Waveguide effective index changes with mesh size


I want to calculate the effective index of a signle strip waveguide 410nm*220nm.
However, the effective index changes with meshsize.
The simulation wavelength is set as 1310nm.
When meshsize=10nm, Neff=2.585929
When meshsize=20nm, Neff=2.581679

Which value is more accurate? How to solve this problem?
The neff difference may lead to a big variation in some phase sensitive device.
swg_neff.lms (238.1 KB)


Hi @lishifeng,

It is normal that you see a variation in the effective index calculated with FDE as you refine the mesh. This just reflects the effect of the spatial discretization. Ideally, what you should see as you refine the mesh is that the change in the results becomes smaller, i.e. there is convergence towards some value. It is important to do this convergence test to determine the level of precision of your results. In this modified file swg_neff_modFG.lms (241.1 KB) I included a sweep over meshsize in the Optimization and Sweeps tool.

The result of the sweep shows that the calculated neff varies as you refine the mesh but it gradually converges to some value.

You can see this kind of convergence testing in these examples of waveguides.


Thank you very much. I will try as you suggested.