How does this differ from finite-element mode analysis?The quantities obtained are of the same kind, but the discretization and normalization differ. This method divides the cross-section by finite differences on a rectangular mesh, and fields are normalized so that the peak electric field intensity is 1. A finite-element solver can align triangular elements with material interfaces. The frequency sweep capability is on this side.
Are modes found automatically?No. You have to give the number of trial modes to return and the effective index to search around. Take a generous number of trial modes so that a physical mode is not missed. Using symmetric boundaries can leave only one polarization.
Can it give group index and dispersion?There is a frequency sweep capability, and it can output effective index, loss, group index, group velocity, group delay, dispersion and propagation constant as functions of wavelength. There are settings for tracking a particular mode and for computing dispersion in detail.
Can it evaluate bend loss?Modes of a bent waveguide can be solved by specifying the bend radius, orientation and reference point. Radiation from the bend is absorbed at PML boundaries and evaluated as loss. At the junction between curved and straight sections, the loss from mode overlap mismatch is evaluated separately.