Spectral theory of Schrödinger operators with potentials that are measures supported on ℕ {\mathbb{N}}
Oleg Safronov · 2025
Read Spectral theory of Schrödinger operators with potentials that are measures supported on ℕ {\mathbb{N}} by Oleg Safronov (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-20320.txt.
Introduction
A quantum particle in a landscape of sharp spikes is the object of Oleg Safronov's 2025 arXiv paper in spectral theory. The potential is a sum of point masses sitting on the positive integers. Such Schrodinger operators model electrons in certain disordered wires. Safronov asks when their spectrum contains a continuous band.
His main result gives conditions under which the absolutely continuous spectrum covers a whole interval. The conditions involve the decay of the spike strengths and the behavior of a related Fourier series. He also proves the result is sharp. For potentials decaying too slowly, he constructs examples whose spectrum is purely singular, with no continuous band at all. The paper is short, technical, and closes a precise borderline question in the theory of one-dimensional operators with measure potentials.
Catalog record
| CREATOR | Oleg Safronov |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 69 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-20320
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Spectral theory of Schrödinger operators with potentials that are measures supported on ℕ {\mathbb{N}} Oleg Safronov Abstract . We discuss spectral properties of the one-dimensional Schrödinger operator with a potential of the form ∑ V ( n ) δ ( x − n ) \sum V(n)\delta(x-n) . Our main result says that the absolutely continuous spectum of such an operator covers an interval [ α 2 , β 2 ] [\alpha^{2},\beta^{2}] , if V ∈ ℓ 4 V\in\ell^{4} and the Fourier series ∑ e 2 i k n V ( n ) \sum e^{2ikn}V(n) is a function of k k that is square integrable over [ α , β ] [\alpha,\beta] . We prove that this result is sharp by constructing examples of potentials V ∉ ℓ 2 V\notin\ell^{2} for which the spectrum of the Schrödinger operator is singular. 1. Introduction In this paper, we will look at the spectral theory of very special Schrödinger operators on the positive half-line {\mathbb{R}}_{+}=[0,\infty) with the Dirichlet boundary condition at x=0 . Namely, let V:{\mathbb{N}}\to{\mathbb{R}} be a bounded real-valued function. We define H to be the operator whose quadratic form is \int_{0}^{\infty}|u^{\prime}|^{2}dx+\sum_{n=1}^{\infty}V(n)|u(n)|^{2},\qquad u\in W_{0}^{1,2}({\mathbb{R}}_{+}). Let \delta be the standard delta function on {\mathbb{R}} . Then H can be formally understood as the operator H=-\frac{d^{2}}{dx^{2}}+\sum_{n=1}^{\infty}V(n)\delta(x-n). If {\rm Im}\,k>0 , then there is a unique solution of the equation -\psi^{\prime\prime}+\sum_{n=1}^{\infty}V(n)\delta(x-n)\psi=k^{2}\psi,\qquad\text{Im}\,k>0, (1.1) that decays exponentially as x\to\infty . We denote m(k^{2})=\psi^{\prime}(0)/\psi(0) . The spectral measure associated to H is the measure \mu on {\mathbb{R}} having the property \int_{\mathbb{R}}\frac{1}{1+t^{2}}d\mu(t)<\infty, and uniquely defined by m(E)=A+\int_{\mathbb{R}}\Bigl[\frac{1}{t-E}-\frac{t}{1+t^{2}}\Bigr]d\mu(t),\quad E\notin{\rm supp}\,\mu, (1.2) with A\in{\mathbb{R}} . We want to find conditions on V guaranteeing that the absolutely continuous spectrum of H is essentially supported on the positive half-line {\mathbb{R}}_{+}=[0,\infty) . That is, \mu^{\prime}(t)>0 for almost every t>0 . One of such conditions is formulated in terms of the Fourier series \hat{V}(k)=\sum_{n=1}^{\infty}e^{ikn}V(n) whose convergence is understood in the sense of distributions. Theorem 1.1 . Let the interval [\alpha,\beta]\subset{\mathbb{R}}_{+} be free of integer multiples of \pi . Assume that \sum_{n=1}^{\infty}|V(n)|^{4}<\infty,\qquad\text{and}\quad\int_{\alpha}^{\beta}\bigl|\hat{V}(2k)\bigr|^{2}dk<\infty. (1.3) Then the following statements are true: (0) The spectrum of H coincides with the support of \mu which is [0,\infty)\cup\{E_{j}\}_{j=1}^{N} where N is zero , finite , or infinite , and E_{1}<E_{2}<\cdots<0 . If N is infinite , then \lim_{j\to\infty}E_{j}=0 . (1) Let \mu_{\rm ac}(E)=f(E)\,dE where \mu_{\rm ac} is the Lebesgue absolutely continuous component of \mu . Then \int_{a}^{b}\log[f(E)]\,dE>-\infty, (1.4) for any interval [a,b]\subset(\alpha^{2},\beta^{2}) . Remark. Condition (0) is just a quantitative way of writing that the essential spectrum of H is the same as that of H_{0}=-\frac{d^{2}}{dx^{2}} . Theorem 1.2 . Assume that \sum_{n=1}^{\infty}|V(n)|<\infty. (1.5) Then \sum_{j=1}^{N}|E_{j}|^{1/2}\leq\frac{1}{2}\sum_{n=1}^{\infty}|V(n)|. (1.6) The next statement deals with the sums of higher powers of eigenvalues. Theorem 1.3 . Let V\in\ell^{p+1/2}({\mathbb{N}}) for p>1/2 . Let E_{j} be the negative eigenvalues of the operator H . Assume that \|V\|_{\infty}<2 . Then \sum_{j=1}^{N}|E_{j}|^{p}\leq C_{p}\sum_{n=1}^{\infty}|V(n)|^{p+1/2}, (1.7) where C_{p}=\frac{\sqrt{2}\int_{0}^{4}(1-\gamma/4)\gamma^{p-3/2}d\gamma}{\int_{0}^{1}(1-\gamma)^{1/2}\gamma^{p-3/2}d\gamma}. Remark. Inequalities ( 1.6 ), ( 1.7 ) are analogues of the celebrated bounds of Lieb and Thirring [ 17 ] , [ 18 ] for Schrödinger operators with potentials V that are functions on {\mathbb{R}}_{+} . Inequality ( 1.6 ) could be also viewed as an analogue of the bound established by Hundertmark, Lieb, and Thomas in [ 8 ] . By property (1), for any V satisfying the condition ( 1.3 ), the essential support of the a.c. spectrum of H contains the interval [a,b] . That is, \mu_{\rm ac} gives positive weight to any subset of [a,b] of positive Lebesgue measure. This follows from ( 1.4 ) because f cannot vanish on any such set. This observation may be compared with well-known results involving “usual” Schrödinger operators whose potentials are functions V\in L^{p}({\mathbb{R}}_{+}) . It is known that such operators have a.c. spectrum if p=2 (see Deift-Killip [ 3 ] ), but the a.c. spectrum can disappear once p>2 . While the assumption V\in L^{p}({\mathbb{R}}_{+}) with p>2 does not, by itself, guarantee the presence of absolutely continuous spectrum, there are additional conditions that can be imposed to ensure its existence (see [ 10 ] and [ 19 ] ). One should also mention that many properties and results regarding Schrödinger operators have analogous statements or extensions for Jacobi matrices (see, for instance, [ 11 ] and [ 16 ] ). In this sense, one can find the comparison between ( 1.3 ) and the conditions imposed on the perturbation of the free Jacobi matrix in [ 16 ] particulary interesting: any V\in\ell^{4} obeying \sum_{n=1}^{\infty}\bigl(V(n)+V(n+1)\bigr)^{2}<\infty, satisfies assumptions of Theorem 1.1 . However, the closest to Theorem 1.1 is the result of R. Killip [ 10 ] , involving the Fourier transform of the “usual” potential V\in L^{3}({\mathbb{R}}_{+}) (for Jacobi matrices, see also [ 22 ] ). While their final outcomes are similar, the proofs of Theorem 1.1 and Killip’s theorem [ 10 ] differ in two critical ways. First of all, in our case, \hat{V} is the sum of the Fourier series which is a periodic function of k that does not decay at infinity, even if V has a finite support. Secondly, in the proof, one needs to link the spectral properties of H to the properties of some Jacobi matrices. This connection forces us to consider intervals [\alpha,\beta] that are free of integer multiples of \pi . Also note that, once V satisfies the condition ( 1.3 ), it will also satisfy ( 1.3 ) with \alpha and \beta replaced by \alpha+\pi n and \beta+\pi n . Therefore, the a.c. spectrum of H will cover all intervals [(\alpha+\pi n)^{2},(\beta+\pi n)^{2}] with n\in{\mathbb{N}} . The operator H could be viewed as an operator on quantum graph whose vertices are points in {\mathbb{N}}\cup\{0\} and the edges are the unit intervals connecting these verices. It is known that periodic operators defined on such graphs might have infinitely many gaps in their spectra. In particular, for any a\neq 0 , the spectrum of the operator -\frac{d^{2}}{dx^{2}}+a\sum_{n=1}^{\infty}\delta(x-n) (1.8) has infinitely many gaps situated either to the left or to the right of the points (\pi n)^{2} . Even though the absolutely continuous spectrum is no longer {\mathbb{R}}_{+} , it is still natural to ask, what conditions guarantee that the absolutely continuous spectrum of the operator H_{a}=-\frac{d^{2}}{dx^{2}}+a\sum_{n=1}^{\infty}\delta(x-n)+\sum_{n=1}^{\infty}V(n)\delta(x-n) is the same as that of the operator ( 1.8 ). Consider for simplicity the case a>0 . Suppose V obeys the same conditions as before: \sum_{n=1}^{\infty}|V(n)|^{4}<\infty,\qquad\text{and}\quad\int_{\alpha}^{\beta}\bigl|\hat{V}(2k)\bigr|^{2}dk<\infty. (1.9) where the interval [\alpha,\beta]\subset{\mathbb{R}}_{+} is free of integer multiples of \pi . What is the essential support of the absolutely continuous spectrum of the operator H_{a} ? To answer this question, we define the open set K_{\alpha,\beta}=\bigl\{\lambda>0\mid\,\,\,\frac{a\sin(\sqrt{\lambda})}{\sqrt{\lambda}}+2\cos(\sqrt{\lambda})\in S_{\alpha,\beta}\bigr\}, where S_{\alpha,\beta}=\bigl\{\lambda\in{\mathbb{R}}\mid\,\,\,\lambda=2\cos k,\quad k\in(\alpha,\beta)\bigr\}. Theorem 1.4 . Let a>0 and let the interval [\alpha,\beta]\subset{\mathbb{R}}_{+} be free of integer multiples of \pi . Assume that V satisfies conditions ( 1.9 ). Let \mu_{\rm ac}(E)=f(E)\,dE where \mu_{\rm ac} is the Lebesgue absolutely continuous component of the spectral measure \mu of the operator H_{a} . Then \int_{\alpha_{0}}^{\beta_{0}}\log[f(E)]\,dE>-\infty, (1.10) for any interval [\alpha_{0},\beta_{0}]\subset K_{\alpha,\beta} . This theorem holds for a<0 with the set K_{\alpha,\beta} replaced by the union K_{\alpha,\beta}\cup\bigl\{\lambda<0\mid\,\,\,\frac{a\sinh(\sqrt{-\lambda})}{\sqrt{-\lambda}}+2\cosh(\sqrt{-\lambda})\in S_{\alpha,\beta}\bigr\}. The necessity of considering this union is related to the fact that the spectrum of ( 1.8 ) with