The project group focuses on a major problem in earthquake science, namely, to understand the interaction of seismic slip-rupture with geometrical and structural complexities of fault zones. Such interactions include transitions of the failure path among fault strands at bends, branches and stepovers, rupture arrest, and induced inelastic deformations in fault border zones, which are generally damaged (highly cracked and/or granulated) and fluid-saturated. Prior work of the group, on which the current studies build, provided new understanding of how rupture paths are chosen at branch-like geometric complexities, and of how inelastic response of the fault bordering zone affects rupture propagation and shear localizations.
The new areas for theory and modeling in current work are as follows: (1) Understanding how interactions of deformations with ground fluids and frictional elastic-plastic responses in the damage zone couple to the dynamics of rupture propagation. That includes explaining the effects of different types of across-fault material dissimilarity (in elastic properties, strength and extent of damage, and near-fault permeability to fluids); such dissimilarities are common for mature, highly slipped faults. (2) Assessing if, and to what extent, current understanding of how rupture paths are chosen at branch intersections, and of whether rupture passes through or arrests at step-overs, is affected by the presence of extensively damaged material, capable of elastic-plastic response, near such fault junctions. (3) Determining how residual stress states imprinted in fault-border material by the previous rupture affects response in the next event, and how that depends on rupture directivity in the past and pending events; (4) Devising procedures to rigorously analyze strain localizations that arise in modeling inelastic response of damaged/granulated fault border zones, by imposing localization-limiting procedures that eliminate grid dependence, thus ultimately evolving a methodology that can predict spontaneous development of localized fault-rupture paths through damaged material.
Correlation of theory and modeling with field examples and lab experiments is a hallmark of the group's work, and new thrusts in that direction are as follows: (I) Adopting methodology like in (4) above to understanding when a damaged pull-apart stepover, like in the 1992 Landers earthquake between the Johnson and Homestead Valley Faults, and between the Homestead Valley and the Emerson Fault, is breeched by a through-going rupture, and similarly for the 1920 M8 Haiyuan, China event, which ruptured through a sequence of pull-aparts. (II) Understanding mega-branches of great thrust ruptures onto splay faults through the sediment cover of accretionary subduction zones, like documented or suspected at Alaska, Cascadia, Nankai and Sumatra, as well as when and by what processes branching onto landward- versus seaward-vergent splays can occur, and what that means for tsunami generation. (III) Testing the evolving theoretical understanding of rupture branching and interactions with damaged border zones against results of lab experiments (conducted by colleagues elsewhere) which are devised to address the same issues.
The understanding of when and how earthquake ruptures stop, which often involves geometric complexities of the type we address, is central to understanding seismic risk. New ways of using relic fault geometries to constrain directivity and other features of past events is also a potentially valuable outcome.
Our project sought to understand principles controlling earthquake rupture propagation and arrest in typical geometrically complex fault systems. We devised computational modeling to simulate dynamic rupture with inclusion of non-planarity of fault surfaces, step-overs from one fault trace to an adjacent one, and branches in fault networks. In those, the path of the seismic rupture was dynamically self-chosen among available fault segments according to the intensity and distribution of stress changes radiated from the already ruptured part of the fault system. Those provide background for assessing earthquake risk in complex fault systems, where the simplest assumption, that each continuous fault segment defines the greatest possible length of an earthquake rupture involving that segment, may not be valid. Instead, the rupture path may follow a branch off the main fault, ultimately leading to another segment, or may stop at a segment end but radiate enough stress from that site to initiate fresh rupture on a nearby but disconnected segment. Examples of both occurred in the well-documented 1992 Landers earthquake along a complex fault network in the Mohave desert. While conventional analyses had often considered the rock outside the network of fault segments to respond to local stress changes as an elastic solid, a more realistic kind of modeling that we helped to pioneer treats that region as being "elastic-plastic", which means that it could deform permanently under large enough stress changes. That elastic-plastic response is due to material granulation and networks of small fractures in the "damage zones" that inevitably border mature faults that have slipped, over geologic time, in many earthquakes. Such damaged material ceases to behave elastically at much smaller stresses than for pristine rock. Another realistic feature of maturely slipped faults, on which our work focused, is that they often separate terrains with different seismic properties. Such is important because, in certain circumstances, that may favor a rupture which typically propagates in one direction versus the other along the fault. That causes a "directivity" to the radiated ground motions, with stronger radiation, and hence more likely earthquake damage, in the direction of propagation. We showed that elastic-plastic off-fault response could sometimes reverse such predicted directivity based on elastic analyses, as could also the "poroelastic" effects of groundwater presence when one side of the fault is more porous and permeable than the other. Effects of supershear rupture speed (i.e., faster than the propagation of elastic shear strain disturbances) and rupture path branching on radiated ground motion were characterized for surface-breaking dip-slip ruptures, with particular reference to a branched configuration of importance for the proposed radioactive waste repository site at Yucca Mountain, Nevada. The inclusion of elastic-plastic response was shown to significantly diminish the large ground accelerations which would otherwise result, from Mach fronts and from interactions at the branching junction -- although repository ground accelerations two or more times the gravitational acceleration could not be precluded. Our predictions on fault branching and jumping compare favorably to various field examples that we examined, from California, Alaska and Japan, although we often did not have unambiguous enough values of input parameters (e.g., direction of the maximum compressive pre-stress) to be sure. Thus we also used our computational modeling predictions of rupture along kinked or branched fault systems to compare to the observations in laboratory experiments, done by the group of Prof. Ares Rosakis at Caltech. Those studied propagation of shear cracks along non-planar, kinked and branched fault paths in pre-cut and weakly rebonded homalite (photoelastic) plates. We found that dynamic 2D finite element analyses reproduced the qualitative rupture behavior past the bendor branch junctions in most cases and generally reproduced principal features of high-speed photographs of dynamic stress contour patterns. We also presented evidence of coseismic subduction zone splay faulting (a large-scale fault-branching episode) as part of the great 2004 Indian Ocean earthquake. That was recognized in the ocean-bound tsunami wave-form, as detected by satellite altimetry measurements, as a lead wave with two peaks of similar amplitude and wavelength. We showed by analysis of ocean gravity wave dynamics to such was consistent with coseismic activation of such a branch along the updip portion of the subduction interface. Back projection of the observed waveform suggested splay rupture on a fault located ~110 km from the trench. Stress changes and deformation caused by such a hypothesized rupture were shown to be compatible with aftershock locations, geodetic measurements on the closest large island, Simeulue, and direct observation of surface rupture on a small island NNW of Simeulue. Concerning human resources, partial support was provided for students Elizabeth L. Templeton (Ph.D. 2009) and Nora DeDontney (Ph.D. 2011), both now employed in the energy industry, Robert C. Viesca (Ph.D. 2011), now a university faculty member, and John D. Platt, a current Ph.D. candidate.
