Lower bounds for codimension-1 measure in metric manifolds (Q1989541)

The author shows a lower bound for the codimension one Hausdorff measure of separating sets on metric closed manifolds. This lower bound can be seen as a quantitative topological isoperimetric inequality. Two necessary conditions are assumed on the metric closed manifolds: the \(D\)-doubling property, i.e., every ball of radius \(r\) can be covered by at least \(D\) balls of radius \(r/2\), and the \(L\)-linear contractibility, i.e., any ball of radius \(r\) can be contracted to a point in the ball of radius \(L\cdot r\). These conditions, prevent the apparition of cusps, thin necks and other possible pathologies. The in-radius of a set \(S\), called \(\operatorname{in-rad}(S)\), is the supremum of the radii \(r>0\) such that some ball of radius \(r\) are included in \(S\). Observe that this implies that the in-radius is trivial for sets with empty interior. Under these hypotheses, given a Borel set \(E\subset M\), it is shown that \(\mathcal{H}_{n-1}(\partial E)\geq C\cdot(\min\{\operatorname{in-rad}(E),\operatorname{in-rad}(M\setminus E)\})^{n-1}\). Where \(\mathcal{H}_{n-1}\) is the \(n-1\) Hausdorff measure and \(C\) is a constant that does not depend on \(E\). The paper is well written and aimed to those researchers interested in quantitative topology, measure theory and isoperimetric inequalities for Borel sets. It is remarkable that these conditions are invariants by quasisymmetry, this is not the case for classical isoperimetric inequalities. This justifies the fact that the inequality is not sharp for non-euclidean geometries. A corollary is that the Euclidean isoperimetric inequality is, in some sense, universal. The paper is organized as follows: In the first section the author motivates the main result and gives the basic definitions. The second section is devoted to show the following result: if \(S=\partial E\) is a subset of a closed metric \(n\)-manifold \(M\) such that \(M\setminus S\) consists of more than one connected component then \(\check{H}^{n-1}(S,\mathbb{Z}_2)\) is nontrivial. Morover, if \(h:S\to M\) is homotopic to the inclusion then the induced map in cohomology is also nontrivial. Here it is important the fact that \(M\) has no boundary. The third section is a sort of projection lemma. Given a simplex \(\Delta_n\) of dimension \(n\) and a Borel set \(S\) with finite \(k\)-dimensional Haussdorff measure (\(k<n\)) there exists a continuous map \(p:\Delta_n\to\Delta_n\), fixing the boundary and do that \(p(S)\subset \partial \Delta_n\) and its Hausdorff measure is of the same order as the Hausdorff measure of \(S\). This can be seen as a special case of Federer-Flemming theorem [\textit{H. Federer} and \textit{H. W. Flemming}, Ann. Math. (2) 72, 458--520 (1960; Zbl 0187.31301] and a simplest version of Proposition 3.1 in [\textit{G. David} and \textit{S. Semmes}, Mem. Am. Math. Soc. 144, 687, 132 p. (2000; Zbl 0966.49024]. In the fourth section, the metric manifold is approximated by a simplicial complex. It is shown that a separating set \(S\) with sufficiently small \((n-1)\)-dimensional Hausdorff measure can be projected into the \((n-2)\)-dimensional skeleton of the simplicial complex (following the guidelines of the third section). It is also shown that for a sufficiently good approximated simplicial complex the previous projection can be mapped to \(M\) and the resulting transformation \(h:M\to M\) is homotopic to the identity in \(M\) by a homotopy that move points less than a controlled distance. Therefore \(h(S)\) still separates \(M\) (by cohomological arguments), however this is not possible since the \((n-2)\)-dimensional skeleton cannot separate points and this will prevent \(h(S)\) to separate \(M\) as the homotopy is too small.