Discrete approximations of metric measure spaces of controlled geometry (Q2352176)

A metric measure space \((X,d,\mu)\) is said to be doubling if every ball \(B(x,r)\) has positive measure and its measure is comparable to \(B(x,2r)\). The space \((X,d,\mu)\) is said to support a \((1,p)\)-Poincaré inequality if every ball has positive and finite measure and if there exist constants \(C>0\) and \(\lambda>0\) such that \[ \int_B |u-u_B| \, d\mu \leq Cr \left(\int_{\lambda B}g^p \, d\mu\right)^{1/p}, \] for every measurable function \(u:X\to\mathbb{R}\) that is integrable on balls and for every upper gradient \(g\) of \(u\). The classification of metric measure spaces that support a \((1,p)\)-Poincaré inequality is one of the most interesting open questions in modern Analysis on metric spaces. The authors of the paper under review present necessary and sufficient conditions for a doubling metric measure space to carry a \((1,p)\)-Poincaré inequality in terms of a discretization of the metric space and a discrete version of the Poincaré inequality on graphs [\textit{I. Holopainen} and \textit{P. M. Soardi}, Ann. Acad. Sci. Fenn., Math. 22, No. 1, 205--226 (1997; Zbl 0874.31008)]. In particular, the authors show that if \((X,d,\mu)\) is doubling and supports a \((1,p)\)-Poincaré inequality then any discretized space \((V,d_V,m)\) obtained by graphs of maximal \(\epsilon\)-separated sets has the same properties (Theorem 1.1). Conversely, if \((V_i,d_{V_i},m_i)\) is a sequence of nested graphs embedded in a doubling \((X,d,\mu)\) that are ``finer and finer approximations'' of the space, and are doubling and carry a \((1,p)\)-Poincaré inequality with uniform constants then \((X,d,\mu)\) carries a \((1,p)\)-Poincaré inequality (Theorem 1.2).