Discussion Forum

Almost certainly yes.

Almost certainly yes.

But we find that the S-parameter is quite confusing when period length is near or lager than wavelength. There is no problem when the frequency is small that the period length is smaller than wavelength.

The frequency is 1-5GHz and the period size is 72mm.

With application of sufficient knowledge and care in creating the subject model as well as application of sufficient computational resources, you should be able to create an accurate model, with possible exceptions, such as if your model is overly-idealized (e.g., perfect conductors, no losses, which might yield infinite Q conditions), or otherwise demands very accurate computing of tiny differences between relatively-large numbers (thus losing precision). In contrast to Method of Moments (MoM), the Finite Elements (FE) formulation used is Comsol Multiphysics is (unless I am missing something) not based on any physics-level approximation to Maxwell's full classical electrodynamics, so it should work properly where non-classical EM physics can be ignored -- aside from any numerical errors caused by having only a finite number of these finite elements, as well as the digits of precision used in the calculations. You might want to consider posting your .mph file to the forum so that others here can offer suggestions. By the way, do you find that your results depend upon how fine your mesh is? Do your results depend on the discretization (linear, quadratic, cubic,...) of your mesh elements? Do your results depend on the volume used within (or distance to the boundaries of) your computational space?