Why not? FFT is a lot faster than the alternatives, and as long as you have enough vacuum, any boundary condition works just as fine. Of course, if you actually have a gate in any direction, then it changes, but forcing the field to be zero if the vacuum is not large enough (and no gate present) is actually more wrong than a periodic solution which at least would have zero derivative at the edge of the cell (ok, so would Neumann, but you can only place that on one side or else it's equivalent to periodic
). So in short, FFT is the default basically because it's not wrong, and faster.
