On Nyman, Beurling and Baez-Duarte's Hilbert space reformulation of the Riemann hypothesis (Q998214): Difference between revisions

The author's abstract reads as follows: ``There has been a surge of interest of late in an old result of Nyman and Beurling giving a Hilbert space formulation of the Riemann hypothesis. Many authors have contributed to this circle of ideas, culminating in a beautiful refinement due to Baez-Duarte. The purpose of this little survey is to dis-entangle the resulting web of complications, and reveal the essential simplicity of the main results.'' So the author formulates (and reproves) many statements [expressed in the language of functional analysis] equivalent to the truth of the Riemann hypothesis (and so elucidating work of [\textit{B. Nyman}, On the one-dimensional translation group and semi-group in certain function spaces. (Diss.) Uppsala: Appelbergs Boktryckeri AB, 55 p. (1950; Zbl 0037.35401), \textit{A. Beur\-ling}, Proc. Natl. Acad. Sci. 41, 312--314 (1955; Zbl 0065.30303), and \textit{L. Baez-Duarte}, Atti Acad. Naz. Lincei 14, 5--11 (2003; Zbl 1097.11041)]). Denote by \(\mathcal H\) the Hilbert space of sequences \(a= \{a_n\in\mathbb C\); \(n=1,2,\dots\}\) with inner product \(\langle a,b\rangle= \sum_1^\infty \frac1{n(n+1)} \cdot a_n \overline{b_n}\). The Hardy space \(H^2(\Omega)\), where \(\Omega\) is the open half-plane \(\sigma > \frac12\), is the Hilbert space of all holomorphic functions \(F\) on \(\Omega\) with norm \[ \| F\|^2 = \sup_{\sigma > \frac12}\; \frac1{2\pi} \int_{-\infty}^\infty |F(\sigma + it)|^2 \,dt < \infty. \] The functions \(E(s) = \frac1s\), \(F_\lambda(s) = (\lambda^s- \lambda) \cdot \frac{\zeta(s)}s\) (where \(0\leq\lambda \leq 1)\), and \(G_\ell= F_{1/\ell}\), \(\ell= 1,2,\dots\), are in \(H^2(\Omega)\). The functions \(f_\lambda(x)= \{ \frac{\lambda}x\}- \lambda \{ \frac1x\}\), \(x\in (0,1]\), and the constant function \textbf{1} are in the Hilbert-space \(L^2((0,1])\) (note: \(\{\frac \lambda x\}\) is the fractional part of \(\frac \lambda x\)); the same is true for the functions \(g_\ell(x)= \{\frac1{\ell x}\}- \frac1\ell \{\frac1x\}\), \(0<x\leq 1\), \(\ell\in\mathbb N\). The functions \(g_\ell= f_{1/\ell}\) are in the closed subspace \(\mathcal M \subset L^2((0,1])\) of all functions constant (a.e.) on each of the subintervals \((\frac1{n+1}, \frac1n]\), \(n\in \mathbb N\). Then the following statements are equivalent: (1) The Riemann hypothesis for \(\zeta(s)\). (2) In \(\mathcal H\) the vector \(\gamma= \{1,1,1,\dots\}\) belongs to the closed linear span of \(\{\gamma_\ell\); \(\ell=1,2,\dots\}\), where \(\gamma_\ell= \{\{\frac1\ell\}, \{\frac2\ell\}, \dots \}\). (3) \(\{\gamma_\ell\); \(\ell\in\mathbb N\}\) is complete in \(\mathcal H\). (4) \(E\) is in the closed linear span of \(\{ G_\ell\); \(\ell = 1,2,\dots\}\). (5) \(E\) is in the closed linear span of \(\{F_\lambda\); \(0 \leq \lambda \leq 1\}\). (6) In \(L^2((0,1])\), the constant function \textbf{1} is in the closed linear span of \(\{ f_\lambda\); \(= 0\leq \lambda \leq 1\}\). (7) The set \(\{ f_\lambda\); \(0\leq \lambda \leq 1\}\) is complete in \(L^2((0,1])\). (8) \textbf{1} is in the closed linear span of \(\{g_\ell\); \(\ell =1,2,\dots\}\). (9) \(\{g_\ell\); \(\ell =1,2,\dots\}\) is a complete set in \(\mathcal M\).