Summary of the results

We present some properties of a new model of rupture designed to reproduce structural patterns observed in the formation and evolution of a population of strike-slip faults. This model is a multiscale cellular automaton with two states. An active state represents actively slipping fault segments. A stable state represents `intact' zones in which the fracturing process is confined to a smaller scale. At the elementary scale, the transition rates from one state to another are determined with respect to the magnitude of the local strain rate and a time dependent stochastic process. At increasingly larger scales, healing and faulting are described according to geometric rules of interaction between active fault segments based on fracture mechanics. A redistribution of the strain rates in the neighbourhood of active faults at all scales ensures long range interactions and non-linear feedback processes are incorporated in the fault growth mechanism.

Typical patterns of development of a population of faults are presented involving nucleation, growth, branching, interaction and coalescence. The geometries of the fault populations spontaneously converge to a configuration in which strain is concentrated on a dominant fault. In these numerical simulations, the material properties are uniform, so the complex behaviours results solely from the random fluctuations and physical interactions. Consequently, the entire process of fault development is an emergent property of the model of fault interaction and does not depend on pre-existing material heterogeneity. Thus, homogenisation of the strain rate along faults and structural regularization of the fault trace can be quantified by analysis of the output patterns. The temporally stochastic element allows relocation of faults by branching from bends and irregularities of the fault traces. This relocation mechanism involves partitioning of the strain and competition between faults. Possible relationships between the seismic regime and the geometry of the fault population suggest that the fault system may be attracted by a critical point. Toward this critical point, the correlation length (i.e. the size of the largest fault) increases, the frequency of large event is higher and larger faults are able to slip at lower stresses.