Adjacent cross-sections of the commutant of Hilbert space operators

László Kérchy · 2026

Read Adjacent cross-sections of the commutant of Hilbert space operators by László Kérchy (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-2605-23846.txt.

Introduction

László Kérchy studies cross sections of the commutant of upper triangular operator matrices. He builds similarity models from nearly unchanged half spaces. Squeezing commutant operators through a standard map yields a linear manifold. In that manifold, every element drops to rank at most two. A cyclic trait makes the manifold transitive. Earlier work mapped its three dimensional cross sections and their standard bases in three by three matrices. This paper adds matching rules between side by side cross sections, tightening the picture.

Catalog record

CREATORLászló Kérchy
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38280
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.23846

Browse the full library · About the Tower of Babel

Full text

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

Adjacent cross-sections of the commutant of Hilbert space operators László Kérchy Abstract Applying the techniques resulting the existence of almost invariant half-spaces, similarity models T ^ \widehat{T} can be given for upper triangular operator-matrices T = [ A C 0 B ] T=\left[\begin{matrix}A&C\\ 0&B\end{matrix}\right] . The model T ^ \widehat{T} is also an operator-matrix, containing two diagonal operators in the general case [ K25 ] , and the unilateral shift S S together with a diagonal operator in the particular case when A A is similar to S S [ K26 ] . Well-chosen compressions of operators in the commutant { T ^ } ′ \{\widehat{T}\}^{\prime} form a linear manifold ℒ ^ \widehat{\cal L} satisfying the condition that every X ^ ∈ ℒ ^ \widehat{X}\in\widehat{\cal L} is transformed into Y ^ \widehat{Y} with rank ​ Y ^ ≤ 2 \hbox{\rm rank\,}\widehat{Y}\leq 2 by a canonical mapping. Furthermore, a cyclcity property of { T } ′ \{T\}^{\prime} yields transitivity of ℒ ^ \widehat{\cal L} . In [ K25 ] and [ K26 ] the 3-dimensional cross-sections of ℒ ^ \widehat{\cal L} have been investigated characterizing the canonical bases occurring in the corresponding subspaces of the matrix-algebra M 3 ​ [ ℂ ] M_{3}[\mathbb{C}] . In this paper new conditions are provided by studying matching of adjacent cross-sections. AMS Subject Classification (2020): 47A08, 47A15, 47A20, 47L05. Key words : invariant subspace, hyperinvariant subspace, compression of operator, unilateral shift, transitive subspace of matrices, singular matrices. 1 Introduction Let {\cal H}_{1} and {\cal H}_{2} be complex Hilbert spaces with \dim{\cal H}_{1}=\dim{\cal H}_{2}=\aleph_{0} and let us consider a (bounded, linear) operator T=\left[\begin{matrix}A&C\\ 0&B\end{matrix}\right]\in{\cal L}({\cal H}_{1}\oplus{\cal H}_{2}). It is clear that the subspace {\cal H}_{1}(\equiv{\cal H}_{1}\oplus\{0\}) is invariant for T:\,T{\cal H}_{1}\subset{\cal H}_{1} . The commutant \{T\}^{\prime} of T consists of those operators Q\in{\cal L}({\cal H}_{1}\oplus{\cal H}_{2}) which commute with T:\,QT=TQ . It will be assumed that T is not scalar multiple of the identity. We are interested in the following basic question: Question 1 Does T have a nontrivial hyperinvariant subspace, that is a subspace {\cal M}\in\hbox{\rm Lat}({\cal H}_{1}\oplus{\cal H}_{2}) , different from \{0\}\oplus\{0\} and {\cal H}_{1}\oplus{\cal H}_{2} , which is invariant for every Q\in\{T\}^{\prime}:\,Q{\cal M}\subset{\cal M} ? The hyperinvariant subspaces of T form the complete lattice \hbox{Hlat\,}T . So the question above asks whether \hbox{Hlat\;}T\neq\left\{\{0\}\oplus\{0\},{\cal H}_{1}\oplus{\cal H}_{2}\right\} . Notice that an affirmative answer for Question 1 yields that every nonscalar operator A has a proper hyperinvariant subspace. Indeed, take T=A\oplus A and observe that \hbox{Hlat\,}T=\left\{{\cal N}\oplus{\cal N}:{\cal N}\in\hbox{Hlat\,}A\right\} . In that way we would obtain positive answers for the hyperinvariant subspace problem (HSP), and in particular for the invariant subspace problem (ISP), which are arguably the most challenging open questions in operator theory. A recent approach to (ISP) in an arbitrary (infinite dimensional) Banach space {\cal X} was initiated in [ APTT ] , and culminated in [ Tc ] proving that every operator T on {\cal X} has an almost invariant half-space . We recall that {\cal M} is a half-space in {\cal X} if {\cal M} and {\cal X}/{\cal M} are infinite dimensional. Furthermore, {\cal M} is almost invariant for T if {\cal M} is invariant for a finite rank perturbation of T . The technique of this approach was refined in the Hilbert space setting in [ JKP ] , which led to a structure theorem in the paper [ HP ] . Applying this method to our block-triangular operator T\in{\cal L}({\cal H}_{1}\oplus{\cal H}_{2}) more delicate similarity model \widehat{T}\in{\cal L}({\cal H}^{(4)}) has been constructed for a translate of T in [ K25 ] . Here {\cal H} is an auxiliary complex Hilbert space with \dim{\cal H}=\aleph_{0} and {\cal H}^{(4)}={\cal H}\oplus{\cal H}\oplus{\cal H}\oplus{\cal H} . The model \widehat{T} contains two diagonal components: D,D_{*} . The central linear manifold \widehat{\cal L}\subset{\cal L}({\cal H}) consists of the entries of the operators in \{\widehat{T}\}^{\prime} belonging to {\cal L}(\hbox{dom\,}D,\hbox{dom\,}D_{*}) . The crucial property of \widehat{\cal L} is that \hbox{\rm rank\,}(\widehat{X}D-D_{*}\widehat{X})\leq 2 holds, for every \widehat{X}\in\widehat{\cal L} . Furthermore, a cyclicity condition of \{T\}^{\prime} , which can be assumed in the quest for proper hyperinvariant subspaces, implies that \widehat{\cal L} is transitive. An orthonormal basis \{e_{k}\}_{k=1}^{\infty} is fixed in {\cal H} . For every r\in\mathbb{N} , we consider the subspace {\cal E}_{3,r}=\vee\{e_{r},e_{r+1},e_{r+2}\} . The cross-section \widehat{\cal L}_{r} consists of the compressions of the operators in \widehat{\cal L} to {\cal E}_{3,r} . This subspace of {\cal L}({\cal E}_{3,r}) inherits the rank-2 property and the transitivity of \widehat{\cal L} . The algebra-isomorphism M_{3,r}\colon{\cal L}({\cal E}_{3,r})\to M_{3}[\mathbb{C}] assigns to each operator its matrix in the orthonormal basis (e_{r},e_{r+1},e_{r+2}) . The main result in [ K25 ] characterizes the arising subspaces M_{3,r}(\widehat{\cal L}_{r}) in terms of properties of a canonical basis. In [ K26 ] the case when A is similar to the unilateral shift S was investigated. Now the similarity model \widehat{T}\in{\cal L}({\cal H}^{(3)}) of a translate of T contains S and a diagonal operator D_{*} . The linear manifold \widehat{\cal L}\subset{\cal L}({\cal H}) consists of those entries of the operators in \{\widehat{T}\}^{\prime} which belong to {\cal L}(\hbox{dom\,}S,\hbox{dom\,}D_{*}) . For every \widehat{X}\in\widehat{\cal L} , we have \hbox{\rm rank\,}(\widehat{X}S-D\widehat{X})\leq 1 , where D is a translate of D_{*} . Furthermore, a cyclicity condition of \{T\}^{\prime} implies that \widehat{\cal L} is transitive. For every r\in\mathbb{N} , let \widehat{\cal L}_{r} be again the compression of \widehat{\cal L} to the subspace {\cal E}_{3,r}=\vee\{e_{r},e_{r+1},e_{r+2}\} . The cross-section \widehat{\cal L}_{r} inherits the transitivity of \widehat{\cal L} , and \hbox{\rm rank\,}\Gamma_{r}(X)\leq 2 is true for every X\in\widehat{\cal L}_{r} with a canonical mapping \Gamma_{r}\in{\cal L}({\cal L}({\cal E}_{3,r})) . The main result of [ K26 ] provides complete characterization of the arising subspaces M_{3,r}(\widehat{\cal L}_{r}) in terms of conditions concerning the elements of a canonical basis. Our aim in this paper is to give further conditions for these bases by studying matching of adjacent cross-sections. In Section 2 compressions of operators are studied. In Section 3 we examine the general case, while in Sections 4 and 5 we discuss the case when A is similar to S . The conditions obtained simplify the featured canonical bases in the compressions, and establish connections between the parameters in the neighboring cross-sections. If any of the resulting conditions fails, then T has a nontrivial hyperinvariant subspace; more precisely, there exists a nonzero vector x\in{\cal H}_{1} such that \{T\}^{\prime}x is not dense in {\cal H}_{1}\oplus{\cal H}_{2} . This is because the rules concerning ranks cannot be broken. 2 Compressions Let \{e_{k}\}_{k=1}^{\infty} be an orthonormal basis in the Hilbert space {\cal H} . For any subspace {\cal E}\in\hbox{\rm Lat\,}{\cal H} , the compression mapping C_{{\cal H},{\cal E}}\colon{\cal L}({\cal H})\to{\cal L}({\cal E}),\;X\mapsto P_{\cal E}X\big|_{\cal E} is a bounded linear transformation. (Here P_{\cal E}\in{\cal L}({\cal H}) stands for the orthogonal projection onto {\cal E} .) Given any {\cal F}\in\hbox{\rm Lat\,}{\cal E} , we may consider the mapping C_{{\cal E},{\cal F}}\colon{\cal L}({\cal E})\to{\cal L}({\cal F}),\;X\mapsto P_{\cal F}X\big|_{\cal F}. The following simple statement establishes connection among these transformations. Lemma 2 We have \;C_{{\cal E},{\cal F}}\circ C_{{\cal H},{\cal E}}=C_{{\cal H},{\cal F}} . Proof. For completeness we sketch the easy proof. Given any X\in{\cal L}({\cal H}) , let Y=C_{{\cal H},{\cal E}}(X)\in{\cal L}({\cal E}) and Z=C_{{\cal E},{\cal F}}(Y)\in{\cal L}({\cal F}) . For every x\in{\cal F} we have Zx=P_{\cal F}Yx=P_{\cal F}P_{\cal E}Xx=P_{\cal F}Xx

Read the complete text