A Hilbert space of Dirichlet series and systems of dilated functions in \(L^ 2(0,1)\) (Q677109): Difference between revisions

The authors investigate the following two problems: For which functions \(\varphi\) is the system \((\varphi_n)_{n\in\mathbb{N}}\), \(\varphi_n(x):= \varphi(nx)\), a Riesz basis in \(L^2(0, 1)\) (i.e., an orthonormal basis with respect to an equivalent norm)? And for which functions \(\varphi\) is this system \((\varphi_n)_n\) a complete sequence in \(L^2(0, 1)\)? The second of these problems was stated by Beurling in his seminar in 1945, and the authors follow Beurling's suggestion to associate to the given function \(\varphi\), \(\varphi(x)= \sum^\infty_{n=1} a_n\sqrt 2\sin(n\pi x)\), the Dirichlet series \(S\varphi\), \(S\varphi(s)= \sum^\infty_{n= 1} a_nn^{-s}\). They solve completely the first of these problems by showing that \((\varphi_n)_n\) is a Riesz basis if and only if \(S\varphi\) defines an analytic function bounded away from zero and infinity in the (open) right half-plane. For the completeness problem, they prove that it is equivalent to the (rather hard) problem of describing the cyclic vectors in the Hardy space \(H^2(D^\infty)\) on the infinite-dimensional polydisk. (Note that a complete characterization of the cyclic vectors in \(H^2(D^N)\) is known only when \(N= 1\).) But the authors derive a fairly general sufficient condition for completeness and some interesting consequences. A major part of this article consists of a study of the Dirichlet series \(\sum^\infty_{k= 1} a_nn^{-s}\) with \(\sum_n|a_n|^2 <\) \(+\infty\). They form, in a natural way, a Hilbert space \(\mathcal H\) of analytic functions on the half-plane \(\text{Re }s>1/2\). A multiplier \(m\) on \(\mathcal H\) is an analytic function on \(\text{Re }s>1/2\) such that the function \(mf\) belongs to \(\mathcal H\) for each \(f\in{\mathcal H}\). The authors prove that the multipliers are precisely the bounded analytic functions in the right half-plane which can be represented as Dirichlet series, and that the system \((\varphi_n)_n\) is complete if and only if the Dirichlet series \(S\varphi\) is cyclic in \(\mathcal H\); i.e., the collection of functions \(fg\), where \(g\) is a finite Dirichlet series, is dense in \(\mathcal H\). Following an idea of Bohr, they find it ``both convenient and illuminating'' to use the infinite-dimensional polydisk for the study of \(\mathcal H\), which leads to power series in infinitely many variables (a concept studied already by Hilbert) and to the identification of \(\mathcal H\) as \(H^2(D^\infty)\). The article contains a wealth of material which cannot be described here in detail. There is some overlap with results in harmonic analysis due to Helson. Kronecker flows in ergodic theory appear. The kernel function of \(\mathcal H\) satisfies \(K_{\mathcal H}(z, w)= \zeta(z+\overline w)\), where \(\zeta\) is the Riemann zeta-function, and via the infinite-dimensional polydisk this is connected with results of \textit{B. J. Cole} and \textit{J. W. Gamelin} [Proc. Lond. Math. Soc. III. Ser. 53, 112-142 (1986; Zbl 0624.46032)]. A ``probabilistic Riemann hypothesis'' is proved: For almost all characters \(\chi\) on the positive integers, the Dirichlet series \(\zeta_\chi(s)= \sum^\infty_{n= 1}\chi(n)n^{-s}\) converges to a zero-free holomorphic function in \(\text{Re }s> 1/2\). To sum up: this is a very nice paper connecting several fields of mathematics.