Rupture speed and slip velocity: What can we learn from simulated earthquakes? Andrea Bizzarri ⁎ Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Bologna, Italy abstract article info Article history: Received 20 June 2011 Received in revised form 3 November 2011 Accepted 19 November 2011 Available online 27 December 2011 Editor:L. Stixrude Keywords: seismic source rupture velocity computational sesimology rheology of fault zones In this paper we consider a wide catalog of synthetic earthquakes, numerically modeled as spontaneous, fully dynamic, 3-D ruptures on extended faults, governed by different friction laws, including slip-dependent and rate- and state-dependent equations. We analyze the spatial correlations between the peak of fault slip velocity (v peak ) and the rupture speed (v r ) at which the earthquake spreads over the fault. We found that v peak positively correlates with v r and that the increase of v peak is roughly quadratic. We found that near the transition between sub- and supershear regimes v peak significantly diminishes and then starts to increase again with the square of v r . This holds for all the governing models we consider and for both homogeneous and heterogeneous configura- tions. Moreover, we found that, on average, v peak increases with the magnitude of the event (v peak ~ M 0 0.18 ). Our results can be incorporated as constraints in the inverse modeling of faults. © 2011 Elsevier B.V. All rights reserved. 1. Introduction Understanding the physical and chemical dissipative processes taking place during an earthquake is of pivotal importance in the me- chanics of faulting. Fully dynamic models of spontaneously spreading ruptures give us the extraordinary chance to investigate the features of the constitutive law assumed to govern the fault surface, under conditions that are very often far of being properly reproduced in labo- ratory experiments. One of the goals of modern-days seismology is to design robust and computationally efficient numerical codes able to generate a cata- log of synthetic events and to simulate the synthetic motions recorded on the ground (i.e., on the free surface). The physics-based earthquake (forward) source models appear to be crucial for realistic ground mo- tion simulation and seismic hazard analysis; when seismological data are rare (or even non-existent), numerical experiments can be used in order to predict ground motions caused by future earthquakes. At the same time, it is not obvious what is the most appropriated governing model to describe the breakdown mechanism occurring during slip fail- ures (see Bizzarri, 2011b and references cited therein for a discussion). Investigations of possible spatial correlations between the various dynamic variables, such as fracture energy density, stress drop, total developed slip, peak fault slip velocity (v peak ) and rupture speed (v r ) are important because they could be inserted as constraints in kine- matic modeling of faults, on which current practice in seismic engi- neering relies. By performing laboratory experiments of a mode II crack expand- ing in a granite sample, the following direct dependence between v peak and v r has been proposed (Ohnaka et al., 1987): v peak ≅v r Δτ b G ð1Þ where Δτ b is the breakdown stress drop (expressing the difference between the upper and residual stress levels) and G is the rigidity of the elastic medium. Notably, in laboratory only fracture on intact rocks experiments give the rupture speed, contrarily to friction exper- iments, both rotary shear and sandwich-like, where two pre-existing surfaces slide against each other (and thus without the existence of a crack tip). In their pseudodynamic earthquake source modeling Guatteri et al. (2004) try to understand the spatial interdependency of the earth- quake source parameters, such as v r and the total slip (u tot ). Schmedes et al. (2010a) analyze a series of dynamic models obeying the linear slip-weakening friction to find correlations between various source parameters. On the other hand, Song et al. (2009) explore the spatial coherence between u tot and v r , and between u tot and v peak by analyzing kinematic rupture models of two large strike–slip events (this analysis has been then extended to dynamic models by Song and Sommerville, 2010). Bizzarri (2010c) thoroughly discusses the relations between the fracture energy and different physical observables, such as v r , u tot and the dynamic stress drop, by analyzing spontaneous dynamic earth- quake models obeying different governing models. Given the above-mentioned results, with this study we aim to un- derstand whether, and how, v peak and v r correlate. Both of these two source parameters have a fundamental role in ground motion Earth and Planetary Science Letters 317-318 (2012) 196–203 ⁎ Via Donato Creti, 12, 40128 Bologna, Italy. Tel.: +39 051 4151432; fax: +39 051 4151499. E-mail address: bizzarri@bo.ingv.it. 0012-821X/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.epsl.2011.11.023 Contents lists available at SciVerse ScienceDirect Earth and Planetary Science Letters journal homepage: www.elsevier.com/locate/epsl