We analyze the rupture history of this earthquake, the largest thrust earthquake since 1977, primarily using broadband SH wave seismograms. We show that the earthquake occurred on a shallow dipping thrust fault (strike 109¿, dip 9¿, rake 72¿) with a seismic moment of 2.7 ¿ 1021 N m. The rupture propagated bilaterally at an average speed of ~90% of the shear wave speed, on a fault extending 180 km west and 50 km east of the hypocenter, with very variable width ranging from 30 to 100 km at different locations along strike. The mean slip over a 230 km ¿ 100 km fault area was 4 m and the mean stress drop 1.9 MPa. The slip distribution is very nonuniform over the fault, with the largest slip of ~12 m being near the hypocentral depth (~10 km). The rupture is very complex, propagating first to the west and then, after a ~15-s delay, to the east. We interpret this delay as being due to the existence of a inhomogeneous barrier just east of the hypocenter, which initially acted as a barrier to rupture propagation toward the east, but subsequently failed due to stress increase on it generated by the rupture to its west and with the rupture then continuing to propagate eastward. This barrier acted as a stress concentrator before the earthquake, giving rise to foreshocks and was also the initiator of rupture for this earthquake. The aftershock zone closely corresponds to the region in which rupture occurred, the area with greatest aftershock density lying entirely within the area of highest moment release.