Constrained von Neumann inequalities

From MaRDI portal



Abstract: An equivalent formulation of the von Neumann inequality states that the backward shift S∗ on ell2 is extremal, in the sense that if T is a Hilbert space contraction, then |p(T)|leq|p(S∗)| for each polynomial p. We discuss several results of the following type : if T is a Hilbert space contraction satisfying some constraints, then S∗ restricted to a suitable invariant subspace is an extremal operator. Several operator radii are used instead of the operator norm. Applications to inequalities of coefficients of rational functions positive on the torus are given.


Let \(S^*\) be the backward shift operator on \(\ell^2\), that is, \(S^*(x_0,x_1)=(x_1,x_2,\ldots)\). An equivalent version of the classical von Neumann inequality says that for every Hilbert space contraction \(T\) and every polynomial \(p\) we have \(\| p(T)\| _{{\mathcal B}(H)}\leq \| p(S^*)\| _{{\mathcal B}(\ell^2)}\); that is, \(S^*\) is extremal. The paper under review generalizes these type of inequalities. To this end, the notion of admissible operator radius is introduced; this is a map \(w:{\mathcal B}(H)\to [0,\infty]\) such that (i) \(w(U^*TU)=w(T)\) for each unitary \(U\) and \(T\in {\mathcal B}(H)\); (ii) if \(E\) is an invariant subspace for \(T\in{\mathcal B}(H)\), then \(w(T| _E)\leq w(T);\) and (iii) if \(T^{(\infty)}\) denotes the countable orthogonal sum \(T\oplus T\oplus\dots\), then \(w(T^{(\infty)})=w(T)\). The results obtained then roughly read as follows: if \(T\) is a Hilbert space contraction satisfying some constraints and \({\omega}\) is an admissible operator radius, then there exists an invariant subspace \(E\) of \(S^*\) such that for every polynomial \(p\) one has \(\omega(p(T))\leq \omega(p(S^*| _E))\). As one of many interesting applications of their results, the following estimates for the numerical radius \(\omega_2(T)=\sup\{| \langle Tx,x\rangle| : x\in H\), \(\| x\| =1\}\) of \(T\) are given: If \(T\in {\mathcal B}(H)\) satisfies \(\| T\| \leq 1\), \(T^3\geq 0\) and \(I-2T^*T+T^{*2}T^2=0\), then \(\omega_2(T)\leq \sqrt{7/24}\) and \(\omega_2(T^2)\leq \sqrt{1/12}\), where the constants are the best possible. Further applications to harmonic analysis involve generalizations of coefficient inequalities for positive trigonometric polynomials due to Fejer, Egerváry and Százs.



Cites work









This page was built for publication: Constrained von Neumann inequalities

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1599838)