My project concerns the study of the seismic source by dynamic simulations to investigate how complex geometry for the fault and realistic structural models can drive a seismic rupture in subduction zones.
In this work the Spectral Element Method (SEM) for seismological applications (Komatitsch and Vilotte, 1998) is used: it combines the geometrical flexibility of the Finite Element Method with the rapid convergence of Spectral methods (Maday and Patera, 1989) allowing to tackle complex domains as required by realistic applications. SEM also allows for a natural treatment of seismological boundary conditions (free surface) and it enables a dynamic description of the rupture propagating along a fault interface.
The subduction zones are wide areas located at convergent boundaries where a tectonic plate sinks below another one. These areas are responsible for the biggest seismic events occurred on the Earth (e.g. Tohoku 2011). From a dynamic point of view they are characterized by a peculiar coupling between normal traction perturbations and shear stress mainly due to: variable slope of the fault, interaction with standing waves generated at free surface and the so-called “bimaterial effect”.
The influence of bimaterial effect (i.e. propagation of the rupture along interfaces between dissimilar materials) is a challenging aspect for the study of ruptures owing both to the analytical ill-posedness when a Coulomb friction law is used and to the effects of directivity both for source kinematics and emitted radiation.
As shown by analytical results (Ranjith and Rice, 2001) the use of a relaxation mechanism between shear stress and normal traction in the friction law makes the problem well-posed and this delay, expected from laboratory experiments (Prakash and Clifton, 1998), suggests the existence of a length scale (missing in Coulomb friction law) along the interface. The regularizations proposed so far, based on dynamic and/or constant time scales provided numerical well-posed results without caring about the physical implications of the regularization itself.
The aim of the project is the characterization of the physical parameters involved in the regularization to obtain reliable solutions both for acceleration (for which analytical solutions are not available) and stationary phase of the rupture (for which the expected asymptotic speed may be computed) and the characterization of the missing length scale. In particular the parametric studies conducted so far show a strong dependence of this length on the process zone when a linear slip weakening is used.
The case study of Tohoku is interesting because of the asymmetric kinematic description obtained from several inversions of seismic data, which revealed larger slip close to the trench related to low frequency radiation and several high frequency smaller-scale sources in deepest part of the subduction zone inferred from high-frequency inversion. Moreover the same inversions have revealed that the rupture was confined in a small stripe of the fault plane (along-dip) during the initial phase, allowing to perform a large number of 2D dynamic simulations.
The most reliable dynamics models showed a bilateral rupture faster upward as an effect of bimaterial propagation; larger slip, associated with low frequency signals close to the trench as an effect of the coupling between low normal stress and interaction with standing waves; high frequency signals from the deep part of the plate inside the mantle wedge as an effect of the acceleration of rupture due to: an inversion of the impedance contrast, to geometrical discontinuities inside the mantle itself and inhomogeneous initial stress conditions mimicking the areas where recent big events were localized.