Guilin Sun
Ansys Employee

The built-in wavelength sweep is simplified method as mentioned previously. It can be good when the mode dispersion is small, or the bandwidth is small. In your case it seems the bandwidth is large, and I would guess the neff dispersion is large, which causes this problem.

You may try to use script to do the wavelength sweep which should give you better result. The price you pay is significantly long simulaiton time. If you use 100 frequency points, it will be 100 times slower.

 

In addition, you may compare the mode profiles at shorter wavelength and the longer wavelength, as well as neff since you have the data.