In nonlinear simulations, it is necessary to modify some advanced settings to get correct results. For example, if you want to get nice high order harmonics from a material with a chi2 coefficient, it is important to set correctly the sampling rate in the monitors and make sure the mesh generation uses the appropriate bandwidth.
This can be illustrated using the simulation file harmonic.fsp (252.6 KB). If we use the original settings in this file, we get a nice spectrum from the time monitor showing a peak at the source wavelength and peaks at smaller wavelengths, which correspond to the second and higher order harmonics:
If we reduce the “min sampling per cycle” in the advanced settings of the time monitor (for example, to be 10 instead of 100), the spectrum is not as clean as before:
Also, if we are not careful to check the option to override simulation bandwidth for mesh generation in the Advanced Options of the FDTD simulation settings, the mesh will be too coarse to resolve some of the higher order harmonics:
A complete description of this and other special settings for nonlinear simulations can be found in: https://kb.lumerical.com/en/index.html?nonlinear_methodology.html.