Mean quadratic variations and Fourier asymptotics of self-similar measures (Q1261146)

Let \((S_ j)^ m_{j=1}\) be a family of contractive similarities on \(\mathbb{R}^ n\) and \(\mu\) the unique Borel probability measure on \(\mathbb{R}^ n\) such that \(\mu=\sum^ m_{j=1}a_ j \mu \circ S^{-1}_ j\), where the \(a_ j\)s are positive weights summing up to 1. The similarities satisfy the open set condition iff there is a bounded open set \(U\) such that \(S_ jU \subseteq U\) for all \(j\), and \(S_ jU \cap S_ iU \neq \emptyset\) for \(i \neq j\). Strichartz has investigated the Fourier asymptotic behaviour of self- similar measures \(\mu\) defined as above. He has shown that \[ {1\over T^{n-\beta}}\int_{| x | \leq T} | \hat \mu(x) |^ 2dx \] behaves either like a constant or like a multiplicative periodic function for \(T \to \infty\) depending on \(\{-\ln \rho_ j;j=1,\ldots,m\}\) is nonarithmetic or arithmetic, where \(\rho_ j\) is the ratio of \(S_ j\), provided that the family of similarities satisfies the strong open set condition \(S_ j(\overline U) \cap S_ i(\overline U)=\emptyset\) for \(i \neq j\). The authors prove that for a selfsimilar measure under the open set condition either \(\mu(U)=1\) or \(\mu(U)=0\). They show that the strong open set condition allows to choose some \(U\) with \(\mu(U)=1\). This implies that the measure of the boundary \(\partial U\) is zero. The main result is then that the strong open set condition can be replaced by the simple open set condition and \(\mu(\partial U)=0\) in Strichartz's result, but they use mainly the so called \(\alpha\)-mean quadratic variation instead of the above term for the Fourier transform of \(\mu\). Both are asymptotically equal.