Two-dimensional fully dynamic SEM simulations of the 2011 Tohoku earthquake cycle

Earthquake cycle simulations have been performed to successfully reproduce the historical earthquake occurrences. Most of them are quasi-dynamic, where inertial effects are approximated using the radiation damping proposed by Rice [1993]. Lapusta et al. [2000, 2009] developed a methodology capable of the detailed description of seismic and aseismic slip and gradual process of earthquake nucleation in the entire earthquake cycle. Their fully dynamic simulations have produced earthquake cycles considerably different from quasi-dynamic ones. Those simulations have, however, never been performed for interplate earthquakes at subduction zones. Many studies showed that on dipping faults such as interplate earthquakes at subduction zones, normal stress is changed during faulting due to the interaction with Earth's free surface. This change in normal stress not only affects the earthquake rupture process, but also causes the residual stress variation that might affect the long-term histories of earthquake cycle. Accounting for such effects, we perform two-dimensional simulations of the 2011 Tohoku earthquake cycle. Our model is in-plane and a laboratory derived rate and state friction acts on a dipping fault embedded on an elastic half-space that reaches the free surface. We extended the spectral element method (SEM) code [Ampuero, 2002] to incorporate a conforming mesh of triangles and quadrangles introduced in Komatitsch et al. [2001], which enables us to analyze the complex geometry with ease. The problem is solved by the methodology almost the same as Kaneko et al. [2011], which is the combined scheme switching in turn a fully dynamic SEM and a quasi-static SEM. The difference is the dip-slip thrust fault in our study in contrast to the vertical strike slip fault. With this method, we can analyze how the dynamic rupture with surface breakout interacting with the free surface affects the long-term earthquake cycle. We discuss the fully dynamic earthquake cycle results focusing on the differences from previous quasi-dynamic studies such as Kato and Yoshida [2011]. They proposed a shallow strong patch model to explain the observed huge coseismic slip at the shallow portion close to the Japan Trench and the long recurrence time of several hundreds.