Discussion Forum

Adarsha:

To answer your 2nd question first--no, you would not typically use PMC/PEC BCs in addition to PBCs on the same boundary. The PMC/PEC BCs are typically applied to take advantage of symmetry/anti-symmetry within your geometry while under the constraint of a symmetric/anti-symmetric impinging field. Because of this symmetry (say, k(x,y,z) = kz; Ei(x,y) = Ei,y), there is no reason to track the propagation/phase vector. The Floquet PBC does not require this constraint in the field, and can track the progression of the vector. If you were to place a PMC over the PBC, I would imagine you would lose this information.

As for the far-field determination, the region assessed has to be a closed geometry--either a closed loop (for 2D) or a closed surface (for 3D) to encompass the full 2-pi rad or 4-pi sr space. As it is a projection of your E-field solution--which already takes the periodic nature into account, I do not think the fact that a line or boundary is periodic affects the far-field calculation, but I am not 100% sure.

The two optical scattering library models are good examples everything you are asking here about BCs and far-field determination. Keep in mind the scattering problem is generally applied as per the "scatter on a substrate" library model--by first solving for the background field with the substrates but without the scatter using PBCs, THEN with the scatterer surrounded in a PML. This second physical model is where you get your real results, and should be assessing the far field. So, the far field calculation you are trying to accomplish would not even be using PBCs, would they?

Good luck!

-Ado

Hi Bryan,

I am simulating a unit cell of an infinite array using PBCs. For the far field calculation, the boundary should be a closed geometrical structure. But, in this case, since it is virtually infinite in xy plane, how could we consider a closed surface to define far field boundary? Surfaces perpendicular to the xy plane should not be considered for far field projection! Should I?


Adarsha

It's infinite, so the far-field flux is going to be normal to the surface, by symmetry, right? Basically if it's not normal to the surface than it must be at some angle, and since there's no way to decide whether it should be "left" or "right" in any cross-sectional plane, it must be normal. By definition "far field" means the detail of the individual cells is lost, so you can't talk about local flux direction: the EMR must be normal to the surface.

Yeah.. I agree..., which essentially means, I must have a far field boundary defined at both the sides of XY plane.. parallel to it. One to keep track of forward scattering and the other to backward. Is it??

That seems right (you definitely wouldn't want reflective or periodic boundary conditions, obviously!) You want to make sure the EMR leaves the simulation domain. But I've got no direct experience with this module so I'll defer to others.

Hey,

what is your solution finally