2ϕ=ϕ¨,cs2∇2Ψ=Ψ¨.(10)c_{\mathrm{d}}^{2}\nabla^{2}\phi=\ddot{\phi},\qquad c_{\mathrm{s}}^{2}\nabla^{2}\boldsymbol{\Psi}=\ddot{\boldsymbol{\Psi}}. \tag*{(10)}

Equations (S3) imply that isovolumetric deformation propagates at the shear wave-speed csc_{\mathrm{s}}, whereas irrotational deformation travels at the dilatational wave-speed cdc_{\mathrm{d}}, both expressed in terms of Poisson’s ratio ν\nu and the shear modulus μ\mu (appearing in Hooke’s law). Next, we consider a dynamic earthquake rupture propagating in the xx direction at speed cr(t)c_{\mathrm{r}}(t) along the fault separating the two linear elastic solids (located at y=0y=0).

We introduce a Cartesian frame of reference defined by the unit vectors x^\hat{\mathbf{x}} and y^\hat{\mathbf{y}}, whose directions are respectively parallel to the rupture propagation direction and normal to the fault plane, and assume plane-strain conditions (as noted above) such that Ψ=ψz^\boldsymbol{\Psi}=\psi\hat{\mathbf{z}}, with the unit vector z^\hat{\mathbf{z}} being perpendicular to the x^\hat{\mathbf{x}}-y^\hat{\mathbf{y}} plane. We invoke the Lorentz transform, which preserves the shape of wave equations and reads

x−crtαs,d=rs,dcos⁡(θs,d),y=rs,dsin⁡(θs,d),(11)\frac{x-c_{\mathrm{r}}t}{\alpha_{\mathrm{s,d}}}=r_{\mathrm{s,d}}\cos(\theta_{\mathrm{s,d}}),\qquad y=r_{\mathrm{s,d}}\sin(\theta_{\mathrm{s,d}}), \tag*{(11)}

for a Cartesian frame of reference co-moving with the rupture edge. In Eq. (11), we used the definition αs,d=1−cr2/cs,d2\alpha_{\mathrm{s,d}}=\sqrt{1-c_{\mathrm{r}}^{2}/c_{\mathrm{s,d}}^{2}}. Equations (S4) actually correspond to two “relativistic” frames of reference, where distances along the propagation direction significantly contract as the rupture speed approaches cs,dc_{\mathrm{s,d}}, respectively.

rs,d=rαs,d1−cr2sin⁡2(θ)cs,d2≡rαs,dγs,d,(12)r_{\mathrm{s,d}}=\frac{r}{\alpha_{\mathrm{s,d}}}\sqrt{1-\frac{c_{\mathrm{r}}^{2}\sin^{2}(\theta)}{c_{\mathrm{s,d}}^{2}}}\equiv\frac{r}{\alpha_{\mathrm{s,d}}}\gamma_{\mathrm{s,d}}, \tag*{(12)}
θs,d=tan⁡−1(αs,dtan⁡(θ)),\theta_{\mathrm{s,d}}=\tan^{-1}\left(\alpha_{\mathrm{s,d}}\tan(\theta)\right),

with rr denoting the distance from the rupture edge and θ\theta the angle relative to the propagation direction. Next, we assume that time variations are slaved to the smoothly time-varying rupture speed cr(t)c_{\mathrm{r}}(t), such that steady-state solutions can be searched in the co-moving frame of reference, i.e. ∂t=0\partial_t=0. In fact, as shown in [10] and noted in the manuscript, the singular contribution to the solution (which is our main focus, see below and the manuscript), is valid for any cr(t)c_{\mathrm{r}}(t), without invoking the quasi-steady-state assumption.

Consequently, the wave equations (S3) become a set of Laplace equations for ϕ\phi and ψ\psi (recall that Ψ=ψz^\boldsymbol{\Psi}=\psi\hat{\mathbf{z}}). As explained in the manuscript, it is then assumed that these equations admit a power-law series solution of the form ∇u=rd,sξF(θd,s,ξ)\nabla u=r_{\mathrm{d,s}}^{\xi}\mathcal{F}(\theta_{\mathrm{d,s}},\xi), with F(⋅)\mathcal{F}(\cdot) being a function to be determined, leading to a solution of the form

ϕ(rd,θd)=rdξ+2(Asin⁡((ξ+2)θd)+Bcos⁡((ξ+2)θd)),\phi(r_{\mathrm{d}},\theta_{\mathrm{d}})=r_{\mathrm{d}}^{\xi+2}\left(A\sin\left((\xi+2)\theta_{\mathrm{d}}\right)+B\cos\left((\xi+2)\theta_{\mathrm{d}}\right)\right),
ψ(rs,θs)=rsξ+2(Csin⁡((ξ+2)θs)+Dcos⁡((ξ+2)θs)).(13)\psi(r_{\mathrm{s}},\theta_{\mathrm{s}})=r_{\mathrm{s}}^{\xi+2}\left(C\sin\left((\xi+2)\theta_{\mathrm{s}}\right)+D\cos\left((\xi+2)\theta_{\mathrm{s}}\right)\right). \tag*{(13)}

The relations between the constants A,B,C,DA,B,C,D and the exponent ξ\xi can be determined by the boundary conditions. Hereafter, we assume mode-II symmetry, i.e., ux(rs,d,−θs,d)=−ux(rs,d,θs,d)u_x(r_{\mathrm{s,d}},-\theta_{\mathrm{s,d}})=-u_x(r_{\mathrm{s,d}},\theta_{\mathrm{s,d}}), which implies that B=C=0B=C=0. The resulting mode-II displacement solution reads

ux=(ξ+2)(rdξ+1αd−1Asin⁡((ξ+1)θd)−rsξ+1αs−1Dsin⁡((ξ+1)θs)),u_x=(\xi+2)\left(r_{\mathrm{d}}^{\xi+1}\alpha_{\mathrm{d}}^{-1}A\sin\left((\xi+1)\theta_{\mathrm{d}}\right)-r_{\mathrm{s}}^{\xi+1}\alpha_{\mathrm{s}}^{-1}D\sin\left((\xi+1)\theta_{\mathrm{s}}\right)\right),
uy=(ξ+2)(rdξ+1Acos⁡((ξ+1)θd)−rsξ+1αs−1Dcos⁡((ξ+1)θs)).(14)u_y=(\xi+2)\left(r_{\mathrm{d}}^{\xi+1}A\cos\left((\xi+1)\theta_{\mathrm{d}}\right)-r_{\mathrm{s}}^{\xi+1}\alpha_{\mathrm{s}}^{-1}D\cos\left((\xi+1)\theta_{\mathrm{s}}\right)\right). \tag*{(14)}

Consider then a constant normal traction along the fault and assume that the shear traction is dominated by a singular stress contribution of the form

σxy(r,θd=0)=KII(ξ)rξ2π,(15)\sigma_{xy}(r,\theta_{\mathrm{d}}=0)=K_{\mathrm{II}}^{(\xi)}\frac{r^{\xi}}{\sqrt{2\pi}}, \tag*{(15)}

where KII(ξ)K_{\mathrm{II}}^{(\xi)} is a generalized stress intensity factor. Using the above, the ratio between AA and DD can be determined and one ends up with

ux(r,θ)=rξ+1KII(ξ)(ξ+1)μR(cr)2π×(2γdξ+1αssin⁡[(ξ+1)θd]−γsξ+1αs(1+αs2)sin⁡[(ξ+1)θs]),(16)u_x(r,\theta)=\frac{r^{\xi+1}K_{\mathrm{II}}^{(\xi)}}{(\xi+1)\mu R(c_r)\sqrt{2\pi}}\times\left(2\gamma_d^{\xi+1}\alpha_s\sin[(\xi+1)\theta_d]-\gamma_s^{\xi+1}\alpha_s(1+\alpha_s^2)\sin[(\xi+1)\theta_s]\right), \tag*{(16)}
uy(r,θ)=rξ+1KII(ξ)(ξ+1)μR(cr)2π×(2γdξ+1αdαscos⁡[(ξ+1)θd]−γsξ+1(1+αs2)cos⁡[(ξ+1)θs]),(17)u_y(r,\theta)=\frac{r^{\xi+1}K_{\mathrm{II}}^{(\xi)}}{(\xi+1)\mu R(c_r)\sqrt{2\pi}}\times\left(2\gamma_d^{\xi+1}\alpha_d\alpha_s\cos[(\xi+1)\theta_d]-\gamma_s^{\xi+1}(1+\alpha_s^2)\cos[(\xi+1)\theta_s]\right), \tag*{(17)}

for the displacement, with

u˙x(r,θ)=KII(ξ)2πrξcrαsμR(cr)×[−2γdξsin⁡(ξθd)+(1+αs2)γsξsin⁡(ξθs)],\dot{u}_x(r,\theta)=\frac{K_{\mathrm{II}}^{(\xi)}}{\sqrt{2\pi}}r^\xi\frac{c_r\alpha_s}{\mu R(c_r)}\times\left[-2\gamma_d^\xi\sin(\xi\theta_d)+(1+\alpha_s^2)\gamma_s^\xi\sin(\xi\theta_s)\right],
u˙y(r,θ)=KII(ξ)2πrξcrμR(cr)×[−2αdαsγdξcos⁡(ξθd)+(1+αs2)γsξsin⁡(ξθs)],(18)\dot{u}_y(r,\theta)=\frac{K_{\mathrm{II}}^{(\xi)}}{\sqrt{2\pi}}r^\xi\frac{c_r}{\mu R(c_r)}\times\left[-2\alpha_d\alpha_s\gamma_d^\xi\cos(\xi\theta_d)+(1+\alpha_s^2)\gamma_s^\xi\sin(\xi\theta_s)\right], \tag*{(18)}

for the particle velocity, and

σxy(r,θ)=KII(ξ)2πrξR(cr)×[4αdαsγdξcos⁡(ξθd)−(1+αs2)2γsξcos⁡(ξθs)],\sigma_{xy}(r,\theta)=\frac{K_{\mathrm{II}}^{(\xi)}}{\sqrt{2\pi}} \frac{r^\xi}{R(c_r)}\times\left[4\alpha_d\alpha_s\gamma_d^\xi\cos(\xi\theta_d)-(1+\alpha_s^2)^2\gamma_s^\xi\cos(\xi\theta_s)\right],
σyy(r,θ)=KII(ξ)2πrξ2αs(1+αs2)R(cr)×[−γdξsin⁡(ξθd)+γsξsin⁡(ξθs)],(19)\sigma_{yy}(r,\theta)=\frac{K_{\mathrm{II}}^{(\xi)}}{\sqrt{2\pi}}\frac{r^\xi2\alpha_s(1+\alpha_s^2)}{R(c_r)}\times\left[-\gamma_d^\xi\sin(\xi\theta_d)+\gamma_s^\xi\sin(\xi\theta_s)\right], \tag*{(19)}

for the stress. The Rayleigh function reads

R(cr)=4αdαs−(1+αs2)2,(20)R(c_r)=4\alpha_d\alpha_s-(1+\alpha_s^2)^2, \tag*{(20)}

featuring a non-trivial root R(cr=cR)=0R(c_r=c_R)=0 that defines the Rayleigh wave-speed cRc_R. Equations (1)–(3) in the manuscript then follow, where the slip rate is v(r,t)=2u˙x(r,θ=π,t)v(r,t)=2\dot{u}_x(r,\theta=\pi,t).

We next consider the energy associated with rupture propagation. In the co-moving frame of reference, the mechanical energy corresponds to the potential energy Π\Pi. This useful property enables one to apply the original approach proposed by Irwin [82] for quasi-static crack, see Figure S1, to our problem. Irwin considered a pair of opposite tractions applied over a distance Δa\Delta a to initially maintain the crack surfaces closed. In this context, the change in potential energy dΠ(Δa)\mathrm{d}\Pi(\Delta a) resulting from rupture propagation over Δa\Delta a can be expressed as the work done by surface tractions t=σ⋅n\mathbf{t}=\boldsymbol{\sigma}\cdot\mathbf{n} along the fault

Image not extracted

Figure S1. Illustration of Irwin’s original approach for mode-I (tensile) rupture [82] to compute the change of potential energy due to tensile crack advance, i.e., between states (A) and (B). In order to keep the boundary surfaces unchanged, Irwin assumed that a pair of opposite normal tractions (blue arrows) maintains the crack closed over the surface increment Δa\Delta a and mimics the effect of the material’s cohesion. We follow the same approach to integrate the energy release rate of mode-II rupture governed by unconventional singularities.

dΠ(Δa)=−12∫Δad(t⋅u) dS.(21)\mathrm{d}\Pi(\Delta a)=-\frac{1}{2}\int_{\Delta a}\mathrm{d}(\mathbf{t}\cdot\mathbf{u})\,\mathrm{d}S. \tag*{(21)}

Note that Eq. (21) integrates only the flux of mechanical energy towards the near-edge portion of the fault plane of size Δa\Delta a and not over the entire fault surface SS.

Following Irwin’s approach, we consider a pair of opposite shear tractions that exactly balance the singular shear stress existing ahead of the rupture edge, i.e., tx=σxy(rs,d,θs,d=0)t_x=\sigma_{xy}(r_{s,d},\theta_{s,d}=0), and prevent frictional slip. We underline that Eq. (21) has to be evaluated independently for the isovolumetric (shear) and irrotational (dilatational) parts of the solution as they reside in different frame of reference. As a result, the integration of Eq. (21) should also account for the work of tensile traction arising behind the rupture edge, since ty=σyy(rs,d,θs,d=π)t_y=\sigma_{yy}(r_{s,d},\theta_{s,d}=\pi) has finite isovolumetric and irrotational contributions and only their sum is zero. Keeping these properties in mind, the mechanical energy released by growing the rupture over Δa\Delta a can be obtained as

dΠ(Δa)=∑i=s,d∫0Δa[σxy(ri=X;θi=0)ux(ri=Δa−X;θi=π)−uy(ri=X;θi=0)σyy(ri=Δa−X;θi=π)]dX=−sin⁡(−πξ)cr22π(ξ+1)μR(cr)2cs2(Δa)2(ξ+1)Γ(ξ+1)4ξ+1Γ(ξ+32)(4αs2αd2(ξ+1)−αs2(ξ+1)(1+αs2)2)[KII(ξ)]2,(22)\begin{aligned} \mathrm{d}\Pi(\Delta a)=\sum_{i=s,d}\int_0^{\Delta a}\left[\sigma_{xy}(r_i=\mathcal{X};\theta_i=0)u_x(r_i=\Delta a-\mathcal{X};\theta_i=\pi)-u_y(r_i=\mathcal{X};\theta_i=0)\sigma_{yy}(r_i=\Delta a-\mathcal{X};\theta_i=\pi)\right]\mathrm{d}\mathcal{X} \\ &=-\frac{\sin(-\pi\xi)c_r^2}{2\sqrt{\pi}(\xi+1)\mu R(c_r)^2c_s^2}\frac{(\Delta a)^{2(\xi+1)}\Gamma(\xi+1)}{4^{\xi+1}\Gamma\left(\xi+\frac{3}{2}\right)}\left(4\alpha_s^2\alpha_d^{2(\xi+1)}-\alpha_s^{2(\xi+1)}(1+\alpha_s^2)^2\right)\left[K_{\mathrm{II}}^{(\xi)}\right]^2, \tag*{(22)} \end{aligned}

where Γ\Gamma denotes the Gamma function.

The classical rupture theory and generalized energy balance

Before we derive the generalized energy balance for frictional rupture characterized by an unconventional singularity, let us briefly obtain the results corresponding to the classical rupture theory [10]. As discussed extensively in the manuscript, the classical rupture theory is recovered when Eq. (6) in the manuscript is truncated to leading order (the rate-independent limit of the frictional strength τ\tau), i.e., τ(v)≃τ0\tau(v) \simeq\tau_0, where τ0\tau_0 is a constant residual stress. Imposing the boundary condition σxy(r,θ=±π,t)=τ0\sigma_{xy}(r,\theta=\pm\pi,t)=\tau_0, we conclude that the singular contribution in Eqs. (19) must vanish behind the edge, which is ensured by the conventional singularity, ξ=−1/2\xi=-1/2. This is equivalent to setting the RHS of Eq. (7) in the manuscript to zero.

Using then ξ=−1/2\xi=-1/2 in Eqs. (19), one recovers the classical square-root singular fields [10]

σxy(r,θ)=KII2πr1R[4αdαscos⁡12θdγd−(1+αs2)2cos⁡12θsγs],\sigma_{xy}(r,\theta)=\frac{K_{\mathrm{II}}}{\sqrt{2\pi r}}\frac{1}{R}\left[4\alpha_d\alpha_s\frac{\cos\frac{1}{2}\theta_d}{\sqrt{\gamma_d}}-(1+\alpha_s^2)^2\frac{\cos\frac{1}{2}\theta_s}{\gamma_s}\right],
σyy(r,θ)=KII2πr2αs(1+αs2)R[sin⁡12θdγd−sin⁡12θsγs],(23)\sigma_{yy}(r,\theta)=\frac{K_{\mathrm{II}}}{\sqrt{2\pi r}}\frac{2\alpha_s(1+\alpha_s^2)}{R}\left[\frac{\sin\frac{1}{2}\theta_d}{\gamma_d}-\frac{\sin\frac{1}{2}\theta_s}{\gamma_s}\right], \tag*{(23)}

where KII=KII(ξ=−1/2)K_{\mathrm{II}}=K_{\mathrm{II}}^{(\xi=-1/2)}. The computation of the energy release rate GG [10] then follows from Eq. (22)

G≡lim⁡Δa→0(−dΠ(Δa)Δa)=1−ν2EAII(cr,ξ=−1/2)KII2,(24)G\equiv\lim_{\Delta a\to0}\left(-\frac{\mathrm{d}\Pi(\Delta a)}{\Delta a}\right)=\frac{1-\nu^2}{E}A_{\mathrm{II}}(c_r,\xi=-1/2)K_{\mathrm{II}}^2, \tag*{(24)}

where E=2μ(1+ν)E=2\mu(1+\nu) is Young’s modulus and [10]

AII(cr,ξ=−1/2)=cr2αs(1−ν)R(cr)cs2.(25)A_{\mathrm{II}}(c_r,\xi=-1/2)=\frac{c_r^2\alpha_s}{(1-\nu)R(c_r)c_s^2}. \tag*{(25)}

To obtain the energy release rate GG in the generalized case of an unconventional singularity ξ≠−1/2\xi\ne-1/2, note that Δa\Delta a can no longer be arbitrarily small and one needs to introduce a finite cohesive zone of size ℓc\ell_c, as done in the manuscript. Consequently, we obtain

G=lim⁡Δa→ℓc(−dΠ(Δa)Δa)=1−ν2E[KII(ξ)]2AII(cr,ξ)ℓc2ξ+1,(26)G=\lim_{\Delta a\to\ell_c}\left(-\frac{\mathrm{d}\Pi(\Delta a)}{\Delta a}\right)=\frac{1-\nu^2}{E}\left[K_{\mathrm{II}}^{(\xi)}\right]^2 A_{\mathrm{II}}(c_r,\xi)\ell_c^{2\xi+1}, \tag*{(26)}

which identifies with Eq. (5) in the manuscript. The generalized function AII(cr,ξ)A_{\mathrm{II}}(c_r,\xi) reads

AII(cr,ξ)=(crcs)2(4αs2αd2(ξ+1)−αs2(ξ+1)(1+αs2)2)×sin⁡(−πξ)(1−ν)[R(cr)]2Γ(ξ+1)4ξ+1π(ξ+1)Γ(ξ+3/2).(27)A_{\mathrm{II}}(c_r,\xi)=\left(\frac{c_r}{c_s}\right)^2\left(4\alpha_s^2\alpha_d^{2(\xi+1)}-\alpha_s^{2(\xi+1)}(1+\alpha_s^2)^2\right)\times\frac{\sin(-\pi\xi)}{(1-\nu)[R(c_r)]^2}\frac{\Gamma(\xi+1)}{4^{\xi+1}\sqrt{\pi}(\xi+1)\Gamma(\xi+3/2)}. \tag*{(27)}

Note that the Rayleigh function R(cr)R(c_r) enters the denominator of Eq. (27) quadratically, in contrast to the classical result in Eq. (25), where it appears linearly in the denominator.

As discussed in the manuscript, the latter — specifically R(cr=cR)=0R(c_r=c_R)=0 and the sign change of R(cr)R(c_r) for cr>cRc_r>c_R — is the origin of the “forbidden” super-Rayleigh range of propagation wave-speeds in the classical theory. On the other hand, the generalized AII(cr,ξ)A_{\mathrm{II}}(c_r,\xi) in Eq. (27) does not change sign across cRc_R due to R(cr)R(c_r) for ξ≠−1/2\xi\ne-1/2.

Finally, we consider the so-called breakdown energy that is defined in the manuscript as

Gf(δ)≡∫0δ[τ(δ′)−τ(δ)] dδ′,(28)G_f(\delta)\equiv\int_0^\delta[\tau(\delta')-\tau(\delta)]\,\mathrm{d}\delta', \tag*{(28)}

which integrates the local frictional dissipation on top of residual friction, where δ(r)=2ux(r,θ=π)\delta(r)=2u_x(r,\theta=\pi) is the fault slip. The edge-localized dissipation GcG_c corresponds to Gf(δ=δc)=GcG_f(\delta=\delta_c)=G_c, where δc\delta_c is the fault slip associated with ℓc\ell_c. Using then Eqs. (S9) and (S11), one obtains δ(r)∼rξ+1\delta(r)\sim r^{\xi+1} and τ(r)=σxy(r,θ=π)∼πrξ∼δξξ+1\tau(r)=\sigma_{xy}(r,\theta=\pi)\sim\pi r^\xi\sim\delta^{\frac{\xi}{\xi+1}} behind the rupture edge, such that the breakdown energy follows

Gf(δ)=Gc(δδc)1+2ξ1+ξ,(29)G_f(\delta)=G_c\left(\frac{\delta}{\delta_c}\right)^{\frac{1+2\xi}{1+\xi}}, \tag*{(29)}

for δ>δc\delta>\delta_c, which identifies with Eq. (4) in the manuscript. The exponent ξG(ξ)=(1+2ξ)/(1+ξ)\xi_G(\xi)=(1+2\xi)/(1+\xi) is extensively discussed in the manuscript and below.

Numerical simulations

The boundary integral method approach

The linear elastic bodies forming the fault (located at y=0y=0) are assumed in our problem to be infinite in the direction perpendicular to it. Under this condition, the interfacial shear stress σxy(x,t)\sigma_{xy}(x,t) can be computed using the boundary integral approach, where the fault boundary condition τ=σxy(x,t)\tau=\sigma_{xy}(x,t) takes the form [83]

τ[v(x,t),ϕ(x,t)]=τd−μ2csv(x,t)+s(x,t).(30)\tau[v(x,t),\phi(x,t)]=\tau_d-\frac{\mu}{2c_s}v(x,t)+s(x,t). \tag*{(30)}

Here, τd\tau_d is the background stress (prestress) applied far from the fault (the subscript ‘d’ stands for ‘driving’) and (μ/2cs)v(x,t)(\mu/2c_s)v(x,t) is the so-called radiation damping term, which depends on the local slip rate v(x,t)v(x,t). s(x,t)s(x,t) is a spatiotemporal convolutional integral which accounts for the long-range interaction of different parts of the fault, mediated by bulk deformation. In general, s(x,t)s(x,t) does not admit an explicit real-space representation, but is rather expressed in Fourier space [84].

The interfacial shear stress, corresponding to the RHS of Eq. (30), is balanced at each point in time and fault position by the frictional strength τ[v(x,t),ϕ(x,t)]\tau[v(x,t),\phi(x,t)], which depends in addition to v(x,t)v(x,t) on the internal state field ϕ(x,t)\phi(x,t) that admits an evolution equation; both are specified in the next section. Once specified, we solve the problem numerically using the open-source library cRacklet [74], which is based on a spectral representation of Eq. (30). It involves a finite fault of size WW and periodic boundary conditions such that the basic fields in the problem satisfy v(x,t)=v(x+W,t)v(x,t)=v(x+W,t) and ϕ(x,t)=ϕ(x+W,t)\phi(x,t)=\phi(x+W,t). The problem formulation is completed once the initial conditions v(x,t=0)v(x,t=0) and ϕ(x,t=0)\phi(x,t=0) are specified, to be discussed below.

The fault constitutive relation (the friction law)

We employed a nonlinear rate-and-state dependent frictional strength τ[v(x,t),ϕ(x,t)]\tau[v(x,t),\phi(x,t)], which is based on extensive laboratory experiments [42, 45, 47], taking the form [85]

τ(v,ϕ)σ=[1+blog⁡(1+ϕϕ⋆)]×[f01+(v⋆/v)2+αlog⁡(1+∣v∣v⋆)](31)\frac{\tau(v,\phi)}{\sigma}=\left[1+b\log\left(1+\frac{\phi}{\phi^\star}\right)\right]\times\left[\frac{f_0}{\sqrt{1+(v^\star/v)^2}}+\alpha\log\left(1+\frac{|v|}{v^\star}\right)\right] \tag*{(31)}

The values of all parameters appearing in Eq. (31) can be found in Table S-1, along with the value of the employed normal stress σ\sigma and of the linear elastic parameters. The internal state field ϕ(x,t)\phi(x,t), which quantifies the amount of real contact area, satisfies the differential equation (known as the ‘aging-law’ [42, 45])

∂tϕ(x,t)=1−v(x,t)ϕ(x,t)D,(32)\partial_t\phi(x,t)=1-\frac{v(x,t)\phi(x,t)}{D}, \tag*{(32)}

where DD is the characteristic slip distance, whose value is also given in Table S-1.

Under persistent sliding at a slip rate vssv_{\mathrm{ss}}, Eq. (32) attains a steady state of the form ϕ=D/vss\phi=D/v_{\mathrm{ss}}. When the latter is substituted in τ(v,ϕ)\tau(v,\phi) of Eq. (31), the steady-state friction curve is obtained

τss(v)σ=[1+blog⁡(1+Dvssϕ⋆)]×[f01+(v⋆/vss)2+αlog⁡(1+∣vss∣v⋆)](33)\frac{\tau_{\mathrm{ss}}(v)}{\sigma}=\left[1+b\log\left(1+\frac{D}{v_{\mathrm{ss}}\phi^\star}\right)\right]\times\left[\frac{f_0}{\sqrt{1+(v^\star/v_{\mathrm{ss}})^2}}+\alpha\log\left(1+\frac{|v_{\mathrm{ss}}|}{v^\star}\right)\right] \tag*{(33)}

which is NN shaped, as highlighted in the manuscript. It is plotted in Figure S2, where the local minimum at (vmin⁡,τmin⁡)(v_{\min},\tau_{\min}) is marked. The characteristic slip rate vmin⁡=3.4624×10−5 m/sv_{\min}=3.4624\times10^{-5}\ \mathrm{m/s} and characteristic frictional strength τmin⁡=0.3498 MPa\tau_{\min}=0.3498\ \mathrm{MPa} are used to nondimensionalize various quantities hereafter and in the manuscript.

Image not extracted

Figure S2. The steady frictional resistance τss(v)\tau_{\mathrm{ss}}(v) in Eq. (33). The orange square marks the position of the friction curve minimum. The horizontal dashed line denotes the background stress. The green diamond corresponds to the solution τ(vvw,ϕss)=τd\tau(v_{\mathrm{vw}},\phi_{\mathrm{ss}})=\tau_d on the velocity-weakening (vw) branch of the steady-state friction curve.

We employed background stresses in the range τd/τmin⁡∈[1.02,1.078]\tau_d/\tau_{\min}\in[1.02,1.078]. The results plotted in Figure 1 in the manuscript correspond to τd/τmin⁡=1.04\tau_d/\tau_{\min}=1.04 and those in Fig. 2 therein to τd/τmin⁡=1.071\tau_d/\tau_{\min}=1.071. The curve in Figure 4 in the manuscript correspond to τd/τmin⁡=1.070,1.071,1.073,1.075,1.078\tau_d/\tau_{\min}=1.070,1.071,1.073,1.075,1.078, from bottom to top.

ParameterValueUnit
σ\sigma1MPa
ν\nu0.33–
μ\mu9GPa
csc_s2739m/s
DD5×10−75\times10^{-7}m
v⋆v^\star10−810^{-8}m/s
ϕ⋆\phi^\star0.05s
α\alpha0.0075–
bb0.1–
f0f_00.28–

Table S-1. The normal stress σ\sigma, linear elastic bulk parameters, and frictional parameters, cf. Eqs. (31)–(32), used in the computer simulations.

Rupture nucleation procedure

As explained above, Eqs. (30)–(32) for v(x,t)v(x,t) and ϕ(x,t)\phi(x,t) require initial conditions. The latter, which are also used to nucleate rupture, are formulated in two steps. First, for a given τd\tau_d, we set v(x,t=0)=vvwv(x,t=0)=v_{\mathrm{vw}} and ϕ(x,t=0)=ϕss=D/vvw\phi(x,t=0)=\phi_{\mathrm{ss}}=D/v_{\mathrm{vw}} to be space independent, where vvwv_{\mathrm{vw}} is the solution of τ(vvw,ϕss)=τd\tau(v_{\mathrm{vw}},\phi_{\mathrm{ss}})=\tau_d on the velocity-weakening (vw) branch of the steady-state friction curve (illustrated in Figure S2 by a green diamond). Second, at t=0+t=0^+, we introduce a perturbation to ϕ\phi of the form

ϕ(x,t=0+)=[1+ϵsin⁡(2πx/w)]D/vvw,(34)\phi(x,t=0^+)=[1+\epsilon\sin(2\pi x/w)]D/v_{\mathrm{vw}}, \tag*{(34)}

for 0≤x≤w0\le x\le w and ϕ(x,t=0+)=D/vvw\phi(x,t=0^+)=D/v_{\mathrm{vw}} otherwise. Here, ϵ=−0.01\epsilon=-0.01 and ww is the perturbation width, and v(x,t=0+)v(x,t=0^+) is set as if it is slaved to ϕ(x,t=0+)\phi(x,t=0^+), see Figure S3.

Image not extracted

Figure S3. (a) The initial perturbed state field ϕ(x,t=0+)\phi(x,t=0^+), see Eq. (34), (b) the corresponding slip rate v(x,t=0+)v(x,t=0^+) and (c) the corresponding frictional strength τ(x,t=0+)\tau(x,t=0^+), for τd/τmin⁡=1.071\tau_d/\tau_{\min}=1.071. The dashed horizontal lines correspond to D/vvwD/v_{\mathrm{vw}}, vvwv_{\mathrm{vw}} and τd\tau_d, respectively.

The above-described nucleation procedure gives rise to a pair of counter-propagating ruptures. We focus throughout this work on the right-propagating rupture. Yet, due to the employed periodic boundary conditions in the fault direction, the left-propagating rupture eventually interacts with the right-propagating one. Since our entire analysis is performed for an isolated rupture, we increased the propagation distance and time of the right-propagating rupture prior to any interaction with the left-propagating one through the periodic boundary conditions by choosing w=2W/3w=2W/3 and a large enough WW (specifically, set to W=23136W=23136 m). We verified that indeed our analysis of the right-propagating rupture is not affected by any interaction with the left-propagating rupture.

Comparing the theoretical predictions and the numerical observations

In this section, we provide additional details on the comparison between the theoretical predictions and the numerical observations for the right-propagating ruptures, giving rise to the results presented in Figure 1-Figure 3 in the manuscript. In so doing, we also report the values of some of the extracted parameters not reported in the manuscript.

To test Eq. (29) for the breakdown energy Gf(δ)G_f(\delta) (Eq. (4) in the manuscript), we consider the time dependence of the fault slip δ(xi,t)\delta(x_i,t) and stress τ(xi,t)\tau(x_i,t) at a fixed fault location xix_i. Since the fault features a small, yet non-vanishing, background slip rate vvwv_{\mathrm{vw}}, we actually use Gf(δ)=∫δiδ[τ(δ′)−τ(δ)] dδ′G_f(\delta)=\int_{\delta_i}^{\delta}[\tau(\delta')-\tau(\delta)]\,\mathrm{d}\delta', where δi\delta_i corresponds to the slip accumulated prior to the arrival of the rupture to xix_i.

The characteristic slip δc\delta_c is estimated as the slip above which Gf(δ)G_f(\delta) follows a power-law, which is significantly slower than the variation for δ<δc\delta<\delta_c that corresponds to the strong weakening inside the cohesive zone, as illustrated in Figure S4 (and in Figure 1b and 2b in the manuscript). With δc\delta_c at hand, we proceed to determine ξG\xi_G in Eq. (29) (Eq. (4) in the manuscript), as well as the range Δ\Delta of the power-law scaling. We fit Gf(δ)G_f(\delta), in a double-logarithmic representation, over the 4 data points nearest to δc\delta_c (marked by thicker black circles in Figure S4) to extract ξG\xi_G. Then, using ξG\xi_G, δc\delta_c and Gc=Gf(δc)G_c=G_f(\delta_c), we compute the range of δ\delta values — denoted by Δ\Delta — for which the relative difference between the two sides of

Image not extracted

Figure S4. Estimating δc\delta_c and Δ\Delta (the data presented here correspond to Figure 1 in the manuscript). (a) Gf(δ)G_f(\delta) (green circles), as defined in the text, in a double-logarithmic representation. The data are logarithmically spaced for clarity. The 4 points over which the power-law fit parameters were calculated are denoted by thicker black circles. The dashed brown line illustrates the power-law behavior and the corresponding range, Δ\Delta, is marked with a double arrow. (b) F(δ)\mathcal{F}(\delta) of Eq. (35), see text for details. The vertical solid line in both panels corresponds to δc\delta_c, and the vertical dashed line to the right-most edge of the power-law range. The dotted horizontal line in panel (b) corresponds to the chosen threshold of 0.05, cf. Eq. (35).

Eq. (29) satisfies

F(δ)=1−Gc(δ/δc)ξGGf(δ)<0.05,δ∈Δ,(35)\mathcal{F}(\delta)=1-\frac{G_c(\delta/\delta_c)^{\xi_G}}{G_f(\delta)}<0.05,\qquad\delta\in\Delta, \tag*{(35)}

see Figure S4. The values of GcG_c for τd/τmin⁡=1.04\tau_d/\tau_{\min}=1.04 are reported in Figure 1h in the manuscript, and the values of Δ/δc\Delta/\delta_{\mathrm{c}} and ξG\xi_{\mathrm{G}} for all the studied background stresses are reported in Figure 3a,Figure 3b in the manuscript.

Next, we analyze the spatial variation of the slip rate. To that aim, we consider a snapshot v(r,ti)v(r,t_i), e.g., as plotted in Figure 1d in the manuscript, where rr is the distance from the location of the stress peak and tit_i is a fixed time. We then fit the instantaneous snapshot to a power-law v(r,ti)=avrξv(r,t_i)=a_v r^\xi using a double logarithmic representation, as in Figure 1f therein, over the spatial range corresponding to Δ\Delta. The extracted exponent ξ\xi is then compared to the value independently obtained from inverting ξG(ξ)\xi_{\mathrm{G}}(\xi) of Eq. (4) in the manuscript, i.e., ξ=(1−ξG)/(ξG−2)\xi=(1-\xi_{\mathrm{G}})/(\xi_{\mathrm{G}}-2). The comparison is discussed in the manuscript in relation to Figure 1-Figure 3 therein. The stress intensity KII(ξ)(t)K_{\mathrm{II}}^{(\xi)}(t), extracted for various time points tit_i and plotted in Figure 1g in the manuscript, is obtained by equating ava_v to the analytic solution in Eq. (2) in the manuscript.

With KII(ξ)K_{\mathrm{II}}^{(\xi)} and GcG_{\mathrm{c}} at hand, we can test the rupture energy balance prediction in Eq. (5) in the manuscript, with AII(cr,ξ)A_{\mathrm{II}}(c_{\mathrm{r}},\xi) of Eq. (27). To this aim, the size ℓc\ell_{\mathrm{c}} of the cohesion zone is computed by integrating the instantaneous rupture speed over the time it takes the fault slip to reach δc\delta_{\mathrm{c}}, i.e.,

ℓc=∫titccr(t′) dt′,(36)\ell_{\mathrm{c}}=\int_{t_i}^{t_{\mathrm{c}}}c_{\mathrm{r}(t')} \,\mathrm{d}t', \tag*{(36)}

where δ(ti)=δi\delta(t_i)=\delta_i and δ(tc)=δc\delta(t_{\mathrm{c}})=\delta_{\mathrm{c}}. The resulting ℓc\ell_{\mathrm{c}} is plotted in Figure S5 for all τd\tau_{\mathrm{d}} simulations and times. The excellent agreement of the numerical data with the theoretical prediction in Eq. (5) in the manuscript is demonstrated in Figure 1h in the manuscript. Finally, for completeness, we plot in Figure S6 the effective viscosity ηeff\eta_{\mathrm{eff}} (the linear slope of τ(v)\tau(v) in the slip rate range corresponding to Δ\Delta, as shown in Figure 1e and 2a(inset) in the manuscript) for all simulations. These values were used in Figure 3c in the manuscript to obtain ξ(η)\xi(\eta) using Eq. (7) therein.

Image not extracted

Figure S5. The cohesion zone size ℓc\ell_{\mathrm{c}} vs. crc_{\mathrm{r}} for various τd\tau_{\mathrm{d}} values (see text for details). The symbols here are the same as in Figure 3 in the manuscript and correspond to τd/τmin⁡=1.04\tau_{\mathrm{d}}/\tau_{\min}=1.04 (leftmost, circles, cf. Figure 1 in the manuscript), 1.045, 1.050, 1.055, 1.060, 1.065, 1.070 and 1.071 (rightmost, diamonds, cf. Fig. 2).

Image not extracted

Figure S6. The effective viscosity ηeff\eta_{\mathrm{eff}} vs. crc_{\mathrm{r}} for various τd\tau_{\mathrm{d}} values as in Figure S5 (the same symbols are used).

Numerical convergence

In this section, we complement the previous two sections by demonstrating the robustness of our numerical results, i.e., the convergence of the simulations with respect to the spatial and temporal discretizations, Δx\Delta x and Δt\Delta t, respectively. The latter two are related via the Courant-Friedrichs-Lewy (CFL) constant β=csΔt/Δx>1\beta=c_{\mathrm{s}}\Delta t/\Delta x>1. In our simulations, β=0.2\beta=0.2, and Δt\Delta t follows from the spatial resolution. Given the very large fault size W=23136W=23136 m, already stated above, fully resolving the fields inside the cohesion zone is challenging; yet, we confidently determine its edge, which allows to extract ℓc\ell_{\mathrm{c}}, as discussed above. Since we test the broad range of theoretical predictions for δ>δc\delta>\delta_{\mathrm{c}} (correspondingly, r>ℓcr>\ell_{\mathrm{c}}), i.e., out of the cohesive zone, this is enough for our purposes.

To demonstrate this, we compare results for rupture propagation in the low sub-Rayleigh regime (cr≃0.27cRc_{\mathrm{r}}\simeq0.27c_{\mathrm{R}}, cf. Figure 1 in the manuscript) and close to cRc_{\mathrm{R}} (cr≃0.93cRc_{\mathrm{r}}\simeq0.93c_{\mathrm{R}}, cf. Fig. 2 in the manuscript) for two numerical resolutions Δx=0.093\Delta x=0.093 m (and correspondingly, Δt=6.79×10−6\Delta t=6.79\times10^{-6} sec) and Δx=0.0465\Delta x=0.0465 m (and correspondingly, Δt=3.395×10−6\Delta t=3.395\times10^{-6} sec). First, we demonstrate in Figure S7 that our simulations are fully converged in the low sub-Rayleigh regime. Such a quality of convergence persists up to rupture propagation speeds of about 0.9cs0.9c_{\mathrm{s}} or so. Second, for higher propagation speeds, the lower spatial resolution is not sufficient to resolve the edge of the cohesion zone, as demonstrated in Figure S8, where we present results for cr≃0.93cRc_{\mathrm{r}}\simeq0.93c_{\mathrm{R}} and cr≃0.99cRc_{\mathrm{r}}\simeq0.99c_{\mathrm{R}} using the lower resolution. Finally, this issue is rectified by employing the higher resolution, as demonstrated in Figure S9.

Image not extracted

Figure S7. Convergence test for cr/cR≃0.27c_r/c_R \simeq0.27. (a) The slip rate snapshots, (b) the shear stress snapshots, (c) the snapshots of τ(v)\tau(v), and (d) Gf(δ)G_f(\delta) for the two resolutions discussed in the text. Yellow squares denote the lower resolution results (Δx=0.093\Delta x = 0.093 m) and green circles mark the higher resolution results (Δx=0.0465\Delta x = 0.0465 m). The vertical solid line in all panels denotes the values of δc\delta_c obtained from higher resolution simulations. In panel (d), the dotted line marks the value of δc\delta_c in the lower resolution simulations. In panels (c) and (d), the dashed and dot-dashed lines denote the end of the power-law scaling range for higher and lower resolution, respectively.

Image not extracted

Figure S8. The low numerical resolution (cf. Figure S7 and text for definition) is not sufficient to resolve the edge of the cohesive zone very close to cRc_R. We present the breakdown energy Gf(δ)G_f(\delta) for cr≃0.93cRc_r \simeq0.93c_R (green diamonds) and cr≃0.99cRc_r \simeq0.99c_R (black squares). For the former, δc\delta_c is resolved and is marked by the vertical solid line. For the latter, δc\delta_c is not resolved (at best, the results offer an upper bound on δc\delta_c). This issue is rectified by using the higher resolution, as demonstrated in Figure S9.

Image not extracted

Figure S9. The high numerical resolution (cf. Figure S7 and text for definition) is sufficient to resolve the edge of the cohesive zone very close to cRc_R and above it. (a) The slip rate snapshots for cr≃0.93cRc_r \simeq0.93c_R (green diamonds), cr≃0.99cRc_r \simeq0.99c_R (black squares), and cr≃1.014cR>cRc_r \simeq1.014c_R > c_R (brown circles), all symbols as in Fig. 4(inset) in the manuscript. (b) the corresponding shear stress snapshots. (c) the corresponding snapshots of τ(v)\tau(v). (d) The corresponding Gf(δ)G_f(\delta), demonstrating that δc\delta_c (and hence ℓc\ell_c) is resolved at all 3 propagation speeds, as is further highlighted in panel (e). δc\delta_c for cr≃0.93cRc_r \simeq0.93c_R is marked in panel (d) by the vertical solid line, while for both cr≃0.99cRc_r \simeq0.99c_R and cr≃1.014cRc_r \simeq1.014c_R it is marked by the vertical dotted line (its value is the same for the two speeds, within the current uncertainty). The upper limit of the power-law scaling range of Gf(δ)G_f(\delta) for cr≃0.93cRc_r \simeq0.93c_R is marked in panel (d) by the vertical dashed line. The corresponding lines are superposed on panel (c). Finally, the corresponding values of ℓc\ell_c are marked in panels (a)-(b). (e) A zoom-in on the two upper datasets in panel (d), highlighting that the first data point in each slightly deviates from the subsequent data points, which allows to estimate δc\delta_c.

Additional refinement of the spatial resolution is computationally prohibitive for the problem at hand, in light of the physical questions that are addressed. Yet, the above analysis demonstrated the robustness and numerical convergence of our main computational findings, which agree with the theoretical predictions, using the higher resolution scheme. Consequently, all the results presented in the manuscript and in the previous sections in this document correspond to the higher resolution of the two discussed.

References

References

  1. [3]A. Elbanna, M. Abdelmeguid, D. Asimaki, N. Tainpakaidip, G. Lavrentadis, A. Rosakis, and Y. Benzion, Supershear Earthquakes: Their Occurrence and Importance for Seismic Hazard, Early Warning, and Design Standards, Seismological Research Letters 96, 3319 (2025).
  2. [4]A. J. Rosakis, Intersonic shear cracks and fault ruptures, Adv. Phys. 51, 1189 (2002).DOI
  3. [5]A. Bizzarri, The mechanics of supershear earthquake ruptures, Mechanics of Earthquake Faulting , 1 (2019).
  4. [6]P. Bernard and D. Baumont, Shear mach wave characterization for kinematic fault rupture models with constant supershear rupture velocity, Geophysical Journal International 162, 431 (2005).
  5. [7]E. M. Dunham and R. J. Archuleta, Near-source ground motion from steady state dynamic rupture pulses, Geophysical Research Letters 32 (2005).
  6. [8]H. S. Bhat, R. Dmowska, G. C. King, Y. Klinger, and J. R. Rice, Off-fault damage patterns due to supershear ruptures with application to the 2001 Mw 8.1 Kokoxili (Kunlun) Tibet earthquake, Journal of Geophysical Research: Solid Earth 112 (2007).
  7. [9]L. B. Freund, The mechanics of dynamic shear crack propagation, J. Geophys. Res. 84, 2199 (1979).
  8. [10]L. B. Freund, Dynamic Fracture Mechanics (Cambridge university press, Cambridge, 1998).
  9. [11]K. B. Broberg, Cracks and fracture (Academic Press, 1999).
  10. [12]R. Burridge, Admissible speeds for plane-strain self-similar shear cracks with friction but lacking cohesion, Geophysical Journal International 35, 439 (1973).
  11. [13]D. Andrews, Rupture velocity of plane strain shear cracks, Journal of Geophysical Research 81, 5679 (1976).
  12. [14]R. Burridge, G. Conn, and L. B. Freund, The stability of a rapid mode II shear crack with finite cohesive traction, Journal of Geophysical Research: Solid Earth 84, 2210 (1979).
  13. [15]A.-A. Gabriel, J.-P. Ampuero, L. A. Dalguer, and P. M. Mai, The transition of dynamic rupture styles in elastic media under velocity-weakening friction, J. Geophys. Res. Solid Earth 117, B09311 (2012).
  14. [16]I. Svetlitzky, D. Pino Muñoz, M. Radiguet, D. S. Kammer, J.-F. Molinari, and J. Fineberg, Properties of the shear stress peak radiated ahead of rapidly accelerating rupture fronts that mediate frictional slip, Proc. Natl. Acad. Sci. 113, 542 (2016).
  15. [17]E. M. Dunham, Conditions governing the occurrence of supershear ruptures under slip-weakening friction, Journal of Geophysical Research: Solid Earth 112, B07302 (2007).
  16. [18]E. Fukuyama and K. B. Olsen, A condition for supershear rupture propagation in a heterogeneous stress field, Pure and Applied Geophysics 159, 2047 (2002).
  17. [19]E. M. Dunham, P. Favreau, and J. Carlson, A supershear transition mechanism for cracks, Science 299, 1557 (2003).
  18. [20]Y. Liu and N. Lapusta, Transition of mode II cracks from sub-Rayleigh to intersonic speeds in the presence of favorable heterogeneity, J. Mech. Phys. Solids 56, 25 (2008).
  19. [21]H. Weng, J. Huang, and H. Yang, Barrier-induced supershear ruptures on a slip-weakening fault, Geophysical Research Letters 42, 4824 (2015).
  20. [22]F. Barras, P. H. Geubelle, and J.-F. Molinari, Interplay between process zone and material heterogeneities for dynamic cracks, Physical Review Letters 119, 144101 (2017).
  21. [23]K. J. Ryan and D. D. Oglesby, Dynamically modeling fault step overs using various friction laws, Journal of Geophysical Research: Solid Earth 119, 5814 (2014).
  22. [24]F. Hu, J. Xu, Z. Zhang, and X. Chen, Supershear transition mechanism induced by step over geometry, Journal of Geophysical Research: Solid Earth 121, 8738 (2016).
  23. [25]H. Kehoe and E. Kiser, Evidence of a supershear transition across a fault stepover, Geophysical Research Letters 47, e2020GL087400 (2020).
  24. [26]L. Bruhat, Z. Fang, and E. M. Dunham, Rupture complexity and the supershear transition on rough faults, J. Geophys. Res. Solid Earth 121, 210 (2016).
  25. [27]Y. Huang, J.-P. Ampuero, and D. V. Helmberger, The potential for supershear earthquakes in damaged fault zones – theory and observations, Earth Planet. Sci. Lett. 433, 109 (2016).
  26. [28]Y. Kaneko and N. Lapusta, Supershear transition due to a free surface in 3-D simulations of spontaneous dynamic rupture on vertical strike-slip faults, Tectonophysics 493, 272 (2010).
  27. [29]H. Weng and J.-P. Ampuero, Continuum of earthquake rupture speeds enabled by oblique slip, Nature Geoscience 13, 817 (2020).
  28. [30]A. J. Rosakis, O. Samudrala, and D. Coker, Cracks faster than the shear wave speed, Science 284, 1337 (1999).
  29. [31]K. Xia, A. J. Rosakis, and H. Kanamori, Laboratory earthquakes: The sub-Rayleigh-to-supershear rupture transition, Science 303, 1859 (2004).
  30. [32]O. Ben-David, G. Cohen, and J. Fineberg, The Dynamics of the Onset of Frictional Slip, Science 330, 211 (2010).
  31. [33]D. S. Kammer, I. Svetlizky, G. Cohen, and J. Fineberg, The equation of motion for supershear frictional rupture fronts, Sci. Adv. 4, eaet5622 (2018).
  32. [34]F. X. Passelègue, A. Schubnel, S. Nielsen, H. S. Bhat, and R. Madariaga, From sub-Rayleigh to supershear ruptures during stick-slip experiments on crustal rocks, Science 340, 1208 (2013).
  33. [35]S. Xu, E. Fukuyama, F. Yamashita, K. Mizoguchi, S. Takizawa, and H. Kawakata, Strain rate effect on fault slip and rupture evolution: Insight from meter-scale rock friction experiments, Tectonophysics 733, 209 (2018).
  34. [36]J. H. Dieterich, Modeling of rock friction: 1. Experimental results and constitutive equations, J. Geophys. Res. Solid Earth 84, 2161 (1979).
  35. [37]A. L. Ruina, Slip instability and state variable friction laws, J. Geophys. Res. 88, 10359 (1983).
  36. [38]T. Shimamoto, Transition Between Frictional Slip and Ductile Flow for Halite Shear Zones at Room Temperature, Science 231, 711 (1986).
  37. [39]T. E. Tullis and J. D. Weeks, Constitutive behavior and stability of a natural sliding of granite, Pure and Applied Geophysics 124, 383 (1986).
  38. [40]M. Blanpied, D. Lockner, and J. Byerlee, Fault stability inferred from granite sliding experiments at hydrothermal conditions, Geophysical Research Letters 18, 609 (1991).
  39. [41]B. N. J. Persson, Sliding Friction: Physical Principles and Applications (Springer Science & Business Media, 1998).
  40. [42]C. Marone, Laboratory-derived friction laws and their application to seismic faulting, Annu. Rev. Earth Planet. Sci. 26, 643 (1998).
  41. [43]T. Baumberger and P. Berthoud, Physical analysis of the state- and rate-dependent friction law. II. Dynamic friction, Phys. Rev. B 60, 3928 (1999).
  42. [44]M. Nakatani, Conceptual and physical clarification of rate and state friction: Frictional sliding as a thermally activated rheology, J. Geophys. Res. Solid Earth 106, 13347 (2001).
  43. [45]T. Baumberger and C. Caroli, Solid friction from stick-slip down to pinning and aging, Adv. Phys. 55, 279 (2006).
  44. [46]Z. Reches and D. A. Lockner, Fault weakening and earthquake instability by powder lubrication, Nature 467, 452 (2010).
  45. [47]Y. Bar-Sinai, R. Spatschek, E. A. Brener, and E. Bouchbinder, On the velocity-strengthening behavior of dry friction, J. Geophys. Res. Solid Earth 119, 1738 (2014).
  46. [48]H. Gou, W. Hu, Q. Xu, J. Chen, M. McSaveney, E. C. Breard, R. Huang, Y. Wang, X. Jia, and L. Zhou, Variation in granular frictional resistance across nine orders of magnitude in shear velocity, Journal of Geophysical Research: Solid Earth 129, e2023JB028241 (2024).
  47. [49]R. C. Viesca and D. I. Garagash, Ubiquitous weakening of faults due to thermal pressurization, Nat. Geosci. 8, 875 (2015).
  48. [50]N. Brantut and R. C. Viesca, The fracture energy of ruptures driven by flash heating, Geophysical Research Letters 44, 6718 (2017).
  49. [51]Y. Bar-Sinai, M. Adam, R. Spatschek, E. A. Brener, and E. Bouchbinder, Spatiotemporal dynamics of frictional systems: The interplay of interfacial friction and bulk elasticity, Lubricants 7, 91 (2019).
  50. [52]L. Landau and E. Lifshitz, Theory of Elasticity, Third Edition: Volume 7 (Course of Theoretical Physics) (Butterworth-Heinemann, 1986).
  51. [53]See Sect. 4.3 in [10], Eq. (4.3.2) therein in particular.
  52. [54]In Eq. (4.3.2) of [10], displacement potentials are used such that the order of singularity therein is p₀ = ξ + 2, with 1 < p₀ < 2.
  53. [55]A. C. Palmer and J. R. Rice, The growth of slip surfaces in the progressive failure of over-consolidated clay, Proc. R. Soc. A Math. Phys. Eng. Sci. 332, 527 (1973).
  54. [56]N. Berman, G. Cohen, and J. Fineberg, Dynamics and properties of the cohesive zone in rapid fracture and friction, Physical Review Letters 125, 125503 (2020).
  55. [57]H. Kanamori and T. H. Heaton, Microscopic and macroscopic physics of earthquakes, Geocomplexity and the Physics of Earthquakes 120, 147 (2000).
  56. [58]R. E. Abercrombie and J. R. Rice, Can observations of earthquake scaling constrain slip weakening?, Geophys. J. Int. 162, 406 (2005).
  57. [59]E. Tinti, P. Spudich, and M. Cocco, Earthquake fracture energy inferred from kinematic rupture models on extended faults, J. Geophys. Res. 110, B12308 (2005).
  58. [60]S. B. Nielsen, E. Spagnuolo, S. A. F. Smith, M. Violay, G. Di Toro, and A. Bistacchi, Scaling in natural and laboratory earthquakes, Geophys. Res. Lett. 43, 1504 (2016).
  59. [61]E. A. Brener and E. Bouchbinder, Unconventional singularities and energy balance in frictional rupture, Nat. Commun. 12, 1 (2021).
  60. [62]E. A. Brener and E. Bouchbinder, Theory of unconventional singularities of frictional shear cracks, J. Mech. Phys. Solids 153, 104466 (2021).
  61. [63]M. Cocco, S. Aretusini, C. Cornelio, S. B. Nielsen, E. Spagnuolo, E. Tinti, and G. Di Toro, Fracture energy and breakdown work during earthquakes, Annual Review of Earth and Planetary Sciences 51, 217 (2023).
  62. [64]Y. Ida, Slow-moving deformation pulses along tectonic faults, Physics of the Earth and Planetary Interiors 9, 328 (1974).
  63. [65]E. A. Brener and V. I. Marchenko, Frictional shear cracks, J. Exp. Theor. Phys. Lett. 76, 211 (2002).
  64. [66]E. A. Brener, S. V. Malinin, and V. I. Marchenko, Fracture and friction: Stick-slip motion, Eur. Phys. J. E 17, 101 (2005).
  65. [67]J. Desroches, E. Detournay, B. Lenoech, P. Papanastasiou, J. R. A. Pearson, M. Thiercelin, and A. Cheng, The crack tip region in hydraulic fracturing, Proceedings of the Royal Society of London. Series A: Mathematical and Physical Sciences 447, 39 (1994).
  66. [68]D. Garagash and E. Detournay, The tip region of a fluid-driven fracture in an elastic medium, J. Appl. Mech. 67, 183 (2000).
  67. [69]E. Detournay, Mechanics of hydraulic fractures, Annual review of fluid mechanics 48, 311 (2016).
  68. [70]F. Barras, M. Adam, T. Roch, E. A. Brener, E. Bouchbinder, and J.-F. Molinari, The emergence of crack-like behavior of frictional rupture: Edge singularity and energy balance, Earth and Planetary Science Letters 531, 115978 (2020).
  69. [71]F. Paglialunga, F. X. Passelègue, N. Brantut, F. Barras, M. Lebihain, and M. Violay, On the scale dependence in the dynamics of frictional rupture: Constant fracture energy versus size-dependent breakdown work, Earth and Planetary Science Letters 584, 117442 (2022).
  70. [72]F. Paglialunga, F. Passelègue, M. Lebihain, and M. Violay, Frictional weakening leads to unconventional singularities during dynamic rupture propagation, Earth and Planetary Science Letters 626, 118550 (2024).
  71. [73]B. Fryer, M. Lebihain, C. Noël, F. Paglialunga, and F. Passelègue, The effect of stress barriers on unconventional-singularity-driven frictional rupture, Journal of the Mechanics and Physics of Solids 193, 105876 (2024).
  72. [74]T. Roch, F. Barras, P. H. Geubelle, and J.-F. Molinari, cCracklet: a spectral boundary integral method library for interfacial rupture simulation, Journal of Open Source Software 7, 3724 (2022).
  73. [75]G. Festa and J.-P. Vilotte, Influence of the rupture initiation on the intersonic transition: Crack-like versus pulse-like modes, Geophysical Research Letters 33, L15320 (2006).
  74. [76]X. Lu, N. Lapusta, and A. J. Rosakis, Analysis of supershear transition regimes in rupture experiments: the effect of nucleation conditions and friction parameters, Geophysical Journal International 177, 717 (2009).
  75. [77]C. Liu, A. Bizzarri, and S. Das, Progression of spontaneous in-plane shear faults from sub-Rayleigh to compressional wave rupture speeds, Journal of Geophysical Research: Solid Earth 119, 8331 (2014).
  76. [78]A. Bizzarri and C. Liu, Near-field radiated wave field may help to understand the style of the supershear transition of dynamic ruptures, Physics of the Earth and Planetary Interiors 261, 133 (2016).
  77. [79]C. Liang, J.-P. Ampuero, and D. Pino Muñoz, The paucity of supershear earthquakes on large faults governed by rate and state friction, Geophysical Research Letters 49, e2022GL099749 (2022).
  78. [80]A. Chounet, M. Vallée, M. Causse, and F. Courboulex, Global catalog of earthquake rupture velocities shows an- ticorrelation between stress drop and rupture velocity, Tectonophysics 733, 148 (2018).
  79. [81]H. Bao, J.-P. Ampuero, L. Meng, E. J. Fielding, C. Liang, C. W. Milliner, T. Feng, and H. Huang, Early and persistent supershear rupture of the 2018 magnitude 7.5 Palu earthquake, Nature Geoscience 12, 200 (2019).
  80. [82]G. R. Irwin, Analysis of stresses and strains near the end of a crack traversing a plate, J. Appl. Mech. 24, 361 (1957).
  81. [83]S. Das, A numerical method for determination of source time functions for general three-dimensional rupture propagation, Geophysical Journal International 62, 591 (1980).
  82. [84]M. S. Breitenfeld and P. H. Geubelle, Numerical analysis of dynamic debonding under 2D in-plane and 3D loading, Int. J. Fract. 93, 13 (1998).
  83. [85]E. A. Brener, M. Adam, F. Barras, J.-F. Molinari, and E. Bouchbinder, Unstable Slip Pulses and Earthquake Nucleation as a Nonequilibrium First-Order Phase Transition, Phys. Rev. Lett. 121, 234302 (2018).

Paper details

vlnc:W_2jcqc2zd
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents