Structure of the 660-km discontinuity below a deflected slab: complex or simple?

Previous studies using GDSN stations in NE China, Mongolia, and East Siberia suggested a multiple-discontinuity structure from about 660 km to a depth of 780 km in an area occupied by the horizontally deflected Pacific slab. Such complex structure, if confirmed, could lead to new insights into the mineralogical and geodynamic processes related to the interaction between subducted slabs and the 660-km discontinuity (d660). We have collected several hundreds broadband seismograms for each of the GDSN stations (MDJ, HIA, BJT, and ULN) and computed receiver functions. We then stacked the receiver functions based on the predicted travel time of P-to-S converted phases. When teleseismic events from all the back-azimuths are used, there is indeed a multiple-discontinuity structure in the depth range of 650 to 750 km beneath station MDJ. BJT, HIA, and perhaps ULN show a sharp d660. This observation is consistent with previous studies. However, when events from narrow back-azimuthal bands are stacked, the depth of d660 beneath MDJ shows a systematic variation with back-azimuths, ranging from 660 km for events from the west, to 700 km for events from the south. Such a variation is the result of a broad NS-striking depression of d660 associated with the cold slab. We estimated that the magnitude of the depression is about 60 km over an EW distance of about 350 km. For most of the back-azimuthal bands, d660 is a sharp, simple feature. The two bands from the SE show two arrivals at the depths of about 680 and 740 km, respectively.