Convergence rates and explicit error bounds of Hill's method for spectra of self-adjoint differential operators (Q2452286)

The paper is focused on a numerical method for calculating spectra of the ordinary differential operator \[ S_{p} = \frac{d^{p}}{dx^{p}} + \sum_{j = 0}^{p - 1}f_{j}(x)\frac{d^{j}}{dx^{j}}, \] which is referenced to as the Hill's method. The method consists in reducing the initial eigenvalue problem \[ S_{p}\phi(x) = \lambda \phi(x) \] to a finite-dimensional eigenvalue problem \[ S_{p, N}^{\mu}\phi_{N} = \lambda_{N}\phi_{N}, \] where the operator \(S_{p, N}^{\mu}\) is a finite-dimensional projection of the operator \(e^{-i\mu x}S_{p}e^{i\mu x}\). Under some assumptions, among which are those about the functions \(f_{j}(x)\) and spectra properties of \(S_{p},\) the authors prove several statements about approximation properties of the operator \(S_{p, N}^{\mu}\) as related to \(S_{p}\). In particular, the authors find sufficient conditions for the eigenvalues of \(S_{p, N}^{\mu}\) to converge exponentially to those of \(S_{p},\) as \(N\) tends to \(+\infty,\) and provide explicit estimates for the convergence rate. The included numerical examples explain the meaning of the assumptions imposed and confirm theoretical conclusions derived in the paper.