HR: 11:35h
AN: S21G-06    [PDF]
TI: Numerical Models of Stopping Ruptures on a Bimaterial Interface
AU: * Rubin, A M
EM: arubin@princeton.edu
AF: Department of Geosciences, Princeton University, Princeton, NJ 08544 United States
AU: Ampuero, J
EM: jampuero@princeton.edu
AF: Department of Geosciences, Princeton University, Princeton, NJ 08544 United States
AB: Using a cross-correlation earthquake relocation technique, Rubin and Gillard (2000) and Rubin (2002) found that the nearest aftershocks of microearthquakes on the San Andreas fault were much more likely (by a ratio of nearly 3:1) to occur to the NW of the mainshock than to the SE. They attributed this asymmetry to the material contrast across the fault and the resulting dynamical reduction in normal stress near the rupture front propagating to the SE (the front moving in the direction of slip of the more compliant medium). Specifically, it was hypothesized that regions of the fault far enough from failure to resist this extra dynamical "kick" would be that much farther from failure once those dynamical stresses decayed. However, analytical (steady-state) models of propagating slip on a bimaterial interface (Weertman, 1980) show that, as with the static stress field, normal stress changes occur only behind the rupture front. The proposed explanation works most simply if the region ahead of the SE rupture front experiences a transient stress favorable for slip. In principal this stress transient could be associated with either rupture growth or arrest. To investigate this further, we ran 2-D numerical models of slip on a bimaterial interface with slip-weakening friction, using the code of Cochard and Rice (2000). The ruptures spontaneously accelerate to the generalized Rayleigh wave speed of the medium, when such exists. During this growth phase, large tensile stresses are indeed restricted to regions of large slip velocity behind the SE-propagating rupture front. Ahead of the rupture front the normal stresses are smaller and compressive. If the rupture front is stopped abruptly, the short-wavelength tensile stress pulse continues to propagate at roughly the same velocity. The above comments also apply in an anti-symmetric sense to the NW rupture front, although there the slip speeds and normal stress changes are lower. If the rupture is stopped by a more gradual reduction in the loading stress, the moving tensile pulse can spawn a decaying slip pulse at the SE front but not the NW. If this slip pulse marks the furthest extent of slip, the resulting static stress field is quite asymmetric even for a symmetric initial stress, lying on the failure envelope at the NW end of the rupture but well below it at the SE end. These results are at least permissive of the explanation proposed by Rubin and Gillard. For weaker slip pulses (due to any of a number of factors contributing to smaller maximum slip speeds), the furthest extent of slip near the SE rupture front can be driven by the stopping phase arriving from the NW end of the crack. Under such conditions the final stress field is more symmetric. We will be running models using heterogeneous stress fields to explore these questions further, and hope to use rate-and-state friction to investigate the observed temporal decay of the aftershock asymmetry.
DE: 7209 Earthquake dynamics and mechanics
SC: Seismology [S]
MN: 2003 Fall Meeting