Discussion Forum

I'll offer some preliminary suggestions. First, this is easiest if the incident wave direction is normal to the plate. (Otherwise, you may need to set up Floquet boundary conditions, and that is trickier.) Let's assume normal incidence for now and assume a face of the plate lies in the x-y plane and thus the incident wave direction (the k vector) is along z. Construct a "unit cell" of the plate in the x-y plane, including any holes. This is, by definition, a subset of the plate that, if repeated infinitely (i.e., "tiled") in both the x and y directions would yield the overall plate. Construct a rectangular box (for the computation) that has x and y dimensions equal to the x and y dimensions of the unit cell, and z dimensions that extend moderately away from both faces of the plate (e.g., by perhaps lambda/2 for the longest wavelength of interest). Subtract (a Boolean "difference" operation) the plate (which should have a finite thickness) from the rectangular box. This will lead the plate parts to automatically be treated as PEC boundary conditions. Next, let's consider the case with the incident polarization of the wave in the y direction (for example). Thus, the two sides of the rectangular box that are parallel to the y-z plane should be set to perfect magnetic conductor (PMC) boundary conditions, while other two sides of the rectangular box (parallel to the x-z plane) should be set to perfect electric conductor (PEC) conditions.

Now, there are some options for what to do next. You can either try to launch your own plane wave from (for example) the end at –z heading toward +z, or you can employ the scattering-wave formulation (by telling it to solve for the scattered field). If we do the latter, we can simply set the two remaining faces of the rectangular box to scattering boundary conditions. In the EM box (“interface identifier”) box to specify the background electric field, you can specify the background plane wave, Eb. Set the y component equal to (for example): exp(-z*2.0*pi*c_const/freq). That gives you a +z directed background plane wave, with the right kz for each frequency (freq), over which you will be stepping in your frequency-domain study. Now, if doing it that way, Comsol's s-parameters are not explicitly available. But you can still define and compute integrals of the power-flow/area into and out-of the two end faces (the faces parallel to the x-y plane), to determine how much power is reflected and how much is transmitted, as a function of frequency.

And, of course, you will have to tell it the material (air) and mesh it all appropriately, set up your results parameters to plot nicely, etc. Good luck.

Robert:

Thank you for your suggestions. I gave them a shot and have attached my attempt. Is this about what you meant for the model? I am not sure that I meshed correctly....I just picked physics controlled....kept warning edge was much smaller than mesh...until used element size extremely fine.

Also it was difficult to give the hole a material. Every time I attempted to select the hole it only selected a small portion (seems it was broken into quadrants depth wise). I tried to plot the Ey and seems to increase with frequency as one would expect with an aperture/perforation but not sure the magnitude is correct yet.

I also just tried to assign the ends of the lambda/2 box but not sure if this is appropriate. I will look into how to properly set this up.

Any other info is greatly appreciated!