Some remarks on thin operators (Q4831442)

Following \textit{R.~D. Neidinger} [Lect. Notes Math. 1166, 116--128 (1985; Zbl 0587.47023)], an operator (continuous, linear) \(T: X \to Y\) between infinite-dimensional Banach spaces \(X\) and \(Y\) is called thin if, given any infinite-dimensional subspace \(Y_1\) of \(Y\), there exists a \(\delta > 0\) such that \(B_{Y_1} \not \subseteq \lambda T B_X + \delta B_Y\) for every \(\lambda > 0\) (here \(B_Z\) means the closed unit ball of \(Z\)). Neidinger has shown that any compact operator is thin and any thin operator is strictly singular. Later, \textit{V.~Caselles} and \textit{M.~González} [``Compactness properties of strictly singular operators in Banach lattices'', In: Semesterbericht Funktionalanalysis (Universität Tübingen), 175--189 (1987; per bibl.)] showed that if \((\Omega, \Sigma, \mu)\) and \((\Omega', \Sigma', \mu')\) are probability spaces and \(1 \leq q \leq p < \infty\), \(p \neq 1\), then for the class of regular operators \(T: L^p(\mu) \to L^q(\mu')\) the above notions coincide (precisely, each regular strictly singular operator is compact). The main result of the paper under review asserts that a weaker result is true for a more general case of values for \(p\) and \(q\). Namely, let \((\Omega, \Sigma, \mu)\) and \((\Omega', \Sigma', \mu')\) be probability spaces and \(1 < p < \infty, \,\,\, 1 \leq q < \infty\). Then each regular strictly singular operator \(T: L^p(\mu) \to L^q(\mu')\) is thin. As a consequence, it is obtained that if \(T: L^p(\mu) \to L^q(\mu')\) is thin and \(0 \leq S \leq T\), then so is \(S\). Some results then apply to Köthe function spaces.