Grand Sobolev spaces and their applications in geometric function theory and PDEs (Q395439)

The grand Lebesgue space is defined on a bounded domain \(\Omega\) by \[ \| g \|_{L^{p)}(\Omega)} = \sup_{0 < \epsilon \leq p - 1} \epsilon \left(\frac{1}{| \Omega|} \int_{\Omega} |g(x)|^{p - \epsilon} \, dx\right)^{\frac{1}{p - \epsilon} }. \] These are Banach spaces, introduced by \textit{T. Iwaniec} and \textit{C. Sbordone} [Arch. Ration. Mech. Anal. 119, No.~2, 129--143 (1992; Zbl 0766.46016)], which contain the Marcinkiewicz spaces and the \(L^p\) spaces, but they are not reflexive nor is \(L^p\) dense in \(L^{p)}(\Omega)\). The closure of \(L^p\) in \(L^{p)}(\Omega)\) consists of those functions in \(L^{p)}(\Omega)\) with the additional property that \[ \lim_{\epsilon \to 0} \epsilon \left( \frac{1}{| \Omega|} \int_{\Omega} |g(x)|^{p - \epsilon} \, dx\right)^{\frac{1}{p - \epsilon} } = 0. \] This space is denoted by \(L^{p)}_b(\Omega)\). The Orlicz class \(L^p(\log L)^{-1}(\Omega)\) is contained in both \(L^{p)}(\Omega)\) and \(L^{p)}_b\). The grand Sobolev spaces are defined in analogy with the ordinary Sobolev spaces (\(u \in L^p,\, | \nabla u | \in L^p\)). \(W^{1, p)}(\Omega)\) consists of all functions \(u \in L^{p)}(\Omega)\) such that \(| \nabla u | \in L^{p)}(\Omega)\). The authors also introduce the grand Gagliardo space, which corresponds to the fractional-order Sobolev space \(W^{1 - 1/p, p}(\mathbb R^n)\) replacing the \(L^p\) norms by \(L^{p)}\) norms. (With all the indices, the opportunities for typos increase. I~found one significant one: the authors introduce a grand Lorentz space and with their notation \(L^{p)}\) is \(L^{p),1}\) rather than the \(L^{p),0}\) given in the paper.) Among the applications is a proof that, for any \(\mu \in \mathcal{M}(\Omega)\), \(\operatorname{div} F = \mu\) has a vector-valued solution \(F \in L^{\frac{n}{n - 1})}(\Omega, \mathbb R^n)\) (which means that \(|F| \in L^{\frac{n}{n - 1})}(\Omega)\)) with a norm bound. If \(\mu\) is absolutely continuous with respect to the Lebesgue measure, then \(|F| \in L^{\frac{n}{n - 1})}_b(\Omega, \mathbb R^n)\). If \(h\) is a self homeomorphism of the circle \(S \subseteq \mathbb R^2\) and \(H\) is its Poisson extension to all of the unit disc, the authors prove that \(H \in W^{1, 2)}(D)\) and \( \| D H \|_{L^{2)} } \leq c_0\) (Iwaniec and Sbordone proved that actually \( \| D H \|_{L^{2)} } \leq 16 \sqrt{\pi}\)). They show that, if \(f:\Omega \to \mathbb R^n\) is a sense-preserving mapping whose differential is of the class \(L^n (\log L)^{-1}\), then its Jacobian determinant is locally integrable. The result does not hold under a natural weakening (essentially, \( |Df|^n\) in \(L^{1, \infty} (\log L)^{-1}\)). They give a regularity result for the Leray-Lions operator \[ \mathcal{L} u = \operatorname{div} \mathcal{A}(x, \nabla u). \] They conclude with an application of these ideas to the Lusin condition \(\mathcal{N}\): under what conditions on \(f\) does \( | E| = 0\) imply \(| f(E) | = 0\)? It is known that \(f\) Lipschitz suffices, and that \(f\) differentiable a.e.\ does not suffice. \textit{M. Marcus} and \textit{V. J. Mizel} [Bull. Am. Math. Soc. 79, 790--795 (1973; Zbl 0275.49041)] proved that it holds if \(f\) is a continuous map in \(W^{1,p}\) for \(p > n\), and it was shown that Lusin's condition may fail for a continuous mapping in \(W^{1,n}\). Finally, it was observed that if the Jacobian determinant of \(f\) is nonnegative and \(f\) is a homeomorphism, then \(| Df | \in L^{n)}_b\) suffices. The authors show here that the same result holds if \( | Df | \in L^{n)}\) and one of its components is in \(L^{n)}_b\).