Hereditary completeness for systems of exponentials and reproducing kernels (Q1941160)

A system of vectors \(\{x_n\}_{n\in N} \) in a separable Hilbert \(H\) space is \textit{exact} if it is both complete and minimal. Given an exact system \(\{x_n\} \), let \(\{x_n'\}\) denote its (unique) biorthogonal system, i.e., \((x_m,x_n') = \delta_{mn}\). The system \(\{x_n\}_{n\in N}\) is hereditary complete if \(x\in \overline{\operatorname{Span}}\{(x_n,x_n') x_n\}\) for every \(x\in H\). It is known [\textit{A. S. Markus}, Izv. Akad. Nauk SSSR, Ser. Mat. 34, 662--688 (1970; Zbl 0195.13702); translation in Math. USSR, Izv. 4, 670--696 (1970; Zbl 0232.47001)] that hereditary completeness is equivalent to the following: for any partition \( N= N_1 \cup N_2,\, N_1\cap N_2 = \phi\), the system \(\{x_n\}_{n\in N_1} \cup \{x_n'\}_{n\in N_2}\) is complete in \(H\). \smallskip In the paper the authors consider exponential systems in \(L^2(-a,a),\, a>0\). For \(\Lambda = \{\lambda_n\}\subset\mathbb C\), let \(e_\lambda(t) = \exp(i\lambda t)\). It was shown by \textit{R. M. Young} [Proc. Am. Math. Soc. 83, 537--540 (1981; Zbl 0484.42009)] that, in contrast to the general situation, for any exact system of exponentials \(\{e_\lambda\}_{\lambda \in\Lambda}\), its biorthogonal system \(\{e^\prime_\lambda\}\) is complete. In this paper, the authors prove that any exact system of exponentials in \(L^2(-a,a)\) is hereditary complete up to a one-dimensional defect. Specifically, for any partition \(N=N_1 \cup N_2\) of the index set, the orthogonal complement to the system \[ \left\{e^{i\lambda_n t}\right\}_{n\in N_1} \cup \left\{e_n^\prime\right\}_{n\in N_2} \] is at most one-dimensional. It is also proved that this one-dimensional defect is possible, and thus there exist non-hereditary complete exponential systems. Analogous results are also obtained for systems of reproducing kernels in de Branges spaces, and the authors construct non-hereditary complete systems of reproducing kernels for a wide class of de Branges spaces, thus answering a question posed by N. Nikolski.