An extension and refinement of the theorems of Douglas and Sebestyén for unbounded operators
Yosra Barkaoui, Seppo Hassi · 2025
Read An extension and refinement of the theorems of Douglas and Sebestyén for unbounded operators by Yosra Barkaoui, Seppo Hassi (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-2507-13561.txt.
Introduction
Yosra Barkaoui and Seppo Hassi extend two classic operator theory results to unbounded operators, in a 2025 math paper. The Douglas theorem ties an operator inequality to a factorization, a way of writing one operator as a product of two. Sebestyén's theorem adds a middle step to that picture. Both were known for bounded operators, which keep vectors under control. Many operators in physics and engineering are unbounded, so the gap matters.
The pair give exact conditions for factorizing a closed, densely defined operator with a bounded non-negative partner. Their main result inserts a self-adjoint middle operator between the two sides of the Douglas inequality. The proof first lifts Sebestyén's result to the unbounded case. The authors say the sharper forms should serve work on operator equations and control.
Catalog record
| CREATOR | Yosra Barkaoui, Seppo Hassi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 27 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-13561
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
An extension and refinement of the theorems of Douglas and Sebestyén for unbounded operators Yosra Barkaoui, Seppo Hassi Abstract . For a closed densely defined operator T T from a Hilbert space ℌ \mathfrak{H} to a Hilbert space 𝔎 , \mathfrak{K}, necessary and sufficient conditions are established for the factorization of T T with a bounded nonnegative operator X X on 𝔎 . \mathfrak{K}. This results yields a new extension and a refinement of a well-known theorem of R.G. Douglas, which shows that the operator inequality A ∗ A ≤ λ 2 B ∗ B , λ ≥ 0 A^{*}A\leq\lambda^{2}B^{*}B,\lambda\geq 0 , is equivalent to the factorization A = C B A=CB with ‖ C ‖ ≤ λ \|C\|\leq\lambda . The main results give necessary and sufficient conditions for the existence of an intermediate selfadjoint operator H ≥ 0 H\geq 0 , such that A ∗ A ≤ λ H ≤ λ 2 B ∗ B A^{*}A\leq\lambda H\leq\lambda^{2}B^{*}B . The key results are proved by first extending a theorem of Z. Sebestyén to the setting of unbounded operators. 1991 Mathematics Subject Classification 47A07, 47A62, 47A63, 47B02, 47B25 1. Introduction Let (\mathfrak{H},(.,.)_{\mathfrak{H}}) be a Hilbert space and denote by \mathbf{B}(\mathfrak{H}) the class of bounded everywhere defined operators on \mathfrak{H}. For T,B\in\mathbf{B}(\mathfrak{H}), R.G. Douglas [ 6 , Theorem 1] showed, in 1966, that the following equivalence holds for some \lambda\geq 0: (1.1) TT^{*}\leq\lambda^{2}\,BB^{*}\quad\Leftrightarrow\quad T=BC,\ C\in\mathbf{B}(\mathfrak{H})\quad\Leftrightarrow\quad\mathrm{ran}\,T\subseteq\mathrm{ran}\,B. Later, in 1983, Z. Sebestyén [ 10 ] established the following characterization for a related problem: (1.2) T^{*}T\leq\lambda\,T^{*}B\textrm{ for some }\,\lambda\geq 0\quad\Leftrightarrow\quad T=XB,\ X\in\bm{B^{+}}(\mathfrak{H}), where T,B\in\mathbf{B}(\mathfrak{H}) and \bm{B^{+}}(\mathfrak{H}) stands for the class of bounded nonnegative operators on \mathfrak{H}. In fact, the inequality in ( 1.2 ) is closely connected to the one in ( 1.1 ), since the identity T=XB implies the following two inequalities: (1.3) T^{*}T\leq\lambda T^{*}B\leq\lambda^{2}B^{*}B. Therefore, the existence of a product presentation T=XB , where the factor X is not only bounded, but also nonnegative involves an intermediate nonnegative selfadjoint operator \lambda T^{*}B lying between T^{*}T and \lambda^{2}B^{*}B in the theorem of Douglas. The study of such factorizations has since been extended to more general settings. For instance, in 2013, D. Popovici and Z. Sebestyén [ 9 , Theorem 2.2] generalized the second equivalence in ( 1.1 ) to multivalued linear operators (linear relations) and showed that T\subseteq BC for some liner relation C if and only if \mathrm{ran}\,T\subseteq\mathrm{ran}\,B. On the other hand, the first equivalence in ( 1.1 ) was established by S. Hassi and H.S.V. de Snoo [ 7 ] for both unbounded linear operators and linear relations in 2015. As to ( 1.2 ), its extension has been recently studied by the present authors in the context of closed unbounded densely defined operators T,B from the Hilbert space \mathfrak{H} to another Hilbert space \mathfrak{K} with domains \mathrm{dom}\,T and \mathrm{dom}\,B. More precisely, it is shown in [ 3 , Theorem 2.7] that (1.4) T^{*}T\leq\lambda\,T^{*}B\quad\Leftrightarrow\quad X\overline{B_{0}}\subseteq T,\ X\in\bm{B^{+}}(\mathfrak{H}), where B_{0}:=B\upharpoonright\mathrm{dom}\,(T^{*}B) and T^{*}B\geq 0 is selfadjoint. Moreover, when in addition, B is nonnegative and selfadjoint, also the factorization T=XB is characterized in [ 3 ] by means of quasi-affinity to a nonnegative selfadjoint operator. Furthermore, such a factorization is shown to imply several local spectral properties for T studied further in [ 2 ] . In the case of bounded operators Sebestyén’s result in ( 1.2 ) has also been studied in recent papers by M. L. Arias, G. Corach, and M. C. Gonzalez [ 1 ] and by M. Contino, M. A. Dritschel, A. Maestripieri, and S. Marcantognini [ 5 ] . In the last paper also some local spectral theoretic results for T=XB\in\mathbf{B}(\mathfrak{H}) have been established. In this paper, we improve ( 1.4 ) and establish a complete unbounded analog of ( 1.2 ) without the core condition B=\overline{B_{0}} and the selfadjointness assumption on T^{*}B . Inspired by the work of Sebestyén, our approach hinges on the construction of an auxiliary Hilbert space by means of a sesquilinear form \tau_{\scriptstyle T,B}[f,g]:=(Tf,Bg)_{\mathfrak{K}}, f,g\in\mathrm{dom}\,\tau_{\scriptstyle T,B}=\mathrm{dom}\,B\subseteq\mathrm{dom}\,T . In the first step, the following equivalences are proved in 2.1 : (1.5) XB\subseteq T\Leftrightarrow\tau_{T}\leq\lambda\,\tau_{\scriptstyle T,B}\ \Leftrightarrow\|Tf\|_{\mathfrak{K}}^{2}\leq\,{\lambda}(Tf,Bf)_{\mathfrak{K}}, where the form \tau_{T} is defined by \tau_{T}[f]=(Tf,Tf)_{\mathfrak{K}} . An important further result is that the form \tau_{\scriptstyle T,B} is closable and, therefore, its closure gives rise to a nonnegative selfadjoint operator H , and the inequality in ( 1.5 ) can be described explicitly by means of H , X and B as follows: (1.6) T^{*}T\leq\,\lambda H,\quad H=B^{*}X^{\frac{1}{2}}\overline{X^{\frac{1}{2}}B}; see 2.2 . An essential difference here is that in ( 1.4 ) the operator T^{*}B is assumed to be selfadjoint, while ( 1.6 ) shows that T^{*}B is in general just a symmetric restriction of H=H^{*}\geq 0 . Moreover, here \mathrm{dom}\,B is a core for the form \overline{\tau_{\scriptstyle T,B}} generating the operator H . A further study of the inequality T^{*}T\leq\,\lambda H , where H is only assumed to be a selfadjoint operator (i.e. without the specific formula for H in ( 1.6 )), is carried out and yields further equivalent conditions for ( 1.5 ) in 2.1 , for instance: \displaystyle XB\subseteq T\Leftrightarrow\,T^{*}T\leq\,\lambda H,\text{ where }H\subseteq B^{*}T\text{ and }\mathrm{dom}\,B\subseteq\mathrm{dom}\,H^{\frac{1}{2}}. This not only completes the extension of Sebestyén’s result to the present unbounded framework, but the above mentioned results motivate the investigation of the reversed version of the above inequalities to be studied in Section 3 . Namely, also in the present case of unbounded operators, the inequality T^{*}T\leq\lambda H implies the inequality H\leq\lambda B^{*}B ; see 2.1 . This second inequality will be characterized in 3.1 with some further results, completing the study of the inequalities ( 1.1 )–( 1.3 ) in the case of unbounded operators. 2. An extension of Sebestyén’s theorem for unbounded operators The main purpose of this article is in fact to solve the above problem and present analogue characterizations for the factorization of T as in ( 1.2 ). Our first approach is inspired by Sebestyén theorem [ 10 ] , which we now extend to the unbounded case. Theorem 2.1 . Let (\mathfrak{H},(\cdot,\cdot)_{\mathfrak{H}}),(\mathfrak{K},(\cdot,\cdot)_{\mathfrak{K}}) be two complex Hilbert spaces and let T,B:H\rightarrow K be closed densely defined operators. Then the following statements are equivalent for some \lambda\geq 0: (1) \|Tf\|_{\mathfrak{K}}^{2}\leq\,\lambda\,(Tf,Bf)_{\mathfrak{K}} for all f\in\mathrm{dom}\,B\subseteq\mathrm{dom}\,T; (2) there exists X\in\bm{B^{+}}(\mathfrak{K}) such that \|X\|_{\mathfrak{K}}\leq\lambda and XB\subseteq T . In this case, X can be selected such that \mathrm{ran}\,X\subseteq\overline{\mathrm{ran}}\,T . Assume that (1) holds. Then (Tf,Bg)_{\mathfrak{K}} , f,g\in\mathrm{dom}\,B , defines a nonnegative sesquilinear form in the Hilbert space \mathfrak{H} . Observe, that (2.1) (Tf,Bf)_{\mathfrak{K}}=0\quad\Leftrightarrow\quad Tf=0,\quad f\in\mathrm{dom}\,B. By completing the quotient space [\mathrm{dom}\,B/(\ker T\cap\mathrm{dom}\,B)] one obtains a Hilbert space \mathfrak{K}_{B} whose inner product is denoted by \langle\widetilde{f},\widetilde{g}\rangle_{\mathfrak{K}_{B}} , \widetilde{f},\widetilde{g}\in\mathfrak{K}_{B} , such that (2.2) \langle\widetilde{f},\widetilde{g}\rangle_{\mathfrak{K}_{B}}=(Tf,Bg)_{\mathfrak{K}},\quad f,g\in\mathrm{dom}\,B. Now let V:\mathfrak{K}_{B}\rightarrow\mathfrak{K} be defined by V\widetilde{f}=Tf\quad\text{for all }f\in\mathrm{dom}\,B. Then V is a well-defined linear operator by ( 2.1 ) and ( 2.2 ), and it follows from (1) that it is bounded by \sqrt{\lambda}. It is claimed that V^{*}Bf=\widetilde{f}\quad\text{for all }f\in\mathrm{dom}\,B. To see this, let f,g\in\mathrm{dom}\,B and \widetilde{g}\in\mathfrak{K}_{B}. Then, \langle\widetilde{g},V^{*}Bf\,\rangle_{\mathfrak{K}_{B}}=(V\widetilde{g},Bf)_{\mathfrak{K}}=(Tg,Bf)_{\mathfrak{K}}=\langle\widetilde{g},\widetilde{f}\,\rangle_{\mathfrak{K}_{B}}, and therefore V^{*}Bf=\widetilde{f}, as claimed. Consequently, X:=VV^{*}\in\bm{B^{+}}(\mathfrak{K}) and one has \|X\|_{\mathfrak{K}}\leq\lambda, \overline{\mathrm{ran}}\,X\subseteq\mathrm{\overline{ran}}\,T, by construction, and XBf=V\widetilde{f}=Tf\quad\text{for all }f\in\mathrm{dom}\,B. This proves that XB\subseteq T. For the converse, assume (2) and let f\in\mathrm{dom}\,B\subseteq\mathrm{dom}\,T. Then, \|Tf\|^{2}=\|XBf\|^{2}\leq\,\|X^{{}^{\frac{1}{2}}}\|^{2}\|X^{{}^{\frac{1}{2}}}Bf\|^{2}\leq\lambda(Tf,Bf)_{\mathfrak{K}}, which completes the proof