We determined the source process of the 1994 Northridge earthquake in relation to the aftershock mechanisms. To study the source complexity of the mainshock, we inverted the P and SH waveforms recorded by the IRIS and IDA/IRIS networks, using the method of Kikuchi and Kanamori (1991) in which the rupture is represented by a series of discrete subevents with varying mechanisms. The waveforms show that the rupture consisted of several subevents with about 2 sec in between. Our solution consists of three subevents with essentially the same mechanism, viz., strike, dip, and slip of 130°, 42°, and 116°, respectively. The first subevent occurred at a depth of about 19 km, followed after 2 sec by the second and largest subevent at a depth of 17 km and the third subevent again 2 sec after the second at a depth of about 13 km. The total moment from the body waves of this sequence is about 1.1 × 1026 dyne · cm (Mw = 6.6) with a source duration of 7 sec.
The large depths of these subevents explain the lack of any surface rupture. Furthermore, the upward propagation of the subevents is consistent with the depth of the hypocenter and the distribution of the aftershocks, which are shallower and more northerly than the mainshock hypocenter.
The aftershocks were analyzed using data from the TERRAscope network. We inverted short-period surface waves to determine the moment tensor for 70 events with Mw > 3.5. The aftershocks can be grouped into three regions based on the mechanisms: the eastern part of the aftershock zone, where we find thrust events with mechanisms very similar to the main event; a central area with predominantly strike-slip events; and an area to the west, where we find oblique thrust events but with more northerly P axes than in the eastern region. This distribution suggests that the fault system on which the Northridge earthquake occurred is segmented and that the extent of the Northridge rupture is controlled by a change in geometry of the fault. We find a high stress drop (270 bar) for the mainshock; we propose that the changes in the fault geometry prevented a slip pulse from propagating, thereby causing a high ratio of slip-to-rupture length.