Finiteness of lacunae in the spectrum of the Schrödinger operator and that of Dirac with periodic magnetic and electric potentials (Q1326739)

It is intuitively clear to a physicist that the spectrum of a Schrödinger operator with a periodic potential will have band structure with at most finitely many gaps. Mathematically the situation is as follows. If \(V: \mathbb{R}^ n\to \mathbb{R}\) is periodic with respect to a lattice \(\Gamma\) and \(D:= {1\over i}\nabla\), the Schrödinger operator \(H= D^ 2+ V\) in \(L^ 2(\mathbb{R}^ n)\) can be decomposed into a family, indexed by the so-called quasimomentum \(\theta\in \mathbb{R}^ n\), of self- adjoint operators \(H^ \theta_ V= (D- \theta)^ 2+ V\) in \(L^ 2(\mathbb{R}^ n/\Gamma)\) which have a purely discrete spectrum. Let \((\lambda_ k(\theta))\) \((k\in \mathbb{N})\) be a numbering of the eigenvalues of \(H^ \theta_ V\) in non-decreasing order, counting multiplicities. For \(k\in \mathbb{N}\), \(\lambda_ k(\cdot)\) is continuous and periodic with respect to the dual lattice of \(\Gamma\). As a consequence, the spectrum of \(H\) is the union of the closed intervals \(b_ k:= R(\lambda_ k(\cdot))\). For \(n= 1\) the \(b_ k\) are non-overlapping and in fact generically disjoint. For \(n\geq 2\) these intervals tend to overlap in a mode that is quite sensitive to the dimension. In the early eighties \textit{M. M. Skriganov} [see, e.g. Invent. Math. 80, 107-121 (1985; Zbl 0578.47003) and the literature cited there]\ showed for arbitrary lattices in dimensions \(n= 2,3\) and for rational \(\Gamma\) if \(n\geq 4\) that at most finitely many \(b_ k\) are disjoint, thus proving a long-standing conjecture of Bethe and Sommerfeld. The restriction on \(\Gamma\) in higher dimensions was necessitated by intricate number-theoretic considerations when counting the number of points of \(\Gamma\) in spheres of varying centres \(\theta\). The aim of the present paper is to prove the Bethe-Sommerfeld conjecture for arbitrary lattices in arbitrary dimensions. In fact the authors also consider periodic vector potentials \(c: \mathbb{R}^ n\to \mathbb{R}^ n\) in which case \(D-c\) replaces \(D\). (In a final section the \(n\)-dimensional Dirac operator is treated.) To this end they directly estimate the minimal and maximal number of eigenvalues of \(H^ \theta_ V\) that are less than a given number \(\lambda^ 2\) when \(\theta\) varies in \(\mathbb{R}^ n\). This is done by applying the method of a famous paper of \textit{L. Hörmander's} [Acta Math. 121, 193-218 (1968; Zbl 0164.132)] to the Fourier integral operators \(\exp(\text{it}| H^ \theta_ V|^{1/2})\) \((t\in \mathbb{R})\). While the reviewer does not wish to cast any doubt on the validity of the authors' results, he must confess that he is unable to keep track of the dependence or independence of the multitude of terms on the parameters involved. \{Reviewer's remark: Bethe and Sommerfeld were guided by perturbation theory, and such an approach to the problem was advocated by \textit{O. A. Veliev} [Sov. Math., Dokl. 27, 234-237 (1983; Zbl 0541.47017)]. This approach has, of course, its own difficulties [see \textit{Yu. E. Karpeshina}, Math. USSR, Sb. 71, 101-123 (1992; Zbl 0719.35061)], and the status of Veliev's proof in J. Sov. Math. 49, 1045-1058 (1990; Zbl 0664.47005) does not seem to be entirely undisputed. It would be desirable to have a detailed account of this deep and fascinating topic that is accessible to analysts outside the very small circle of specialists\}.