Three-dimensional finite-different time-domain (3D FDTD) simulation of
photonic crystal devices often demands large amount of computational
resources. In many cases it is unlikely to carry out the task on a serial
computer. We have therefore parallelized a 3D FDTD code using MPI. Initially
we used a one-dimensional topology so that the computational domain was
divided into slices perpendicular to the direction of the power flow. Even
though the speed-up of this implementation left considerable room for
improvement, we were nevertheless able to solve largescale and long-running
problems. Two such cases were studied: the power transmission in a
two-dimensional photonic crystal waveguide in a multilayered structure, and
the power coupling from a wire waveguide to a photonic crystal slab. In the
first case, a power dip due to TE/TM modes conversion is observed and in the
second case, the structure is optimized to improve the coupling. We have
also recently completed a full three-dimensional topology parallelization of
the FDTD code.