Constraints on the Physical Mechanism of Deep Earthquakes from Observations of Source Finiteness

Since the discovery of deep earthquakes in the 1920s, their physical mechanism has been debated. We seek to place constraints on their mechanism with observations of source finiteness from many deep events. Is the rupture process primarily isobaric, predicting primarily horizontal rupture propatation? Do these deep events represent the reactivation of faults created prior to subduction, or the creation of new faults? Fault reactivation would be most consistent with a characteristic fault-plane orientation for outer rise events that persists with depth in the slab. To test this hypothesis, it is necessary to identify the rupture plane. Resolving the fault plane ambiguity is a classical problem in seismology, and we present a semi-automated method, designed to be applied to large numbers of events, to accomplish this task. Source finiteness is observable on seismograms at different azimuths and distances as variations in the apparent rupture duration. For each earthquake, the rupture duration will be shortest in the direction of rupture propagation and longest in the opposite direction. Rather than measuring the actual rupture duration at each station, we use a cross-correlation technique that includes a stretching factor to measure the differential rupture duration between each pair of stations. These differential measurements then allow us to estimate the rupture direction and rupture velocity, thereby identifying the fault plane as the nodal plane that contains the rupture vector. To demonstrate the method, we apply it to P waves from broadband seismograms from four intermediate- and deep-focus earthquakes composed of two subevents: the 23 January 1997 Bolivian earthquake (M_W 7.1, 276 km depth), the 27 October 1994 earthquake south of the Fiji Islands (M_W 6.7, 549 km depth), the 21 July 1994 Japan Sea earthquake (M_W 7.3, 471 km depth), and the 11 November 1998 Fiji Islands earthquake (M_W 6.3, 149 km depth). Each focal mechanism contains a subvertical and a subhorizontal nodal plane. For three of the events, our analysis shows that rupture propagated subhorizontally, and we identify the subhorizontal nodal plane as the fault plane. For the smallest event, the rupture azimuth, but not the rupture dip, is well-constrained, so we cannot conclusively identify the fault plane. Given these initial results, we conclude that this method holds promise for testing whether or not deep earthquakes occur on faults that formed in the oceanic lithosphere prior to subduction.