An improved multilevel green´s function interpolation scheme for 3D full-wave problems

A series of oscillatory radial basis functions (RBFs) with Bessel functions of different orders are compared at different cube lengths. An adaptive choice of the interpolation functions with the best orders is then adopted to enhance the interpolation accuracy when implementing the multilevel Green´s function interpolation method (MLGFIM). In order to further improve the interpolation accuracy, a new type of boundary clustering procedure is proposed to reduce the error near the cube boundary which dominates the total error. Numerical results reveal that the interpolation accuracy is significantly improved using the adaptive oscillatory RBFs with non-uniform new staggered Tartan grid which is generated by the boundary clustering procedure.