Common-conversion-point (CCP) stacking of receiver functions is a widely used technique to image velocity discontinuities in the mantle. The CCP imaging technique assumes that receiver functions are composed solely of P to S conversions at velocity boundaries, whose depths can be mapped out through their arrival times. The multiple reflections at shallow boundaries with large velocity contrasts, such as the base of unconsolidated sediments and the Moho, can lead to artificial structures in the CCP images. We develop a refined CCP stacking method that uses relative slowness as a weighting factor to suppress the multiples (slowness weighted CCP stacking; SWCCP). We conduct extensive numerical tests with synthetic data to seek the best weighting scheme and to verify the robustness of the images. We apply this technique to receiver function data of NECESSArray in China and the transportable array in western US, and find most of the events in the depth range of 200-400 km shown in the regular CCP images are eliminated. The SWCCP images, on the other hand, reveal a clear negative event under some parts of the two arrays, indicating the presence of low velocity layer above the 410-km discontinuity, which was reported by previous studies.

In CCP stacking of receiver functions, most of the current studies computed 3D relative Pds traveltime corrections by integrating traveltime anomalies along 1D ray paths. This ray-tracing approach is generally time consuming and less accurate when prominent velocity anomalies exist and effects of the 3D ray paths become significant. In this study we introduce a new scheme that utilizes the fast-marching method eikonal solver to improve both the efficiency and accuracy of 3D Pds traveltime computation. We first employ a 1D ray tracing method and the iasp91 model to calibrate the accuracy of the new scheme and optimize the parameters of the numerical solver. We then apply the new scheme to compute a massive number of Pds traveltimes using two 3D synthetic models, one with a high-velocity slab and another one with a low-velocity plume, and compare these 3D traveltimes with those computed with the ray-tracing approach. We found 3% ray paths in the slab model and 12% ray paths in the plume model show a 3D traveltime difference of more than 0.5s. We apply the proposed scheme to the TA receiver functions that sample the transition zone structure beneath the Yellowstone hotspot and find that CCP stack using 3D Pds traveltimes computed by eikonal solver has better focused P660s than that by the ray-tracing method. Finally we illustrate that the computational times can be reduced by 1 to 2 orders of magnitude with the new scheme to compute the 3D Pds traveltimes of 20,000-200,000 receiver functions.