Learning Forum

[emh711]

Lumerical recommends extending structure through the PML boundary in z-direction.

I notice that many samples posted on Lumerical website doesn't necessarily have their structures extend through the PML boundary in z-direction. When do you need to extend structure through PML boundary in z-direction? When is it okay to not have the structure extend through PML boundary in z-direction?

[Greg Baethge]

Thank you for posting your question. Whether the structure should extend through the PML(s) will depend on your simulation. PMLs are used to simulate the propagation in an infinite space while using a finite sized simulation region: light reaching the PMLs will be absorbed as if it would simply propagates out of the simulation region and never be reflected back.
If the structure doesn't extend through the PML, there is an additional interface between the structure and the background material that can cause some reflections. The question is then, are these reflections wanted or not. For example, let's look at this simulation:
This is a 50nm layer of silicon on a glass substrate. Here, the substrate extends through the bottom PML, so the light reflected by the silicon layer will be absorbed by the PML. If I reduce the thickness of the substrate:
It added an interface that will change the reflection seen by the monitor:
In this example, we extend the substrate through the bottom PML as we only want 2 interfaces: SiO2/Si and Si/vacuum.
Note if the structure touches the inner PML boundary, by default, the solver will automatically extend it (see this page).
I hope this will help.

[emh711]

So, the concern of not extending the structure through the PML boundary is the reflection.
Would it help if the PML boundary is placed far away from the structure in the z-direction?

[Greg Baethge]

Not necessarily: if the PML is far away, it means the simulation region is large which impacts the memory requirements and simulation time. That said, the PML should be "far enough" so it doesn't affect the evanescent fields at the surface of the structure. Typically, a distance of half the maximum wavelength is usually enough, but some convergence testing will validate this.
Again, whether the structure should extend through the PML depends on what you are trying to model and is not limited to the PML in the z direction (it also depend on the propagation direction).