Wave operators for Jacobi matrices

Sergey A. Denisov, Giorgio Young · 2026

Read Wave operators for Jacobi matrices by Sergey A. Denisov, Giorgio Young (2026) 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-2604-18442.txt.

Introduction

Denisov and Young study wave operators of a Jacobi matrix whose spectral measure meets the Szegö rule. They prove they exist and are complete under a mild extra proviso on the Verblunsky coefficients. Those numbers belong to the linked measure on the unit circle, tying the spectral data to them.

Catalog record

CREATORSergey A. Denisov, Giorgio Young
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE75707
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.18442

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Wave operators for Jacobi matrices Thanks: The research of S.D. was supported by the grant NSF-DMS-2450716, Simons Fellowship in Mathematics, Simons Travel Support for Mathematicians Award, and by the Van Vleck Professorship Research Award. He gratefully acknowledges the hospitality of IHES where part of this work was done. Thanks: G.Y. acknowledges the support of the National Science Foundation through grant DMS–2303363. Sergey A. Denisov, Giorgio Young Abstract . We study the wave operators for a Jacobi matrix J J whose canonical spectral measure ρ \rho satisfies the Szegő condition. In particular, for γ s \gamma_{s} the Verblunsky coefficients of the measure associated to ρ \rho by the Szegő mapping, we prove existence and completeness of wave operators under the condition lim n → ∞ log ⁡ ( n ) ​ ∑ s = n 2 ​ n | γ s | 2 = 0 \lim_{n\to\infty}\log(n)\sum_{s=n}^{2n}|\gamma_{s}|^{2}=0 . In the course of proving our main theorem, we establish an estimate on the norm of orthogonal polynomials on the unit circle localized to arcs that may be of independent interest. 1. Introduction We consider semi-infinite Jacobi matrices \displaystyle J=\begin{pmatrix}v_{1}&b_{1}&0&&&&\\ b_{1}&v_{2}&b_{2}&0&&&\\ 0&b_{2}&v_{3}&b_{3}&0&&\\ &\ddots&\ddots&\ddots&\ddots&\ddots\\ \end{pmatrix} in the bounded and self-adjoint setting, where the Jacobi coefficients \{b_{n}\} and \{v_{n}\} satisfy b_{n}>0,v_{n}\in\mathbb{R} for all n\in\mathbb{N} , and \{b_{n}\},\{v_{n}\}\in\ell^{\infty}(\mathbb{N}) . Under these conditions, J defines a bounded self-adjoint operator J:\ell^{2}(\mathbb{N})\to\ell^{2}(\mathbb{N}) . Through the special case b_{n}=1/2 for all n\in\mathbb{N} , one recovers discrete Schrödinger operators, which serve as Hamiltonians for lattice models in quantum mechanics. Jacobi matrices arise as a natural generalization of these operators, particularly from the perspective of inverse spectral theory discussed below. Our main interest in this work is the dynamics generated by J through the discrete time-dependent Schrödinger equation: \displaystyle-i\partial_{t}\psi=J\psi,\quad\psi(0)=f\in\ell^{2}(\mathbb{N}), whose solution may be written in terms of the one-parameter unitary group as \psi(t)=e^{itJ}f . By Favard’s theorem of classical analysis, there is a bijective correspondence between bounded semi-infinite Jacobi matrices J and compactly and infinitely supported (in the sense of cardinality) probability measures on \mathbb{R} . In the forward direction of this correspondence, one maps J to its canonical spectral measure, \rho . In this work, we consider Jacobi matrices whose corresponding measures \rho are supported on [-1,1] , and are in the Szegő class: given the decomposition of \rho with respect to the equilibrium measure \omega for the interval [-1,1] , \displaystyle d\rho=\rho^{\prime}d\omega+\rho_{s},\,\,d\omega(x)=\pi^{-1}(1-x^{2})^{-\frac{1}{2}}\mathbb{1}_{[-1,1]}(x)dx, we assume that \rho satisfies the Szegő condition: (1.1) \displaystyle\int\limits_{\mathbb{R}}\log\rho^{\prime}d\omega>-\infty\,. Despite the inverse direction of this correspondence passing through the orthogonal polynomials on the real line (OPRL), much of our work here makes use of orthogonal polynomials on the unit circle (OPUC). In particular, exploiting a well-known relationship between OPRL with respect to measures supported on the interval [-1,1] , and OPUC with respect to a related measure, we may work on the circle and apply harmonic analysis techniques. Thus, our first result is most naturally stated in the OPUC terminology, and we introduce it now. We refer the reader to [ 23 , 24 , 25 , 26 ] for further background on the brief description that follows. Let \sigma be an infinitely supported probability measure on the unit circle \mathbb{T} . Denote the orthonormal polynomials by \phi_{n}(z) and the monic orthogonal polynomials by \Phi_{n}(z) . Clearly, \phi_{n}=\Phi_{n}/\|\Phi_{n}\|_{2,\sigma} . We will use notation f^{*}(z)=z^{n}\overline{f(\bar{z}^{-1})} for a polynomial f of degree n . The polynomials \{\Phi_{n}\} satisfy recurrence [ 23 , Theorem 1.5.2, p. 57] (1.2) \left\{\begin{array}[]{cc}\Phi_{n+1}(z)=z\Phi_{n}(z)-\bar{\gamma}_{n}\Phi_{n}^{*}(z),&\Phi_{0}(z)=1,\\ \Phi_{n+1}^{*}(z)=\Phi_{n}^{*}(z)-z\gamma_{n}\Phi_{n}(z),&\Phi_{0}^{*}(z)=1\,,\end{array}\right. where \gamma_{n} are the Verblunsky coefficients corresponding to the measure \sigma . The map \sigma\mapsto\{\gamma_{n}\}|_{n=0}^{\infty}\in\mathbb{D}^{\infty} is a bijection, cf. [ 23 , p. 2] . Measures \rho supported on [-1,1] may be related to even measures \sigma on \mathbb{T} through the Szegő mapping Sz . We will write \sigma=Sz^{-1}(\rho) for the measure \sigma on \mathbb{T} that satisfies d\sigma(z)=d\sigma(1/z) and (1.3) \displaystyle\int\limits_{\mathbb{T}}f(z)d\sigma(z)=\int\limits_{[-1,1]}f(e^{i\arccos(x)})d\rho(x) for any f\in C(\mathbb{T}) with f(z)=f(1/z),z\in\mathbb{T} [ 25 , p. 31, eq. (1.9.5)] . Throughout the text, \sigma will refer to the measure related to \rho in this way. For such measures, all \gamma_{n} are real, and they may be related to the Jacobi coefficients by means of the direct Geronimus relations [ 25 , p. 33] (1.4) \displaystyle\begin{split}b_{k+1}&=\frac{1}{2}\left((1-\gamma_{2k-1})(1-\gamma_{2k}^{2})(1+\gamma_{2k+1})\right)^{\frac{1}{2}},\\ v_{k+1}&=-\frac{1}{2}(\gamma_{2k-2}(1+\gamma_{2k-1})-\gamma_{2k}(1-\gamma_{2k-1}))\,,\end{split} with the convention \gamma_{-1}:=-1 . If we denote the polynomials orthonormal with respect to \rho by p_{n}(x) , then [ 25 , p. 32, eq. (1.9.12)] (1.5) \displaystyle p_{n}((z+1/z)/2)=(2(1-{\gamma}_{2n-1}))^{-1/2}z^{-n}(\phi_{2n}(z)+\phi_{2n}^{*}(z)),\,z\in\mathbb{T}\,. Szegő’s theorem, cf. [ 25 , Theorem 1.8.6] , is a cornerstone of the theory of OPUC; for the decomposition of \sigma with respect to the normalized Lebesgue measure dm(z) on \mathbb{T} , d\sigma(z)=\sigma^{\prime}(z)dm(z)+\sigma_{s}(z) , it states \displaystyle\prod_{n=0}^{\infty}(1-|\gamma_{n}|^{2})=\exp\left(\int\limits_{\mathbb{T}}\log(\sigma^{\prime}(z))dm(z)\right)\,, which in turn implies \displaystyle\{\gamma_{n}\}\in\ell^{2}(\mathbb{Z}^{+})\iff\int\limits_{\mathbb{T}}\log(\sigma^{\prime}(z))dm(z)>-\infty. With this in hand, the identity ( 1.3 ) implies that the Szegő condition ( 1.1 ) on J is equivalent to the corresponding Verblunsky parameters satisfying \{\gamma_{n}\}\in\ell^{2}(\mathbb{Z}^{+}) . Denoting by J_{0} the free operator with Jacobi coefficients b_{n}=\frac{1}{2},v_{n}=0 for all n , we may combine this observation with the Geronimus relations to find that the condition ( 1.1 ) implies that J-J_{0} is a Hilbert-Schmidt operator. Thus, in addition to being a qualitative condition on the size and distribution of the absolutely continuous part of the measure \rho , this condition allows for J to be viewed as a perturbation of the free operator. Both of these perspectives offer a first motivation for the study of the evolution generated by J , particularly in comparison to the free evolution J_{0} . Motivated by classical results of [ 2 , 20 ] and already given the condition ( 1.1 ), one may expect the typical state to spread out over \mathbb{N} , behaving as in the free evolution in a qualitative sense, meanwhile the perturbative perspective suggests a stronger and more direct comparison may hold between the two evolutions. Such a comparison is often encoded through the existence and completeness of wave operators \Omega_{\pm} , a by now well-studied and classical topic. We refer the reader to the foundational scattering theory papers [ 1 , 5 , 11 ] , as well as the textbook treatments of the subject found in [ 14 , 17 , 22 , 27 ] . When they exist, wave operators are defined as the strong \ell^{2}(\mathbb{N}) -limits (1.6) \displaystyle\lim_{T\to\pm\infty}e^{-iTJ}e^{iTJ_{0}}=:\Omega_{\pm}. We denote by \ell^{2}_{\rm ac}(\mathbb{N}) the absolutely continuous subspace for J , or set of vectors whose spectral measure is absolutely continuous, cf. [ 18 , 21 ] : \displaystyle\ell^{2}_{\rm ac}(\mathbb{N}):=\{\psi\in\ell^{2}(\mathbb{N}):\mu_{\psi}\text{ is a.c.}\}, where \mu_{\psi} is spectral measure of \psi relative to J . When \Omega_{\pm} exist, the inclusion \text{Ran}(\Omega_{\pm})\subseteq\ell^{2}_{\rm ac}(\mathbb{N}) always holds. Completeness is defined as the equality \text{Ran}(\Omega_{\pm})=\ell^{2}_{\rm ac}(\mathbb{N}) , so that \Omega_{\pm} are unitary maps from \ell^{2}(\mathbb{N}) onto \ell^{2}_{\rm ac}(\mathbb{N}) . This equality yields the comparison between the two evolutions alluded to above; indeed, once this is established, one has that every state in \ell^{2}_{\rm ac}(\mathbb{N}) behaves asymptotically freely. Establishing these results under ( 1.1 ), along with a mild quantitative condition on the tail of the series for \{\gamma_{n}\} , will be the main goal of our work here. For the reader more familiar with the OPUC literature, we note that this is a significantly more delicate task than existence and completeness of the CMV wave operators, where the free dynamics are given by translation, cf. [ 24 ,

Read the complete text