Hardy Operators: In the footsteps of Brown, Halmos, and Shields
Eva A. Gallardo-Gutiérrez, Jonathan R. Partington, William T. Ross · 2025
Read Hardy Operators: In the footsteps of Brown, Halmos, and Shields by Eva A. Gallardo-Gutiérrez, Jonathan R. Partington, William T. Ross (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-2501-14706.txt.
Introduction
Eva Gallardo-Gutiérrez, Jonathan Partington, and William Ross revisit two classic Hardy operators in a 2025 mathematics paper. These averaging operators, first studied by Brown, Halmos, and Shields, act on functions over a finite interval and over the half-line. They are simple to state yet rich in structure, a combination that has drawn analysts for decades.
The authors build explicit cyclic vectors, starting points from which the whole space can be generated by repeated application. They also characterize the subspaces the operators leave unchanged. Frame vectors get a full description as well. Such complete maps of an operator's behavior are rare and useful: Hardy operators model averaging processes across analysis and probability.
Catalog record
| CREATOR | Eva A. Gallardo-Gutiérrez, Jonathan R. Partington, William T. Ross |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 61 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-14706
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Hardy Operators: In the footsteps of Brown, Halmos, and Shields Thanks: First two authors are partially supported by Plan Nacional I+D grant no. PID2022-137294NB-I00, Spain. First author also acknowledges support from the Spanish Ministry of Science and Innovation, through the “Severo Ochoa Programme for Centres of Excellence in R&D” (CEX2019-000904-S & CEX2023- 001347) and from the Spanish National Research Council, through the “Ayuda extraordinaria a Centros de Excelencia Severo Ochoa” (20205CEX001). Eva A. Gallardo-Gutiérrez, Jonathan R. Partington, William T. Ross Abstract . This paper discusses the two classical Hardy operators ℋ 1 \mathcal{H}_{1} on L 2 ( 0 , 1 ) L^{2}(0,1) and ℋ ∞ \mathcal{H}_{\infty} on L 2 ( 0 , ∞ ) L^{2}(0,\infty) initially studied by Brown, Halmos and Shields. Particular emphasis is given to the construction of explicit cyclic and ∗ * -cyclic vectors in conjunction with a characterization of their invariant and reducing subspaces. We also provide a complete description of the frame vectors for I − ℋ 1 ∗ I-\mathcal{H}_{1}^{*} . 2010 Mathematics Subject Classification Primary 47A15, 47A16, 47B15, 30H10 1. Introduction This paper explores the invariant subspaces as well as the cyclic and frame vectors for the two classical Hardy integral operators. The first, known as the finite Hardy operator \mathcal{H}_{1} , is defined on the Lebesgue space L^{2}(0,1) by the formula (\mathcal{H}_{1}f)(x)=\frac{1}{x}\int_{0}^{x}f(t)dt,\quad 0<x<1, while the second, known as the infinite Hardy operator \mathcal{H}_{\infty} , is defined on L^{2}(0,\infty) by (\mathcal{H}_{\infty}g)(x)=\frac{1}{x}\int_{0}^{x}g(t)dt,\quad 0<x<\infty. Although the formulas for these two operators are the same and have similar looking adjoint formulas (\mathcal{H}_{1}^{*}f)(x)=\int_{x}^{1}\frac{f(t)}{t}dt,\quad 0<x<1, (\mathcal{H}^{*}_{\infty}g)(x)=\int_{x}^{\infty}\frac{g(t)}{t}dt,\quad 0<x<\infty, and even have the same operator norm equal to 2 (stemming from Hardy’s integral inequality), they act boundedly on different Hilbert spaces and have distinct properties. In 1965, Brown, Halmos, and Shields [ 4 ] proved that the spectrum of \mathcal{H}_{1} is the closed disk \{z:|z-1|\leqslant 1\} while the spectrum of \mathcal{H}_{\infty} is its boundary \{z:|z-1|=1\}. A more salient difference between \mathcal{H}_{1} and \mathcal{H}_{\infty} , also from [ 4 ] , comes from their representations as shifts, namely I-\mathcal{H}_{1}^{*} is unitarily equivalent to the unilateral shift (Sf)(z)=zf(z) on the classical Hardy space of the open unit disk \mathbb{D}:=\{z:|z|<1\} (and denoted by H^{2} ), while I-\mathcal{H}_{\infty}^{*} is unitarily equivalent to the bilateral shift (Mf)(\xi)=\xi f(\xi) on the Lebesgue space L^{2}(\mathbb{T},m) , where \mathbb{T}:=\{z:|z|=1\} is the unit circle and dm=d\theta/(2\pi) is normalized Lebesgue measure on \mathbb{T} . As a straightforward consequence, \mathcal{H}_{1}^{*} is a subnormal operator, while \mathcal{H}_{\infty} is normal. Accordingly, the invariant subspace structure for the operators \mathcal{H}_{1} and \mathcal{H}_{\infty} is essentially understood since, in the first case, the invariant subspaces of \mathcal{H}_{1} correspond, via Beurling’s theorem on the invariant subspaces of the shift on H^{2} , to the model spaces (\Theta H^{2})^{\perp}:=H^{2}\ominus\Theta H^{2} (where \Theta is an inner function), while in the second, via results of Wiener and Helson on the invariant subspaces of the bilateral shift, the invariant subspaces of \mathcal{H}_{\infty} correspond to orthogonal complements of spaces of the form qH^{2} (where q is a unimodular function on \mathbb{T} ) or \chi_{E}L^{2}(\mathbb{T},m) (where E is a measurable subset of \mathbb{T} ). See [ 14 , Ch. II] for the details on this. More recently, Agler and McCarthy [ 1 ] have shown that the invariant subspaces of {\mathcal{H}}_{1} can also be expressed as limits of sequences of finite dimensional spaces spanned by eigenfunctions for \mathcal{H}_{1} , namely the monomial functions x\mapsto x^{s} for s\in\mathbb{C} with \mathop{\rm Re}\nolimits s>-\frac{1}{2} , as so–called “liminf spaces”. This is in contrast to the related Volterra operator f(x)\mapsto\int_{0}^{x}f(t)dt on L^{2}(0,1) , which has no finite dimensional invariant subspaces (see, for example, [ 12 , Chap. 7] ). Indeed, every invariant subspace of the Volterra operator takes the form \chi_{[a,1]}L^{2}(0,1) for some 0\leqslant a\leqslant 1 . The above descriptions of the invariant subspaces of \mathcal{H}_{1} and \mathcal{H}_{\infty} rely either on the unitary operators involved in identifying I-\mathcal{H}_{1}^{*} with S and I-\mathcal{H}_{\infty}^{*} with M or, in the somewhat implicit approach of Agler and McCarthy, as limits of functions via “liminf spaces”. This paper starts by giving an explicit description of the invariant subspaces in terms of the functions in the spaces L^{2}(0,1) or L^{2}(0,\infty) and not their realizations on H^{2} or L^{2}(\mathbb{T},m) . The ideas of Brown, Halmos and Shields, along with an explicit inversion formula for the Mellin transform (Proposition 2.6 ), allows us to characterize the reducing subspaces for \mathcal{H}_{\infty} (i.e., those invariant for both \mathcal{H}_{\infty} and \mathcal{H}_{\infty}^{*} ) as \left\{t\mapsto\int_{E}t^{-ix-\frac{1}{2}}f(ix)dx:f\in L^{2}(i\mathbb{R})\right\} for some measurable subset E\subseteq\mathbb{R} (see Theorem 4.2 ). Note that \mathcal{H}_{1} lacks reducing subspaces. This, in turn, leads us to a characterization in Theorem 4.3 of the \ast -cyclic vectors for \mathcal{H}_{\infty} , i.e., those f\in L^{2}(0,\infty) not contained in any nontrivial reducing subspace. Specifically, f\in L^{2}(0,\infty) is \ast -cyclic for \mathcal{H}_{\infty} if and only if the function \beta\mapsto\int_{-\infty}^{\infty}f(e^{x})e^{(i\beta+\frac{1}{2})x}dx does not vanish on any subset of \mathbb{R} with positive measure. Some particular classes (Proposition 4.7 ) of \ast -cyclic vectors include all nontrivial L^{2}(0,\infty) functions supported on [0,a] or supported on [a,\infty) , 0<a<\infty . The study of the cyclic vectors (vectors not contained in any nontrivial invariant subspace) for \mathcal{H}_{1}^{*} and \mathcal{H}_{\infty}^{*} is a complex undertaking as the provided criteria are difficult to verify. As particular instances, we show in § 5 that the functions t\mapsto t^{\alpha} , where \mathop{\rm Re}\nolimits\alpha>-\frac{1}{2} , are cyclic vectors for \mathcal{H}_{1}^{*} while the functions \frac{1}{(t+1)^{n}},\;n=1,2,\ldots,\;\;\frac{1}{\sqrt{t}(1+(\log t)^{2})},\;\;\mbox{and}\;e^{-t} are cyclic for both \mathcal{H}_{\infty} and \mathcal{H}^{*}_{\infty} . In fact, \mathcal{H}_{\infty} and \mathcal{H}_{\infty}^{*} share the same cyclic vectors (Proposition 5.6 ). This paper will also make connections to semigroups of composition operators on L^{2}(0,\infty) , the Laguerre shift, and, using our Mellin transform analysis, the commutants of \mathcal{H}_{1} and \mathcal{H}_{\infty} . These results will also help deminstrate the variety of interesting invariant subspaces for our two Hardy operators (see Examples 6.2 and 6.3 ). The final section of the paper follows the work by Cabrelli, Molter, Paternostro, and Philipp [ 5 ] and characterizes the frame vectors for a particular variation of \mathcal{H}_{1} . Here, for a bounded linear operator T on separable Hilbert space H , a vector {\bf v}\in H is a frame vector for T if the sequence (T^{n}{\bf v})_{n\geqslant 0} is a frame in that there are positive constants c_{1},c_{2} so that c_{1}\|{\bf x}\|^{2}\leqslant\sum_{n=0}^{\infty}|\langle T^{n}{\bf v},{\bf x}\rangle|^{2}\leqslant c_{2}\|{\bf x}\|^{2}\;\;\mbox{for all ${\bf x}\in H$}. In particular, [ 5 , Thm. 3.3] asserts that for T to have frame vectors at all, is it necessary that T^{*} is similar to a strongly stable contraction, in other words, \lim_{n\to\infty}\|T^{*n}{\bf y}\|=0\;\;\mbox{ for all ${\bf y}\in H$.} By the uniform boundedness principle, it follows that such T must satisfy \sup_{n\geqslant 0}\|T^{n}\|<\infty and hence, in particular, \sigma(T)\subseteq\overline{\mathbb{D}}:=\{z:|z|\leqslant 1\} . Thus, we cannot hope to find frame vectors for the Hardy operators \mathcal{H}_{1} and \mathcal{H}_{\infty} , since, as just mentioned, their spectra are not contained in \overline{\mathbb{D}} . However, we will show in § 8 that I-\mathcal{H}_{1}^{*} does possess frame vectors while neither I-\mathcal{H}_{1} nor I-\mathcal{H}_{\infty} do. Moreover, frame vectors for I-\mathcal{H}_{1}^{*} exist in abundance and we give plenty of specific examples as well as a general characterization (Corollary 8.8 ). Along the way, we show there are no frame vectors for I-\mathcal{C} , where \mathcal{C} is the classical (discrete) Cesàro