On Besicovitch's ½-Problem (Q4024896)

This nice paper deals with the connection between densities and rectifiability. The authors consider the problem in general metric spaces \(M\) and define \(\sigma_ n(M)\) as the smallest number having the property that every subset \(E\subset M\) of finite \(n\)-dimensional Hausdorff measure \({\mathcal H}^ n\) satisfying at \({\mathcal H}^ n\)-almost all of its points \(\underline{D}_ n(E,x)>\sigma_ n(M)\) is rectifiable, where \(\underline{D}_ n(E,x)\) denotes the lower density at \(x\) with respect to \({\mathcal H}^ n\). The main goal is to improve the upper bound \({3\over 4}\) for \(\sigma_ 1(M)\) which was found by Besicovitch for the plane and who has conjectured that \({1\over 2}\) is the best value. The authors show in a tricky proof that \[ \sigma_ 1(M)\leq{2+\sqrt{46}\over 12}<{59\over 80}<{3\over 4}. \] They introduce the notion of the Besicovitch pair condition which plays the same key role as the double circle pair in Besicovitch's proof.