Broadband fixed angle source (BFAST) support

[tomflexcompute]

The current Bloch boundary causes a wavelength-dependent incident angle, which is not ideal in broadband simulations with oblique incidence. Users usually want to fix the incident angle as it's commonly the case in experiments. Do we have any plans to implement BFAST in the future?

[momchil-flex]

One thing to note: our GaussianBeamSource, when used in the broadband setting, i.e. with sufficiently large num_freqs for a given bandwidth (which would typically be no more than ~10 even for large bandwidth), is actually fixed-angle. This will only work well if the beam decays by the simulation domain boundaries though. The problem with e.g. emulating a fixed-angle plane wave by making a very large-area gaussian beam source is the boundary conditions. BFAST requires a special type of boundary conditions for periodic domains with fixed-angle illumination. But, in some cases, the focused fixed-angle Gaussian beam could still be useful (before we implement BFAST).

[tomflexcompute]

One thing to note: our GaussianBeamSource, when used in the broadband setting, i.e. with sufficiently large num_freqs for a given bandwidth (which would typically be no more than ~10 even for large bandwidth), is actually fixed-angle. This will only work well if the beam decays by the simulation domain boundaries though. The problem with e.g. emulating a fixed-angle plane wave by making a very large-area gaussian beam source is the boundary conditions. BFAST requires a special type of boundary conditions for periodic domains with fixed-angle illumination. But, in some cases, the focused fixed-angle Gaussian beam could still be useful (before we implement BFAST).

Right this is exactly my current recommendation for users. Unfortunately in many cases the beam has to be very large and thus the simulation cost is prohibitively high.