Removable sets for Orlicz-Sobolev spaces (Q907314): Difference between revisions

In this nicely illustrated paper, the author considers the problem of removability in Orlicz-Sobolev spaces. The Orlicz-Sobolev spaces \(W^{1,\Psi}\) considered are those where \(\Psi(t)\) is of the form \(t^p\log^\lambda(e+t)\). The classical case was treated in [\textit{P. Koskela}, Ark. Mat. 37, No. 2, 291--304 (1999; Zbl 1070.46502)]. For dealing with the Orlicz case, new methods are applied. The author introduces an adapted generalization of porosity; there are six different cases according to which values \(p\) and \(\lambda\) take. The author proves that if a set \(E\subset \mathbb{R}^{n-1}\) is \((p,\lambda)\)-porous, then it is removable for \(W^{1,\Psi}\), that is \(W^{1,\Psi}(\mathbb{R}^n\setminus E)=W^{1,\Psi}(\mathbb{R}^n)\) as sets. He also verifies the sharpness of his result. A corollary is the existence of spaces that satisfy certain \(\Psi\)-Poincaré inequalities, but no stronger ones of the same type. (Note that we are in the non-complete setting, the complete setting is different: see [\textit{N. DeJarnette}, J. Math. Anal. Appl. 423, No. 1, 358--376 (2015; Zbl 1333.46034)] for details.) First, the two-dimensional case is looked at. In this case, line segments do not have \(1\)-capacity zero, allowing for an easier argument based on the study of directional limits in points on lines. The idea of the proof in the higher-dimensional case is based on continuity on almost all lines. Let \(E\) be the porous set to which we want to extend a Sobolev function. First, a \(W^{1,1}\)-extension is constructed. Then the points in \(E\) are considered where certain one-sided limits do not agree but the porosity condition holds. Note that \[ \int_{B(x,r)} \Psi|\nabla u(z)|)\, dz \] tends to zero \(\mathcal{H}^{n-1}\)-almost everywhere as \(r\) approaches zero. The idea is to show that this fails in the selected points. This implies that \(E\) must have \(\mathcal{H}^{n-1}\)-measure zero.