Guanhui Wang
Ansys Employee
For a single waveguide or fiber structure which includes a bent region, if the bending is circular, it is possible to simulate this by setting up the structure geometry of the bent portion as straight, but in the custom settings for that cell group region, select the “bent waveguide” option and specify the bending radius. This will find the bent waveguide modes in the selected cells similarly to the bent waveguide modes found by the FDE solver:

https://kb.lumerical.com/en/index.html?solvers_finite_difference_eigenmode_bend.html 135

For a more complex device like a ring resonator it is possible for the EME solver to simulate devices where light propagates at a steep angle, and even at 90 degrees like in the ring portion of a ring resonator, but simulating steep angles comes at the cost of increased simulation time and memory requirements.

The modes used in each cell are calculated by finding the Eigenmodes supported by the x-normal cross section of structure at center of each cell – although the individual modes won’t look like the actual fields supported by the waveguide, as long as enough modes are used, they can form a basis set of modes for the actual fields to be expanded onto in the EME propagation step. The problem is that since many modes are required for accurate results, there may be no advantage to using the EME solver compared to the varFDTD or FDTD solvers when simulating propagation at steep angles.

For example, some analysis of a grating coupler device is presented in the following whitepaper where it is found that since the grating coupler supports steep angle propagation, 2D FDTD is faster than EME for this application:

https://www.lumerical.com/learn/whitepapers/optical-solvers-for-integrated-optical-components/