On Fourier frame of absolutely continuous measures (Q647600)

Let \(L^2(K,d\mu)\) denote the \(L^2\)-space of a compactly supported probability measure \(\mu\) where \(K:= {\mathrm{supp}}\,\mu\). A sequence of complex exponentials \(\{e^{2\pi i\,\langle\lambda, \bullet\rangle}\,:\, \lambda\in \Lambda\}\) is said to form a Fourier frame of \(L^2(K,d\mu)\) if it is a frame for \(L^2(K,d\mu)\). In this case, \(\Lambda\) is called the frame spectrum. The author proves that for compactly supported absolutely continuous probability measures \(\mu\) on \(\mathbb{R}^n\) with \(d\mu = \varphi (x) dx\), where \(\varphi\in L^1(\mathbb{R}^n, dx)\) is compactly supported and \(dx\) denotes Lebesgue measure on \(\mathbb{R}^n\), \(L^2(K,d\mu)\) admits a Fourier frame if and only if there exist positive constants \(m\) and \(M\) such that \(m\leq \varphi (x) \leq M\), a.e. on \(K\). In addition, it is shown that if \(\mu\) is an equal weight absolutely continuous self-similar measure on \(\mathbb{R}\) and \(L^2(K,d\mu)\) admits a Fourier frame, then the density \(\varphi\) of \(\mu\) must be the characteristic function of the self-similar tile \(K\). As a consequence one obtains the result that for a.e. \(\lambda\in (\frac{1}{2}, 1)\), the \(L^2\)-space of \(\lambda\)-Bernoulli convolutions can not admit a Fourier frame.