A proof of Parrott's theorem on quotient norms (Q1123376): Difference between revisions

Let H, K be two Hilbert spaces and \(H_ 1\), \(K_ 1\) be two subspaces of H and K respectively. Denote by B(H,K) the Banach space of all linear bounded operators from H into K endowed with the operator norm. Let \(B(H_ 1,K_ 1)\) be the subspace of those \(L\in B(H,K)\) for which \(L(H_ 1)\subset K_ 1\) and \(L(H^{\perp}_ 1)=0\), where \(H^{\perp}_ 1\) means the orthocomplement of \(H_ 1\). Let \(p: B(H,K)\to B(H,K)/B(H_ 1,K_ 1)\) denote the quotient mapping. It can be easily checked that \(\| p(T)\| \geq \max \{\| T|_{H_ 0}\|,\| T^*|_{K_ 0}\| \}.\) \textit{S. Parrott} [J. Funct. Anal. 30, 311-328 (1978; Zbl 0409.47004)] has proved a remarkable result that this inequality is actually an equality and that \(\| p(T)\| =\| T+X\|\) for some \(X\in B(H_ 1,K_ 1)\). The paper under review is devoted to another proof of this fact. The author's approach is based on his earlier investigations on an extension of various kinds of operators acting from a subset of H into H to operators defined on the whole H [Acta Sci. Math. (Szeged) 46, 299-301 (1983; Zbl 0551.47005), Acta Math. Hungar. 47, 29-32 (1986; Zbl 0621.47027)]. Both these papers are recommended for better understanding the proof in the paper under review.