Characterization of functions with zero traces via the distance function and Lorentz spaces (Q6074489)

The authors begin by recalling a result, nowadays classical, in Sobolev space theory: \(u\in W_0^{m,p}(\Omega)\) if and only if \(u/d^m\in L^p(\Omega)\) and \(|\nabla^m u|\in L^p(\Omega)\), where \(m\in {\mathbb N}\), \(p\in (1,\infty)\), \(\Omega\subset {\mathbb R}^N\) is sufficiently regular (for instance, it is a Lipschitz domain) and \(d\) is the distance function defined in \(\Omega\) as the distance from its boundary. Several comments about the existing literature concerning variants, extensions and improvements of this result are given, so that the reader is optimally introduced into the topic. In a first key theorem (Theorem 3.1) the authors prove that for \(p\in [1,\infty)\) and a general open set \(\Omega\subset {\mathbb R}^N\) (hence it can be an unbounded domain), if \(u\in W^{1,p}(\Omega)\) and \(u/d\in L_a^{p,\infty}(\Omega)\), then \(u\in W_0^{1,p}(\Omega)\). Here \(L_a^{p,\infty}(\Omega)\) denotes the collection of all functions \(f\) in the Lorentz space \(L^{p,\infty}(\Omega)\) having absolutely continuous quasinorm, i.e., \(\|f\chi_{E_k}\|_{L^{p,\infty}(\Omega)}\to 0\) for every sequence \(\{E_k\}_{k=1}^\infty\) satisfying \(E_k\to\emptyset\), where \(E_k\to\emptyset\) denotes the fact that \(\chi_{E_k}\to 0\) a.e. on \(\Omega\). Then, in Theorem~3.2, the same conclusion has been obtained in the case of bounded domains satisfying the condition \(W^{1,p}(\Omega)\cap W_0^{1,1}(\Omega)=W_0^{1,p}(\Omega)\), assuming \(u\in L^1(\Omega)\), \(|\nabla u|\in L^p(\Omega)\), and \(u/d\in L_a^{1,\infty}(\Omega)\). Both results are tools for the proof of the main result (Theorem~1.1) which reads as follows. Let \(p\in (1,\infty)\) and let \(\Omega\subset {\mathbb R}^N\) be a bounded domain (i.e., an open and connected set) obeying simultaneously the outer ball portion property (i.e., there exist positive constants \(b\) and \(r_0\) such that, for all \(x\in\partial\Omega\) and all \(r\in (0,r_0]\), one has \( {\lambda^N\left(B(x,r)\cap ({\mathbb R}^N\setminus\Omega)\right)}/{\lambda^N\left(B(x,r)\right)}\ge b \)), and the estimate \( I_\Omega(s)\ge Cs\) for \( s\in\left[0,{\lambda^N(\Omega)}/{2}\right] \), where \(I_\Omega\) is the isoperimetric function of \(\Omega\) and \(\lambda^N\) is the \(N\)-dimensional Lebesgue measure. Then \[ \frac{u}{d}\in L_a^{1,\infty}(\Omega) \quad \textrm{and}\quad |\nabla u|\in L^p(\Omega) \] if and only if \[ u\in W_0^{1,p}(\Omega). \] In Example 6.1, considering \(\Omega=(0,1)^N\), \(N\in {\mathbb N}\), and \(u(x)=1\) for each \(x\in\Omega\), the authors show also that \(L_a^{1,\infty}(\Omega)\) cannot be replaced by \(L^{1,\infty}(\Omega)\). The paper ends with an Appendix devoted to background material, of independent interest, about spaces of functions with absolutely continuous (quasi)norms, namely their equivalent definitions and their relations to Lorentz spaces.