Non-Hermitian quantum mechanics

You have to excite the edge mode selectively through a port in an open system that's leaking the whole time.

Which means, and this is the honest framing, a photonic topological device is non-Hermitian from the start.

Treating gain and loss as a small perturbation on some Hermitian ideal was always a convenient fiction.

Useful fiction, but fiction.

Okay, now this is where it gets genuinely interesting because the field's response to all of this was not to retreat.

It was to lean in.

If the device is non-Hermitian anyway, stop fighting it.

Drop the fiction and write down the Hamiltonian the device actually has.