From the result, it is most likely due to short of simulation time and/or larger autoshutoff min. Please significantly increase the simulation time, eg 100 times of your current setting, and reduce the autoshutoff min to 1e-6 (or so), run the simulation till it reaches the autoshutoff min. If it terminates at 100% of the progress, it means the simulation time is too short.
In addition, it seems this device has very large Q at certain wavelengths, which means it will need smaller autoshutoff min.
Please visit this Chinese link for more information:
Ansys Insight: 为什么说仿真时间要足够长才能得到正确的频域结果 and this one
Ansys Insight: 透射率反射率大于1，或者透射率反射率之和大于1的原因与解决方案--综合贴
A side note: not sure how long the xspan of FDTD. It seems there are ling uniform waveguides at the input and output sides. Maybe you can make it smaller so the simulation can be faster. When dipole source is used, it can have larger radiation angles, maybe you use larger length in purpose to avoid simulation diverging?
Second note: Not sure what the quantity you want. But for such devices, usually a specified mode source is used to excite in experiments. So it might be better to use a mode source on the input waveguide.
You can ignore those notes as it is purely my observations. If you have questions regarding to those, please write a new post.
You are navigating away from the AIS Discovery experience