Can't find modes in narrow (120 nm) waveguide using PML BC


I was trying to find the foundamental TE mode in a 120nm wide, 220 nm thick Si waveguide (with SiO2 everywhere) using Mode solver. However, I can’t find any mode when using the PML boundary condition. I used the default PML settings and I extended my calculation domain to be 60 x 60 um. Really want to know what should I do and what went wrong…


Hi @yug096,

The issue might be related to the mesh size as well. I managed to get the mode using a FDE region of 10um x 10um, 200 mesh cells in x and y, and a mesh override region over the waveguide (mesh of 5nm, with the mesh override surrounding the wg with a 100nm buffer.


Hi @gbaethge,

Thank you so much for the reply! It indeed works! Previously my mesh size is 20 nm over the waveguide. Is it true that when using PML BC, the mesh size has to be very small?


Hi @yug096,

It is not necessarily linked to the PML, your waveguide is fairly small, and you need enough mesh cells to revolve it and the modes.

You get the same issue when using metallic BC: with a coarse mesh, you only get the modes in the waveguide created by the BC (you can recognize them as the effective index is lower than the cladding material index).

You also need the FDE region to be large enough. It’s another tricky point with your waveguide, as due to the size, the mode is far from being confined in the waveguide. Ideally, you want to have the fields as small as possible at the boundaries.


Hi @gbaethge,

Thanks for getting back to me so quickly.

I was asking this because when I used coarse mesh with metallic BC in this case,I was able to find the foundamental mode. However, I couldn’t find the mode using PML unless I try the very fine mesh as you suggested.

How small the fields should be at the boundary for metallic BC, or PML BC? Like, -5 or -10 db?


There’s no absolute rules regarding how small the fields should be, but a rule of thumb would be something like -9dB. Ideally, you would do some testing, like increasing the FDE region size to see how the mode properties vary. You should see some convergence at some point.


Ok, I see. Thanks a lot for the great help!