Discrete length-volume inequalities and lower volume bounds in metric spaces (Q269877)

The author establishes a discrete analog of Derrick's inequality for weighted open covers of the \(n\)-cube by generalizing his previous work [J. Differ. Geom. 100, No. 2, 349--388 (2015; Zbl 1328.53052)]. It is given a Riemannian metric \(g\) on the \(n\)-dimensional unit cube \([0, 1]^n\), and for each open cover \(\mathcal{U}=\{U_i, i\in I\}\) of \([0, 1]^n\) a weight function \(\omega:I\to [0, \infty)\), where \(\omega(i)\) is the weight of \(U_i\). The discrete distance of two elements \(A,B \in[0, 1]^n\) is defined by the infimum of the lengths of connecting chains \(U_{i_j}\), where \(A\in U_{i_1}\) and \(B\in U_{i_m}\) \[ \mathrm{dist}_{\omega}(A,B)=\inf\{\sum_j \omega(i_j): U_{i_1}, \ldots, U_{i_m} \text{ chain connecting \(A\) and \(B\)} \}. \] In the main Theorem 1.4 the author proves that if \(d_k=\mathrm{dist}_{\omega}(F_k, F'_k)\), \(1\leq k\leq n\), denotes the distance of pairs of opposite codimension-one faces \(F_k,F'_k\) of \([0,1]^n\), then \[ \sum_{i\in I} \omega(i)^n\geq d_1\ldots d_n. \] Theorem 1.5 states a generalization of Theorem 1.4 by taking possibly different weights \(\omega_k\) for each \(k\) and \(d_k\) w.r.t. \(\omega_k\) obtaining \(\sum_i(\prod_{k=1}^n \omega_k(i))\geq d_1\ldots d_k\). The proof is inspired on ideas on squares tilings, that associates to each triangulation of a quadrilateral \(Q\) a tiling of a rectangle by squares with a suitable correspondence of edges, vertices and corners, where the size of the squares are solutions of a discrete version of a extermal problem for length in conformal function theory. Theorem 1.5 is first shown to be true in case no \(U_i\) intersects simultaneously \(F_k, F'_k\) for some \(k\) (already assuming the cover to be a finite one) and from the fact that \(d_k\leq d_k(i)+\omega_k(i)\) where \(d_k(i)\) is the distance \( \mathrm{dist}_{\omega_k}\) of \(F_k\) to \(U_i\), the author uses this \(d_k(i)\) to built a \(n\)-rectangle \(R_i\) with side length \(\omega_k(i)\) proxy to \(U_i\) with similar combinatorial properties. The general case also follows by slightly modifying the cover and the weights reducing the proof to the first case. Using a partition of the unity subordinated to \(\mathcal{U}\), the author builds a continuous map \(\phi\) from \([0,1]^n\) to the nerve \(Ner(\mathcal{U})\) of the cover, that is a complex, the union of all the \(m\)-dimensional simplex in \(\mathbb{R}^M\) (of canonic basis \(e_1, \ldots, e_M\) where \(m\leq M\)) defined by the convex hull generated by \(e_{i_j}\), \(j=i_0,\ldots, i_m\) whenever \(\bigcap_{j=i_0,\ldots, i_m}U_j\) is not empty. Then the nerve \(Ner(\mathcal{U})\) is mapped into the union of rectangles \(\bigcup_i R_i\) by using a barycentric subdivision of the nerve and a map sending each vertex to a well chosen point of the corresponding intersection of the rectangles according a rule, and then extended to be affine on each simplex, and finally shown the image contains \(\prod_k[0, d_k]\), using some topological result based on a Brouwer fixed point theorem. In Corollary 4.2 a sharp lower bound for the \(n\)-dimensional Hausdorff content of the image of the \(n\)-cube by any continuous function \(g:[0,1]^n\to X\) into a (pseudo)metric space \((X,d)\) is derived, namely \[ \mathcal{H}_n^{\infty} (g( [0,1]^n))\geq \prod_{k=1}^n \mathrm{dist}(g(F_k),g(F'_k)), \] with equality if and only if \(d=l_{\infty}\) and \(g\) is the identity map. As a consequence the author concludes this inequality is valid for the Hausdorff content but not to the Hausdorff measure, giving an answer to a question of \textit{Yu. D. Burago} and \textit{V. A. Zalgaller} [Geometric inequalities. Transl. from the Russian by A. B. Sossinsky. Berlin etc.: Springer-Verlag (1988; Zbl 0633.53002)] for a pseudometric \(\rho\) on \([0,1]^n\) defining a weaker topology than the Euclidean one. Finally, metric spaces \((X,d)\) with fat \(n\)-cubes are introduced and some connectivity conditions are defined that ensure Euclidean-type lower volume bounds, that is \(\mathcal{H}^{\infty}_n(B(x,r))\gtrsim r^n\), for \(r\leq \mathrm{diam}(X)\). Some further conditions of discrete nature are stated that guarantee the existence of a metric \(\rho\) on \(X\) satisfying \(\rho\approx d^{1/\epsilon}\). As an application, conditions are stated for a homeomorphic \(X\) to the 2-sphere to be a quasi-sphere.