Splitting error reductions for the ADI-FDTD method

The ADI-FDTD method has attracted much attention for its unconditional stability and efficient simulations in the time domain with large time-steps. However, the associated errors are found to be relatively large in comparisons with another unconditionally stable FDTD scheme, the Crank-Nicolson (CN) technique, although the ADI-FDTD method presents higher computational efficiency. In this paper, we propose new ADI-FDTD methods that are based on the CN method but with the computational efficiency similar to that of the original ADI scheme. Numerical results are used to validate the methods.