Hardy's inequality and fractal measures (Q1196575)

The paper deals with the following generalized Hardy inequality \[ \int^ \infty_{-\infty}{| f(x)| d\mu_ \alpha(x)\over \sigma_ \alpha(E,x)}\leq C \liminf_{L\to\infty} L^{\alpha-1}\int^ L_{-L} | \widehat{fd\mu_ \alpha}(x)| dx \] for \(0\leq\alpha\leq 1\), where \(C\) is a constant that may depend on \(E\subseteq\mathbb{R}^ 1\) but not on \(f\). Further \(\mu_ \alpha\) is the \(\alpha\)-dimensional Hausdorff measure including the special cases \(\alpha=0\) for some discrete measure and \(\alpha=1\) for Lebesgue measure. Finally, let \(\sigma_ \alpha(E,x)=\mu_ \alpha(E\cap(-\infty,x])\) and \(\widehat{fd\mu_ \alpha}\) the Fourier transform. This inequality holds provided that \(E\) is \(\alpha\)-coherent. This notion is mainly defined in terms of the upper density of points in \(E\) w.r.t. to \(\mu_ \alpha\). Self-similar sets \(E\) in \(\mathbb{R}\) with equal contraction ratios are examples for \(\alpha\)- coherent sets, where \(\alpha\) is the Hausdorff dimension of \(E\). For \(0<\alpha<1\) the property to be \(\alpha\)-coherent can be replaced by the simpler conditions that the set \(E\) satisfies \(\mu_ \alpha(E)<\infty\) and \(E\) is quasi-regular (the lower \(\alpha\)-dimensional density w.r.t. \(\mu_ \alpha\) is bounded away from zero \(\mu_ \alpha\) a.e. on \(E\)). Nevertheless, there are \(\alpha\)-coherent sets which are not quasi- regular. The authors show also that the above inequality fails without restrictions such as to be \(\alpha\)-coherent in the general case.