Kadec-Klee property for convergence in measure of noncommutative Orlicz spaces (Q1682102): Difference between revisions

If \(\sigma\) is a vector space topology on a Banach space \(X\) which is weaker than the norm topology of \(X\), then \(X\) is said to have the Kadec-Klee property for \(\sigma\) (KKP-\(\sigma\)) if in the unit sphere of \(X\) every sequence \(\sigma\)-converging to a norm one element converges in norm. In the original KKP, \(\sigma\) is the weak topology \(w\) but in many papers concerning Banach function spaces and, in particular, in the paper under review, \(\sigma\) is the topology of convergence in measure \(m\). The classical Lebesgue spaces \(L^p\) have KKP-\(w\) and KKP-\(m\) for \(1<p<\infty\), while \(L^1\) fails KKP-\(w\) but has KKP-\(m\). The setting of the paper is the class of Orlicz spaces \(L_\varphi\) and their noncommutative analogues \(L_\varphi(\widetilde{\mathcal{M}},\tau)\), where \((\mathcal{M},\tau)\) is a semifinite von Neumann algebra with a faithful normal semifinite trace \(\tau\). The authors consider KKP-\(m\) for such spaces with respect to the \(\Delta_2\)-condition. The main result reads: If \(\varphi\) satisfies the \(\Delta_2\)-condition, then \(L_\varphi(\widetilde{\mathcal{M}},\tau)\) has KKP-\(m\) and, if \(\mathcal{M}\) is non-atomic, then the converse implication holds, too. (In the proof of the converse part, positive numbers \(u_k\) and projections \(e_k\) are constructed such that \(2^k \tau(e_k)=1/\varphi(u_k)\) converges to \(0\) because \(\varphi(u_k)\to\infty\) which contradicts the claim \(2^k \tau(e_k)\to\infty\); but it seems that this does not affect the validity of the proof because the latter claim apparently is not used. A slightly misleading typo appears in the definition of order continuity in the preliminaries.) Amongst others, results by \textit{T. Fack} and \textit{H. Kosaki} [Pac. J. Math. 123, 269--300 (1986; Zbl 0617.46063)] are used and, in turn, their result that \(L_p(\widetilde{\mathcal{M}},\tau)\)-spaces have KKP-\(m\) for \(1<p<\infty\) is recovered; another consequence is that if \(\varphi\) satisfies the \(\Delta_2\)-condition, then \(L_\varphi(\widetilde{\mathcal{M}},\tau)\) has a local uniform monotonicity property. Further, as in the commutative case, the noncommutative Orlicz spaces \(L_\varphi(\widetilde{\mathcal{M}},\tau)\) turn out to be reflexive if and only if \(\varphi\) and its complementary function satisfy the \(\Delta_2\)-condition. The authors briefly mention three different natural approaches to non-commutative counterparts of Orlicz-spaces one of which -- the one followed by the authors -- considers Orlicz spaces as modular spaces [\textit{G. Sadeghi}, J. Math. Anal. Appl. 395, No. 2, 705--715 (2012; Zbl 1258.46025)]. The authors seem to assume the equivalence of the three approaches but without explicit reference. (But cf., for example, page 1056 of \textit{L. E. Labuschagne} and \textit{W. A. Majewski} [Ill. J. Math. 55, No. 3, 1053--1081 (2011; Zbl 1272.46052)]). Whence the question: to which extent is there an intersection between the main result of the paper under review and known results of Chilin, Dodds et al.? More specifically, what is the intersection with Corollary 4 of [\textit{F. A. Sukochev}, J. Aust. Math. Soc., Ser. A 59, No. 3, 343--352 (1995; Zbl 0854.46015)]?