I am trying to resolve a narrow resonance in a photonic crystal slab. The attached jpg-file compares the power spectrum of my optiFDTD simulations with a plane-wave-expansion method. In order to attain the needed spectral resolution I’ve concluded that I must refine the mesh grid (make dx and dy smaller) AND increase the signal length (as given by the three reasons listed at the end of my post). This leads to an incredible increase in needed timesteps. I have successfully simulated with N=360’000 timesteps using a time increment dt=113fs, HOWEVER, when I try to increase the number of timesteps to N=480’000 the simulation fails and I end up with an analyzer file that cannot be opened. What is the reason for this? The program seems to crash after having run for several hours: I have recorded memory usage, and this does not seem to be the issue. The analyzer file size is 1603824 KB; I’ve earlier obtained files significantly larger than this.
In order to obtain fine spectral resolution one must think about:
1) Discrete Fourier Transform requires long simulation time: T=N*dt, where N is the number of timesteps and dt is the timesteps. The spectral resolution d\lambda is proportional to 1/T.
2) Artifacts may occur when the spectral resolution d\lambda is smaller than the mesh-discretization dx, dy.
3) Finer discretization dx, dy, dz requires smaller timesteps dt.
Does this sound right?
.