The sharp Riesz potential estimates in metric spaces (Q4545182)

The classical Sobolev inequality for \(C^{\infty}_0(\Omega)\)-functions \(u\) has a slightly sharper version\break \[ \int_0^{\infty} \alpha^{p-1} \operatorname{mes} (| u| > \alpha)^{1-p/n}\,d\alpha \leq C\tag \(*\) \] with \(\| \nabla u\|_p \leq 1\) and \(1 < p < n\) in terms of the distribution function of \(| u| \). There is a corresponding exponential result, due to \textit{N. S. Trudinger} [J. Math. Mech. 17, 473--483 (1967; Zbl 0163.36402)], in the boarderline case \(p = n\). The authors extend these results to metric measure spaces \((X,d,\mu)\) where \(\mu\) is a doubling measure and satisfies a lower estimate \(\mu (B(x,r)) \geq cr^n\); it is enough to assume that \(d\) is a quasimetric, see \textit{J.\ Heinonen} [Lectures on Analysis on Metric Spaces, Springer-Verlag (2001; Zbl 0985.46008)]. Inequality \((*)\) is replaced by \(\int_0^{\infty} \alpha^{p-1}\mu (G_{\alpha})^{1-p/n}d\alpha \leq C\) where \(\| g\|_p \leq 1\) , \(G_{\alpha}=\{x \in X: (Ig)(x) > \alpha\}\) and \(Ig\) refers to a Riesz potential of the form \(\int_0^r \int_{B(x,t)} g\,dy\,dt\), \(r = \text{diam}(X)/2\). The proof makes use of the Hardy-Littlewood maximal function; the main difficulty is that there is no upper bound for \(\mu (B(x,r))\).