I have in the past worked on them. So, what's the real problem?
An FSS is part of a radome near an antenna that rejects certain frequencies (it can be put in the antenna too, but carefully), it's really just a filter implemented in the antenna, usually it's rolloff's are weak compared to normal circuit based filters. You use them to cut out interference on a budget, or for high power transmissions where discrete devices would blow up easily.
kchiggins.

Thank you, Higgler,
You're right, those are some of the applications for the FSS's. What I am doing is a code to analyse them and when the element is a square or a symmetrical cross, and the unit cell is square, I should get the same reflection coefficient for both TE and TM polarization, but I don’t, and it seems that my problem is in the TE polarization and it worsen when I am approaching to its cut-off frequency.
Any idea?
Thank you very much.

On boresight it should be the same, off angle it's very different, even though you're using square pieces. Maybe you have simple approximations that lose their luster at cutoff.

as an example off boresight, the impedance of a thick slab of dielectric is n/cos(theta) versus n*cos(theta)for the two different cases, and the thick slab is homogeneous.
Maybe you are doing it correctly.
kch
There are reference materials on them.

Thank you for your reply Higgler.
I'm just working on boresight by the moment.
Maybe, there's something with the condition number of the matrix. I've seen that the variations in this number along the frequency have influence in the final TE reflection result.  I'll keep working on it.

Thank you again

Your comments are interesting.
Have you found the edge on performance of any square tic to be direction insensitive?