In non-Hermitian photonic crystals, the spectrum of the wave operator forms closed loops in the complex plane rather than lying on the real line. When these loops have nontrivial topology — when they wind around points in the complex plane — edge modes appear that are exponentially localized at interfaces. This is the non-Hermitian skin effect.
Note (2026-11-12; read against the paper's full text): the paper requires broken spectral reciprocity in the continuous photonic crystals it studies (dispersion curves that form closed loops with a point gap), so the essay's contrast, that in continuous models the skin effect comes from the topology of the transfer-matrix eigenvalues rather than from non-reciprocity, is not the paper's. The skin-mode spectrum is the set of points enclosed by the dispersion curve with nonzero winding number, the same form as in the Toeplitz theory for discrete models; the paper also says Toeplitz-operator theory had already been developed for some continuous wave models, and that its new transfer-matrix invariant coincides with the classical winding number for simple bounded components.
The paper on spectral topology and edge modes (arXiv: 2603.22515) provides a rigorous mathematical framework for this effect in continuous wave models, where the discrete-lattice theory (based on Toeplitz matrices) breaks down.
The transfer matrix approach describes wave propagation through a one-dimensional periodic medium. The eigenvalues of the transfer matrix — which track how waves grow or decay across each unit cell — define a new topological invariant equivalent to the winding number of the non-Hermitian spectrum. This invariant counts how many times the spectrum wraps around the origin as the quasimomentum varies.
A nonzero winding number implies edge modes: waves trapped at the boundary because the bulk spectrum's topology prevents them from propagating. The mathematical theory works for continuous media — no discretization or tight-binding approximation needed.
The through-claim: the skin effect is topological, not just a lattice artifact. In discrete models, the skin effect comes from the non-reciprocity of hopping amplitudes. In continuous models, it comes from the topology of the transfer matrix's eigenvalues. The phenomenon survives the continuum limit because it's protected by a winding number, not by the lattice.
2603.22515. Photonics / non-Hermitian physics / skin effect / topological invariants / transfer matrices.