On geometric properties of the spaces \(L^{p(x)}\) (Q2430523): Difference between revisions

Recently, the so-called variable exponent (or generalised Lebesgue) spaces \(L^{p(x)}(\mu)\) have attracted attention in the theory of PDEs [cf. \textit{L.~Diening, P.~Harjulehto, P. Hästö} and \textit{M.~Růžička}, Lebesgue and Sobolev spaces with variable exponents. Lecture Notes in Mathematics 2017. Berlin: Springer (2011; Zbl 1222.46002)]. The authors of the present paper look at their Banach space properties. Let us recall that \(L^{p(x)}(\mu)\) consists of all measurable functions such that for some \(\lambda>0\) \[ \int_{\Omega_0} \Bigl| \frac{f(x)}\lambda \Bigr|^{p(x)} \,d\mu(x) + \mathop{\text{ess\,sup}}_{x\in\Omega_\infty} \Bigl| \frac{f(x)}\lambda \Bigr| <\infty, \] where \(p: \Omega\to [1,\infty]\) is a measurable function, \(\Omega_0=\{x: p(x)<\infty\}\) and \(\Omega_\infty =\{x: p(x)=\infty\}\); it is normed by the corresponding Luxemburg norm, i.e., the infimum of all \(\lambda\) for which the above expression is \({\leq1}\). To formulate the authors' results, one needs the notation \(p_{\min} = \mathop{\text{ess\,inf}} p(x)\), \(p_{\max} = \mathop{\text{ess\,sup}} p(x)\), \(\Omega_1=\{x: p(x)=1\}\). The following theorem is proved for nonatomic measure spaces: (1)~\(L^{p(x)}(\mu)\) is reflexive if and only if \(L^{p(x)}\) is uniformly convex if and only if \(1<p_{\min} \leq p_{\max} <\infty\). (2)~\(L^{p(x)}(\mu)\) has the Radon-Nikodym property if and only if \(\mu(\Omega_1)=0\) and \(p_{\max}<\infty\). (3)~\(L^{p(x)}(\mu)\) has the Daugavet property if and only if \(\mu(\{x: 1<p(x)<\infty\})=0\). The paper is very clearly written. As the authors point out, parts of (1) have already been known.