After the 2011 Mw 9.1 Tohoku earthquake, numerous intraplate earthquakes occurred beneath the outer slope of the Japan Trench. Based on ocean bottom seismograph observations, these earthquakes occurred in the oceanic crust and uppermost mantle of the Pacific plate at depths shallower than about 40 km and had normal-faulting focal mechanisms at all depths. Before the 2011 earthquake, normal-faulting earthquakes beneath the outer trench slope occurred only at depths shallower than 20 km, whereas those at depths of around 40 km had reverse-faulting mechanisms. These observations suggest that the stress regime in the Pacific plate was changed by the 2011 earthquake. The tensional stresses that now extend to depths of about 40 km may play an important role not only in the occurrence of large normal-faulting earthquakes but also in hydration of the uppermost mantle of the incoming Pacific plate prior to the subduction.
 The 2011 off the Pacific coast of Tohoku earthquake (Mw 9.1) occurred on 11 March, 2011, along the boundary between the subducting Pacific and overriding North American plates [e.g., Nettles et al., 2011]. The earthquake had a shallow-thrusting focal mechanism reflecting megathrust displacements of several tens of meters [e.g., Yagi and Fukahata, 2011]. About 40 minutes after the mainshock, a Mw 7.6 earthquake occurred beneath the outer slope of the Japan Trench (Figure 1). This earthquake was one of numerous normal-faulting earthquakes that occurred beneath the outer trench slope of the incoming Pacific plate after the 2011 Tohoku earthquake [e.g., Asano et al., 2011].
 Large intraplate normal-faulting earthquakes are common beneath outer trench slopes after large interplate thrust earthquakes and they can cause large tsunamis [e.g., Lay et al., 2011]. The fault configuration and timing of intraplate normal-faulting earthquakes following interplate thrust earthquakes vary. The 1933 Sanriku earthquake (Mw 8.4 intraplate normal-faulting earthquake) occurred 37 years after the 1896 Sanriku earthquake (Mw 8.0 interplate thrust earthquake) [Kanamori, 1971]. The rupture of the 1933 Sanriku earthquake likely extended through the entire lithosphere, reaching a depth of about 70 km. During the 2006–2007 Kuril Islands earthquake sequence, a Mw 8.3 interplate thrust earthquake was followed two months later by a Mw 8.1 intraplate normal-faulting earthquake under the outer trench slope [Lay et al., 2009]. The rupture zone of this normal-faulting earthquake extended to a depth of about 33 km. Two years later, a Mw 7.4 intraplate reverse-faulting earthquake occurred below the rupture zone of the intraplate normal-faulting earthquake.
 Differences in the both extent of rupture and the interval between related interplate and intraplate earthquakes are important in the evaluation of hazards posed by outer trench slope earthquakes. Lay et al.  noted the potential for large extensional earthquakes in the outer trench slope area following the 2011 Tohoku earthquake. Whether such potentially tsunamigenic earthquakes will occur is unclear, as the geometry of the outer trench slope faults and their relationships to crustal structures are not well understood. Furthermore, because the source areas of the outer trench slope earthquakes are distant from the coast, it is difficult to obtain accurate hypocenter locations (especially hypocentral depth) from land-based seismic networks. Determination of accurate hypocenter locations and focal mechanisms based on offshore observations is key to understanding the stress regime of the incoming Pacific plate and the potential for large extensional earthquakes following the 2011 Tohoku earthquake.
 Understanding normal-faulting earthquakes in outer trench slope areas is also important for estimation of water budgets in subduction zones. Earthquakes in the lower plane of double seismic zones in subducting oceanic mantle may be triggered by dehydration of partially hydrated oceanic mantle [Peacock, 2001]. Hydration of oceanic mantle might occur in the trench–outer rise region, where normal-faulting prior to the subduction may promote infiltration of seawater several tens of kilometers into the oceanic lithosphere of the incoming plate.
 Precise fault geometries and the stress regime in the incoming Pacific plate are important to investigate occurrence of large intraplate normal-faulting earthquake and hydration of the incoming oceanic plate prior to the subduction. To address these matters, we used data from an array of ocean bottom seismographs (OBSs) deployed from late April to early July 2011 around the source area of the Mw 7.6 normal-faulting event that followed the 2011 Tohoku earthquake.
2. Observations and Analysis
 In late April 2011, we used R/V Kairei (Japan Agency for Marine-Earth Science and Technology; JAMSTEC) to deploy an array of 20 OBSs on the outer slope of the Japan Trench (Figure 1). The spatial interval between OBSs on the seafloor was approximately 25 km. The placement of the array of OBSs was determined on the basis of preliminary aftershock distributions determined by the Japan Meteorological Agency (JMA) and the Global Centroid-Moment-Tensor project (GCMT, http://www.globalcmt.org/) taking into consideration the water depth limit of 6000 m for OBS operation. Each OBS was equipped with a three-component 4.5-Hz short-period seismometer and hydrophone, and seismic signals were recorded continuously (sampling frequency 200 Hz with 24-bit analog to digital converter) until the OBSs were recovered by JAMSTEC's R/V Yokosuka in early July 2011.
 We manually picked P- and S-wave arrival times for seismic events detected in the continuous OBS records on the basis of the short-term and long-term amplitude ratios, in addition to earthquakes listed in the JMA catalogue. Where possible, we noted the polarity of P-wave first-motions.
 We estimated preliminary hypocenter locations using the program hypomh [Hirata and Matsu'ura, 1987] in a 1-D Vp structure with a fixed Vp/Vs value of 1.78 based on the structure determined by Hino et al.  using previous OBS data from the northeastern part of our study area (144.9°–145.3°E, 38.0°–38.4°N). This structure was used as the initial velocity model for tomographic analysis. Station corrections for arrival time delays caused by sediments beneath OBSs were estimated from arrival time differences between P-waves and PS-waves converted from P-wave at the base of the sedimentary layer. We assumed Vp/Vs to be 8.0 in the sediments on the basis of the seismic velocity structure in the trench–outer rise area of the northwestern Pacific [Fujie et al., 2010].
 After our initial estimation of hypocenters, we simultaneously estimated hypocenters and 1-D Vp and Vp/Vs structures by using the double-difference tomography method of Zhang and Thurber . Velocity grids for the tomography were centered on the OBS array at depths of 5, 7, 9, 11, 13, 15, 18, 21, 24, 27, 30, 35, and 50 km. We included in the tomography only earthquakes for which (1) the minimum S–P arrival time difference was less than 10 seconds, and (2) there were at least five arrivals for both P- and S-waves. According to these criteria, we picked P- and S-wave arrivals at 16 or more OBSs for 75% and 90% of events, respectively. Magnitudes were determined from the maximum amplitudes of the seismograms [Watanabe, 1971]. We located 1685 earthquakes for the period from 1 May to 30 June, 2011 with magnitudes ranging from 0.9 to 6.1. The velocity structure we obtained was similar to the model of Hino et al.  (Figure S1 in the auxiliary material).
 To examine the uncertainties of our estimated hypocenter locations, especially for locations outside the OBS array, we compared hypocenters estimated using data from all OBSs with those estimated using two subsets of OBS data. We found that when we excluded data from the four westernmost OBSs, or from the three northernmost OBSs, the differences between the hypocenter locations so determined and those determined from all OBS data were less than 5 km within the area 20 km from the OBS array (Figures S2 and S3 in the auxiliary material). Epicentral differences were less than 5 km for most earthquakes. However, depth differences tended to increase markedly away from the OBS array (more than 10 km for some events). Therefore, epicenters were estimated with uncertainties of less than 5 km, even though the depth uncertainties were more than 5 km for events outside the OBS array.
 Focal mechanisms were computed from the polarity of the P-wave first-motions by using the technique of Hardebeck and Shearer . Ray azimuths and take-off angles were based on the results of our tomography. We obtained good-quality solutions for 50 earthquakes with fault plane uncertainties of less than 35°, corresponding to quality categories A and B of Hardebeck and Shearer .
 Most of earthquakes recorded by the OBS array occurred at depths shallower than 20 km (Figure 2). Beneath the area within 20 km of the OBS array, where hypocentral depths were well determined, 70% of earthquakes were shallower than 14 km. We consider that these earthquakes occurred in the oceanic crust of the Pacific plate. Twenty percent of the observed earthquakes occurred at depths between 14 and 20 km, and only 5% occurred at depths greater than 30 km.
 Earthquake activity extended about 110 km east of the trench axis (Figure 2). Most of the earthquakes within the oceanic crust were laterally evenly distributed east of the trench. They also show several epicentral lineations that coincide either with horst and graben structures that are subparallel to the trench axis, or to other topographic lineations that are oblique to the trench axis. On the other hand, the earthquakes at depths greater than 20 km were less evenly distributed. Those west of the trench axis are part of the aftershock activity near the up-dip edge of the 2011 Tohoku earthquake rupture area, which had a maximum slip of several tens of meters along the plate interface [e.g., Yagi and Fukahata, 2011]. Although the uncertainties in the estimated depths of these earthquakes are greater than those closer to the OBS array, the estimated epicenters are just as reliable.
 Most of the focal mechanism solutions for events beneath the outer trench slope indicate normal-faulting (Figure 3). T-axis azimuths are oriented essentially E–W, except for the earthquakes shallower than 14 km (Figure 3b), which correspond to fault activity in oceanic crust and have T-axes oriented NW–SE, that is, rotated about 40° clockwise from those at greater depths.
 The focal mechanisms obtained from our OBS data suggest different stress regimes existed in the incoming Pacific plate before and after the 2011 Tohoku earthquake. The epicenter of the Mw 7.6 normal-faulting earthquake that occurred beneath the outer trench slope after the 2011 Tohoku earthquake is close to that of a Mw 7.0 normal-faulting event in 2005 [Nettles et al., 2011]. Aftershock activity of the 2005 earthquake was recorded by OBSs about a year and a half later in 2007 [Hino et al., 2009]. The aftershocks of the 2005 event were almost all shallower than 20 km, whereas the activity after the 2011 Tohoku earthquake extended to depths of around 40 km (Figure 4). The P-wave first-motion polarities of the aftershocks of the 2005 event at depths shallower than 20 km are consistent with the normal-faulting focal mechanism of the 2005 event [Hino et al., 2009]. In contrast, several events at depths of around 40 km have P-wave first-motion polarities that are consistent with reverse faulting [Hino et al., 2009]. These observations suggest that in 2007 (i.e., before the 2011 Tohoku earthquake), stresses in the Pacific plate were tensional in the upper layers of the oceanic lithosphere and compressional beneath. This stress regime can be explained by bending of the incoming/subducting lithosphere. Similar regimes of shallow tensional stress and deep compressional stress in the Pacific plate along the Japan Trench have been reported on the basis of data before the 2011 Tohoku earthquake [Seno and Gonzalez, 1987; Gamage et al., 2009]. However, focal mechanisms obtained from OBS data after the 2011 Tohoku earthquake show tensional stresses extending to depths of about 40 km in the Pacific plate without a change to compressional stress (Figure 3). The 2011 Tohoku earthquake would have caused Coulomb stress increases of about 0.5–1 MPa on normal-faults at the depths of the earthquakes recorded beneath the outer trench slope [Lay et al., 2011]. Therefore, we consider that the observed change of stress regime in the Pacific plate was caused by the 2011 Tohoku earthquake.
 The possible rupture area of the Mw 7.6 outer trench slope normal-faulting earthquake can be recognized from the hypocenter distribution and focal mechanisms obtained from our OBS data (Figure 4). There is an 80-km-long linear pattern of epicenters trending N12°E near the GCMT solution, roughly parallel to the trench axis. The hypocenter distribution indicates that the seismic activity occurred on a plane that dips to the west at about 45° and extends to a depth of about 35 km (Figure 4). Both the trend of the lineation and the dip of the plane are fairly consistent with the GCMT solutions. The westward dipping plane is clearly evident only for the southern half of the lineation. Because of their distance from the OBS array, the uncertainties of the hypocenters determined along the northern half of the lineation were greater (for depth in particular). The GCMT solution for the Mw 7.6 normal-faulting earthquake beneath the outer trench slope was about 50 km northwest of the epicenter determined by JMA (Figure 4). We consider that the GCMT solution is more reliable than the JMA epicenter, which was estimated using data recorded by the regional land seismic network of Japan, providing inferior azimuthal coverage compared to the GCMT solution calculated from the global seismic network. Hence, we consider that the westward dipping plane approximates the fault plane of the Mw 7.6 normal-faulting event that followed the 2011 Tohoku earthquake.
 The focal mechanisms of the earthquakes along the possible fault plane indicate normal-faulting, even though the T-axes of the shallow earthquakes are oriented NW–SE, similar to the orientation of other shallow earthquakes in the oceanic crust (Figure 3b). The fault plane is 80 km long and 40 km wide. An average slip of 1.4 m with rigidity of 70 GPa satisfies the seismic moment of 3.1 × 1020 Nm estimated by the GCMT solution. The Mw 7.6 outer trench slope normal-faulting earthquake, for which the rupture probably extended to a depth of around 35 km, occurred about 40 minutes after the 2011 Tohoku earthquake as a rapid lithospheric response to the change of stress in the incoming Pacific plate from compressional to tensional at a depth of around 40 km.
 The T-axes rotations of the earthquakes in the oceanic crust relative to the deeper earthquakes can be explained by taking into consideration pre-existing structures in the oceanic crust. The earthquakes in the oceanic crust are widely distributed in the area about 110 km east of the trench axis (Figure 2). A multichannel seismic profile across the Japan Trench at a latitude of 39°N shows the development of normal faults cutting the oceanic crust and forming horst and graben structures about 110 km seaward of the trench axis [Tsuru et al., 2000]. The epicentral lineations of the earthquakes in the oceanic crust along the N–S extending horst and graben structures are especially prominent near the trench axis, where fault throws of the horst and graben structures increase rapidly. On the other hand, the T-axes azimuths of the earthquakes in the oceanic crust are oriented NW–SE, although those of the deeper earthquakes are oriented E–W, almost normal to the horst and graben structures. The epicenters of the earthquakes in the oceanic crust show another lineation parallel to NE–SW trending topographic features. Magnetic anomaly lineations strike N50°–65°E in this area [Nakanishi, 2011]. The T-axes of the shallow focal mechanisms are almost normal to the magnetic anomaly lineations. The NE–SW trending topographic features can be explained by reactivation of the abyssal hill fabric [e.g., Nakanishi, 2011]. The earthquakes within the oceanic crust are likely caused by E–W tensional stresses related to the bending of the incoming Pacific plate, where NE–SW trending structures have existed. Hence, the earthquakes in the oceanic crust tend to follow the trend of pre-existing NE–SW trending structures and their T-axes appear to be rotated with respect to those of the deeper earthquakes.
 Hydration of the oceanic mantle might have occurred at depths corresponding to the lower plane of the double seismic zone under the stress regime that existed after the 2011 Tohoku earthquake. The earthquakes at depths of around 40 km in the incoming Pacific plate, 20–30 km below the oceanic Moho, correspond to the lower plane of the double seismic zone in the subducting plate [Gamage et al., 2009]. Dehydration embrittlement of the oceanic mantle, which may be hydrated in the trench–outer rise region, could have triggered the lower plane earthquakes [Peacock, 2001]. Before the 2011 Tohoku earthquake, stresses in the incoming Pacific plate were tensional in the shallow part of the plate and compressional at greater depth, a response to bending of the lithosphere [Seno and Gonzalez, 1987; Gamage et al., 2009]. Under these conditions, fluids would penetrate downwards only through the shallow tensional faults and hydration of the incoming plate would be limited to depths shallower than the zone of transition from tensional to compressional stress [Lefeldt et al., 2009]. However, after the 2011 Tohoku earthquake, tensional stresses extending to depths of around 40 km in the incoming Pacific plate might have allowed downward infiltration of seawater via normal faults, reaching depths corresponding to the lower plane of the double seismic zone.
 The earthquakes at depths of around 40 km were limited in areal extent although the shallow earthquakes in the oceanic crust were widely distributed. Seismicity at depths greater than 25 km is more active in the source area of the 1933 Mw 8.4 Sanriku earthquake than in the surrounding area [Gamage et al., 2009]. Furthermore, Nakajima et al.  pointed out that the subducting oceanic mantle is not uniformly hydrated and that large intraplate earthquakes within the subducting slab are a result of reactivation of buried hydrated faults in the oceanic mantle formed prior to the subduction. Although there have been few such deep earthquakes, their distribution suggests that the faults on which they occur are related to the heterogeneous structure of the oceanic mantle and that rupturing to these greater depths may have produced the large (M ∼ 8) normal-faulting earthquakes, such as pointed out by Lay et al. .
 Our OBS observations on the outer slope of the Japan Trench after the 2011 Tohoku earthquake showed that earthquakes occurred in the oceanic crust and the uppermost mantle of the Pacific plate at depths shallower than about 40 km. These earthquakes had normal-faulting focal mechanisms at all depths. Before the 2011 Tohoku earthquake, the stresses in the incoming Pacific plate were tensional and compressional in the shallow and deep parts of the plate, respectively. Our analyses indicate that the stress regime at around 40 km depth in the incoming Pacific plate had changed from compression to tension after the 2011 Tohoku earthquake. In addition to the tensional stress regime in the outer rise following a large interplate thrust earthquakes, Christensen and Ruff  reported that the stress regimes in the outer rise have changed with time from tensional, following a previous large interplate thrust earthquake, to compressional in the Kamchatka and New Hebrides. Temporal changes of stress regime are important factors that must be considered in the evaluation of the potential for the occurrence of large outer trench slope normal-faulting earthquakes in the Pacific plate, and for understanding hydration of oceanic mantle prior to the subduction.
 This work was supported in part by the Special Coordination Funds for the Promotion of Science and Technology (MEXT, Japan) titled “the integrated research for the 2011 off the Pacific coast of Tohoku earthquake”. A part of this work is contribution of the JAMSTEC's rapid response research project for the 2011 Tohoku-oki earthquake.
 The Editor thanks Anne Trehu for her assistance in evaluating this paper.