I. Introduction
Since the finite-diffence time-domain (FDTD) algorithm was developed by Yee [1] in the middle 1960s, it has become one of most powerful tools in full-wave analysis for time-domain electromagnetic solutions in device, material, antenna, and scattering simulations [2]–[5]. On many occasions, the FDTD provides accurate results both in time and frequency domains. However, its computational efficiency is limited by two inherent physical constraints: the numerical dispersion and the Courant–Friedrich–Levy (CFL) stability condition. The former requires fine spatial discretization for a particular accuracy, while the latter demands a proper time step for computational stability. Both of them lead to large memory and CPU time consumptions.