On the band gap structure of Hill's equation (Q2484266)

For Hill's equation it is known that solutions are linear combinations of functions having the form \(y(x,\lambda)=\exp(i 2\pi \theta(\lambda) x)p(x,\lambda)\), in which \(p\) is periodic in \(x\) and \(\exp(i 2\pi \theta(\lambda))\) is a Floquet multiplier. The parameter \(\theta(\lambda)\) is real-valued when \(\lambda\) lies in a stability interval and complex-valued when \(\lambda\) lies in an instability interval. Another way to formulate the stability problem is to consider \(\theta\) as a parameter in \((-1/2,1/2]\) and examine the family, parametrized by \(\theta\), of periodic eigenvalue problems for the function \(p\); the classical work of Bloch shows that the union of the ranges of the eigenvalues \(\lambda_m(\theta)\), \(m\in {\mathbb N}\), \(\theta\in (-1/2,1/2]\), is the same as the union of the stability intervals of the original equation. In this work, the authors show that formally replacing \(\theta\) by \(i\vartheta\), and allowing \(\vartheta\) to range over \({\mathbb R}\), one obtains nonselfadjoint eigenvalue problems for the functions \(p\) such that the union of the ranges of the real eigenvalues as \(\vartheta\) ranges over \({\mathbb R}\) yields precisely the instability intervals of the original problem.