\({\Lambda}(p)\)-spaces (Q2253281)

A closed subspace \(H\) of \(L^p = L^p (0, 1)\), \(1 \leq p < \infty\), is said to be a \(\Lambda (p)\)-space if convergence in the \(L^p\)-norm is equivalent to convergence in measure on \(H\), or, equivalently, if for some (and hence for each) \(0 < q < p\), the \(L^p\)-norm is equivalent on \(H\) to the \(L^q\)-``norm''. When \(G\) is an abelian compact group and \(\Lambda \subseteq \hat G\), the space \(L^p_\Lambda (G)\) is a \(\Lambda (p)\)-space if and only if \(\Lambda\) is a \(\Lambda (p)\)-set. In the book of \textit{F. Albiac} and \textit{N. J. Kalton} [Topics in Banach space theory. Graduate Texts in Mathematics 233. Berlin: Springer (2005; Zbl 1094.46002), Definition 6.4.4], such a space is called strongly embedded in \(L^p\). It is known that if \(H\) is a closed subspace of \(L^p\) and its unit ball \(B_H\) is equi-integrable (``has equi-absolutely continuous norms'' in the author's terminology), then \(H\) is a \(\Lambda (p)\)-space. In this paper, the author shows, using \textit{H. P. Rosenthal}'s theorem [Ann. Math. (2) 97, 344--373 (1973; Zbl 0253.46049)], that the converse is true for \(1 \leq p < 2\) (Theorem~2), but that this is not the case for \(p \geq 2\). For \(p > 2\), he uses the space \(L^p_\Lambda\) constructed by \textit{J. Bourgain} [Acta Math. 162, No. 3--4, 227--245 (1989; Zbl 0674.43004)], where \(\Lambda\) is a \(\Lambda (p)\)-set which is \(\Lambda (q)\)-set for no \(q > p\); for \(p = 2\), he uses a space spanned by a suitable sequence of independent random variables. In general, if \(1 \leq p < \infty\) and \(H\) is a closed subspace of \(L^p\), the author shows (Theorem 1) that \(H\) is a \(\Lambda (p)\)-space if there exists a rearrangement invariant (r.i.) space \(X\) such that \(H \subseteq X \subseteq L^p\) and the inclusion operator \(I : X \to L^p\) is disjointly strictly singular (DSS), meaning that there is no disjoint sequence of non-null vectors \(\{x_n\}_{n = 1}^\infty\) for which the norms \(\| \, . \, \|_X\) and \(\| \, . \, \|_p\) are equivalent on their closed linear span. For \(p \geq 2\), one cannot ensure that the unit ball \(B_H\) is equi-integrable, but for \(p = 2\) this happens when \(H\) has a supplementary property: there exists an r.i. space \(X \subseteq L^2\) such that the inclusion operator \(I : X \to L^2\) is DSS and the norms \(\| \, . \, \|_X\) and \(\| \, . \, \|_2\) are equivalent on \(H\) (Theorem 3). For \(p > 2\), this does not suffice: for every \(p > 2\), there exists a \(\Lambda (p)\)-space \(H \subseteq L^p\) and an Orlicz space \(L^F \subseteq L^p\) such that the inclusion operator \(I : L^F \to L^p\) is DSS, the norms \(\| \, . \, \|_F\) and \(\| \, . \, \|_p\) are equivalent on \(H\), but \(B_H\) fails to be equi-integrable (Theorem 4). The construction is based on a result of \textit{F. Hernández} and \textit{B. Rodríguez-Salinas} [Isr. J. Math. 68, No. 1, 27--55 (1989; Zbl 0689.46006)]. In the last part, the author introduces the notion of uniformizable \(\Lambda (2)\)-space and proves that for \(H \subseteq L^2\), \(B_H\) is equi-integrable if and only if \(H\) is a uniformizable \(\Lambda (2)\)-space (see [\textit{J. J. F. Fournier}, Colloq. Math. 51, 119--129 (1987; Zbl 0663.42014)] for uniformizable \(\Lambda (2)\)-sets). He constructs two orthogonal \(\Lambda (2)\)-spaces \(H_1\) and \(H_2\) such that their sum \(H_1 \oplus H_2\) is not a \(\Lambda (2)\)-space and shows that, in general, \(H_1 \oplus H_2\) is a \(\Lambda (2)\)-space if one assumes, moreover, that one of the two spaces \(H_1\) or \(H_2\) is actually a uniformizable \(\Lambda (2)\)-space.