Numerical range and functional calculus in Hilbert space (Q880112): Difference between revisions

Let \(A\) be a bounded linear operator on a complex Hilbert space \(H\) with numerical range \(W(A)=\{\langle Ax,x\rangle: x\in H,\| x\|= 1\}\), where \(\langle\cdot,\cdot\rangle\) denotes the inner product in \(H\). The main result of this paper is that the inequality \[ \|[p_{ij}(A)]\|\leq 11.08\,\sup\{\|[p_{ij}(z)]\|: z\in W(A)\} \] holds for any \(m\)-by-\(n\) polynomial matrix \([p_{ij}]\). In particular, this implies that the homomorphism \(r\mapsto r(A)\) from the algebra of rational functions with poles off \(W(A)\) to \(B(H)\), the \(C^*\)-algebra of all operators on \(H\), is completely bounded with complete bound at most 11.08. This is a major achievement, vastly improving the previous result of \textit{B.\,Delyon} and \textit{F.\,Delyon} [Bull.\ Soc.\ Math.\ Fr.\ 127, No.\,1, 25--41 (1999; Zbl 0937.47004)] showing that the complete bound exists and depends on the area and diameter of \(W(A)\). The bulk of the proof consists, for \(A\) a finite matrix and \(W(A)\) replaced by a larger bounded convex domain \(\Omega\) of the plane with well-behaved boundary, in representing \([p_{ij}(A)]\) as the sum of three operator integrals, one of which has the operator form of the double layer potential of the domain \(\Omega\) as the kernel function, and estimating the magnitude of each of such integrals. The proof itself is intricate and uses classical analysis arguments. Since the constant is universal, independent of the matrix \(A\), its size, the degrees of the polynomials and \(m\) and \(n\), the inequality can be easily generalized to operators on infinite-dimensional spaces. There is even a generalization to closed unbounded operators and an application to the theory of cosine functions. The author conjectures that the smallest complete bound should be ``2'', which is attained for the polynomial \(p(z)= z\) and the matrix \(A=\left[\begin{smallmatrix} 0 & 2\\ 0 & 0\end{smallmatrix}\right]\). The conjecture has been verified by the author before for \(\Omega\) a circular disc [Integral Equations Oper.\ Theory 48, No.\,4, 461--477 (2004; Zbl 1069.47004)].