Unconditionally Stable ADI/Crank–Nicolson Implementation and Lossy Split Error Revisited

This paper first shows that the existing ADI and CN implementations initiated by Namiki , follows an indirect way of solving a linear system. The ADI and CN update equations can and should be implemented following the standard FDTD formulation where the update equations translate directly from Ampere's and Faraday's laws. The advantages leading to a more complete lossy split error analysis and possibilities of mitigating the numerical errors are also presented. In high loss applications the CN method greatly outperforms the ADI method in terms of accuracy. The root cause of the ADI degradation is discussed.