parity at BZ edge

[acerjan]

I can't seem to properly set the parity of a simulation along the edge of the Brillouin zone. For example, if I have a cube-symmetric photonic crystal, we know that along the high-symmetry plane (kx, pi/a, kz), that the eigenstates of the system should be even and odd with respect to the y-axis. However, when I call compute_yparities or display_yparities, the numbers that are returned are not +/- 1. I can plot the fields, compute the parity by hand, and confirm that they are even or odd, but I'd prefer to simply use run_parity to guarantee this symmetry of the system, as this avoids problems when kz is near 0, and some of the field components for each mode vanish.

It seems that the code is checking my choice of k and confirming that whatever parity I've tried to select is appropriate, but I don't understand why these choices are being overridden. Is there something I'm missing? Is there a way to manually force a certain parity? Or, if the fix in the code for this is quick, can you point me to where I should look? I can't quite figure out what's going on in maxwell_constraints.c.

[stevengj]

I think that the issue is that the parity is computed for the Bloch envelope (https://github.com/NanoComp/mpb/blob/master/src/maxwell/maxwell_constraints.c#L254-L312) not including the exp(ikx) factor, which does not have mirror symmetry.

The reason is probably that when I implemented this I was mainly thinking of the eigensolver constraint, which needs to apply to the Bloch envelope and hence I think only works at k=0, and the parity computation was added as an afterthought.