Effects of lateral heterogeneities on receiver-function imaging: the case of subducting slab

Signal enhancement is an important part of receiver-function imaging. One common technique is CCP (common conversion point) stacking, in which receiver-function traces are gathered based on the locations where P to S conversion occurs. The conversion locations are usually computed using a one dimensional reference velocity model. CCP stacking thus favors horizontally layered structure, and may produce an incorrect image of laterally variable and dipping structures, such as a descending slab of oceanic lithosphere. Here we present an investigation of the consequences of lateral heterogeneities on a CCP image. We chose subducting slabs as our test case for several reasons. Subducting slabs are one of the major lateral heterogeneities in tomographic images of the mantle. The large-scale ( ∼100 km) geometry and physical properties of subducting slabs are reasonably well constrained from travel-time tomographic modeling, and geodynamic considerations, yet the fine features, such as the state and behavior of the oceanic-crust portion of a subducting slab, are still unclear. Such fine-scale structures are crucial to understanding material recycling, melting processes, arc volcanism, continental crust formation, and the uneven distribution of subduction zone seismicity. Water is believed to be involved in all these processes; strong P to S conversion and reflectivity are a good tracer for subsurface water, converted wave imaging thus offers the promise of providing understanding of these processes. We generated synthetic seismograms using a two-dimensional elastic wave finite-difference code. Teleseismic wavefields are approximated by an input plane-wave. In our first test cases, slabs descend from surface to the 660-km discontinuity with various dip angles (30° -60° ). The thickness of the slab is assumed to be 100 km, and has a sharp upper boundary but a continuous lower boundary. P-wave velocity within the slab is set to be 2-5% higher than the surrounding mantle, and δ lnVs/δ lnVp is fixed to be 1.5. The dipping slabs were illuminated with inputs at a range of incidence angles (20° -45° ) equivalent to epicentral distances of 30 to 90 degrees, from both the ocean and back-arc sides. Wave fields generated from back-arc events are generally more complicated than those from earthquakes on the ocean side. In the low frequency (~0.15 Hz) band, although images of subducting slabs can be reconstructed from the synthetic seismograms, the dip angle is incorrectly mapped by more than 10° when the CCP stacking technique is applied. It is also impossible to resolve fine structure, such as the oceanic crust, in this frequency band. Artificial structures are also found in the CCP images using back-arc events only. At high frequency, the CCP images are strongly affected by the bin size used, which in turn is directly related to receiver array spacing. A receiver spacing less than 10 km is required in order to obtain an unaliased image. Such a spacing is available in some permanent networks (e.g., the high sensitivity seismograph network, HiNet, in Japan), some PASSCAL experiments, and is possible with the flexible array component of USArray, allowing us to investigate the structural geology of the upper mantle in subduction zones.