rsoft fdtd

Recently, I'm using Rsoft's FDTD to simulate some dispersive media, that is, n+ki. I find when k is very large, suppose, 2+11i, the FDTD will scater. So I want to make sense of the grid size and time step. So far, I define grid size and time step based on lossless media, that is, grid size<lambda/10, time step<grid size/2. For dispersive media, is there any change for time step or grid size? Is there any rule for this? Could anyone help me? Many thanks.

The wavelength varies nonlinearly in dispersive media. You have to consider the longest lambda which can appear in your simulation, in determing the grid size and time step.

I think is not the longest wavelength, but the shortest wavelength in your pulse.

I read that in this case when dispersion occurs, the step size should be very small. But I don't know how small...... Also struggling about the settings similarly.

I have another question though, could you guys get normalized spectrum in the FFT output of the monitor? I always have troubles there??