Some new rich subspaces of C. Applications (Q707258)

Let \(S\) be a compact space and \({\mathcal C}(S)\) be the space of continuous functions on \(S\). A subspace \(X\) of \({\mathcal C}(S)\) is said to be a rich subspace of \({\mathcal C}(S)\) if there exists a probability measure \(\nu\) on \(S\) such that for every \(h\in {\mathcal C}(S)\) and every bounded sequence \((x_n)\) in \(X\) such that \(\| x_n\|_{L^1(\nu)}\to 0\), then \(\text{dist}(hx_n, X)\to 0\). This definition has been coined by \textit{P. Wojtasczyk} in his book ``Banach spaces for analysts'' (Cambridge Univ. Press) (1996; Zbl 0858.46002)], where it is shown that every rich subspace \(X\) of a space \({\mathcal C}(S)\) has Pełczyński's property \((V)\) and its dual \(X^\ast\) has the Dunford--Pettis property (so \(X\) itself has the Dunford--Pettis property). This notion has its origin in a paper of Bourgain (see the comments below). In the present paper, the author gives new examples of rich subspaces. His first main result is that every subspace of \({\mathcal C}(S)\) with reflexive annihilator is rich. This result depends on Rosenthal's theorem on reflexive subspaces of \(L^1\). In particular, he recovers some results of S. Kislyakov (\(L^1/Y\) has the Dunford--Pettis property for every reflexive subspace \(Y\) of \(L^1\)) and of \textit{G. Godefroy} and \textit{P. Saab} [C. R. Acad. Sci., Paris, Sér. I 303, 503--506 (1986; Zbl 0602.46014)]. That allows him to get that for every \(\Lambda(1)\) subset \(E\) of \({\mathbb N}\), \({\mathcal C}_{{\mathbb N}\setminus E} ({\mathbb T})\) is a rich subspace of \({\mathcal C}({\mathbb T})\) and so has Pełczyński's property \((V)\), and its dual has the Dunford--Pettis property. Moreover, if \(\Lambda\subseteq {\mathbb N}\) is such that \({\mathcal C}_\Lambda ({\mathbb T})\) does not contain \(c_0\), then \({\mathbb Z}^-\cup \Lambda\cup E\) is a Riesz subset of \({\mathbb Z}\). In the second part, the author studies the space \(U\) of uniformly convergent Fourier series. It can be viewed as a subspace of \({\mathcal C}(\widetilde{\mathbb T})\), with \(\widetilde{\mathbb T}={\mathbb T}\times \{0,1,1/2,1/3,\ldots\}\). It is shown that \(U\) and \(U_+=U_{\mathbb N}\) (the subspace of \(U\) whose functions have spectra in \({\mathbb N}\)) are rich subspaces of \({\mathcal C}(\widetilde{\mathbb T})\). This result has already been proved by \textit{S. F. Saccone} [Proc. Am. Math. Soc. 128, No. 6, 1813--1823 (2000; Zbl 0953.46014)], but the author improves it: for every \(\Lambda(2)\) subset \(E\) of \({\mathbb Z}\), \(U_{E^c}\) is rich, and for every \(\Lambda(2)\) subset \(E\) of \({\mathbb N}\), \(U_{{\mathbb N}\setminus E}\) is rich. Reviewer's comments: The present notion of rich subspace has no connection with the one studied by \textit{V. Kadets, N. Kalton} and \textit{D. Werner} in [Stud. Math. 159, No. 2, 195--206 (2003; Zbl 1055.46006)]. It comes from a paper of \textit{J. Bourgain} [Stud. Math. 77, 245--253 (1984; Zbl 0576.46040)], but the latter worked with weakly null sequences in the bidual \(X^{\ast\ast}\) of \(X\). More precisely, if one defines, for \(h\in {\mathcal C}(S)\), the map \(S_h: X\to {\mathcal S}/X\) by \(S_h(f)= hf +X\), then Bourgain showed that if \(S_h^{\ast\ast}\) sends weakly convergent sequences into norm convergent sequences (i.e., \(S_h^{\ast\ast}\) is a Dunford--Pettis operator) for every \(h\in {\mathcal C}(S)\), then \(X\) and \(X^\ast\) have the Dunford--Pettis property. This leads to several variants. \textit{B. J. Cole} and \textit{T. W. Gamelin} [J. Funct. Anal. 46, 158--220 (1982; Zbl 0569.46034)] introduced the notion of a tight subspace: \(X\) is called a tight subspace if the operators \(S_h\) are all weakly compact. \textit{S. Saccone} showed in [J. Funct. Anal. 148, No. 1, 86--116 (1997; Zbl 0886.46022)] that \(X\) has property \((V)\) when it is tight. The property asked for the \(S_h\)'s in the definition of rich subspaces (called nearly dominated by Saccone, and which may be traced back to \textit{J. Bourgain} [Bull. Soc. Math. Belg., Sér. B 35, 111--118 (1983; Zbl 0521.46016)] is halfway between being \(1\)-summing and being weakly compact and Dunford--Pettis. To conclude these comments, we note that Bourgain's paper also lead \textit{J. A. Cima} and \textit{R. M.Timoney} [Mich. Math. J. 34, 99--104 (1987; Zbl 0617.46058)] to the notion of a Bourgain algebra: the Bourgain algebra of a subspace \(X\) of \({\mathcal C}(S)\) is the set of functions \(h\in {\mathcal C}(S)\) for which \(S_h^{\ast\ast}\) is a Dunford--Pettis operator.