What is the negative loss in eigensolver analysis



Hi, we simulated a Mach zehnder modulator in MODE. When we set the boundary condition as PML, the eignsolver analysis shows the effective index of to be included imaginary part and negative loss. How the imaginary part is applied if the material data is no imaginary part data such as silicon in the wavelength of 1310 nm to 1550 um. To show the negative loss is wrong simulation or a special physics about PML? Thanks.


Hi @askalee
In Lumerical’s knowledge base there is an example of a Mach Zehnder Modulator :
If you open the “MZI.lms” file you will notice that there are metal boundary conditions. I run the simulation and i didn’t find negative loss. There can be some artificial loss or gain introduced when using PML boundaries with the FDE solver. I would recommend you to use metal boundary conditions.

I hope this helps.


Hi @askalee,

As @konslekk explained, the negative loss that you are seeing is probably the result of using PML as your boundary conditions that introduces artificial loss or gain. A good option to start the simulation would be to use metal. Please refer to this page that further explains the difference between using these two boundary conditions (BCs).
Hope this helps.


Hi, @bkhanaliloo.
Is there any options to reduce this artificial gain from PML BC?
Thank you,


Dear @annakod

I guess one way would to expand the simulation region so that your mode decays enough (few orders of magnitude) before reaching the PML layer. To check and see if the mode has decayed enough, you can plot mode in the log scale. Then perform a convergence test, i.e. increase the simulation region even further until frequency and neff of the mode does not change by increasing the simulation region. You can learn more about convergence test here.