Hermitian indefinite functions and Pontryagin spaces of entire functions. (Q1125289): Difference between revisions

In this paper the author establishes and investigates the relationship between hermitian indefinite functions on a finite or infinite interval on the one hand and de\,Branges spaces of entire functions and chains of entire matrix functions on the other hand. For \(0<a<\infty\) \((a=\infty)\) a function \(f:[-2a,2a]\to\mathbb{C}\) \((f: \mathbb{R}\to\mathbb{C})\) belongs to \({\mathcal P}_{k,a}\), if \(f\) is continuous, satisfies \(f(-t)=\overline{f(t)}\), and if the kernel \(f(t-s)\) has \(k\) negative squares for some \(k\in\mathbb{N}\cup\{0\}\). Let \(\mathcal B\) be a de\,Branges space (for short dB-space). The author proves some properties of the function \(f\in{\mathcal P}_{k,a}\) generated by a dB-space \(\mathcal B\) with mean type \(a\) and \(1\in{\mathcal B}\). Moreover, the model space of \(f\) is identified with a certain dB-subspace of \(\mathcal B\), and the dB-space associated to \(f\in{\mathcal P}_{k,a}\), \(a<\infty\) is studied. The author obtains the following statement: {Theorem 4.4} The relation \[ f(t)=[e^{i\frac t2z}, e^{-i\frac t2z}],\quad t\in[-2a,2a] \] establishes a bijective correspondence between all hermitian functions \(f\in{\mathcal P}_{k,a}\), \(k\in\mathbb{N}\), \(0<a<\infty\) with \(\Delta(f)<\infty\) and all dB-spaces \(\mathcal B\) with \(\text{ind}\_\mathcal B=k\), \(\text{ind}_0\mathcal B=\Delta(f)\), which satisfy \[ \mathcal B=\text{cls}\{e^{izt}:t\in(-a,a)\}\,. \] Finally, the connection of hermitian indefinite functions and maximal chains of matrices is investigated. {Theorem 5.9} Let \(f\in\mathcal P_{k,\infty}\) where \(k\in\mathbb{N}\cup\{0\}\), let \((W_t)_{t\in D^c}\) be the maximal chain of matrices associated to \(f\), and let \((\mathcal B_t)_{t\in M_{\text{reg}}}\) be the chain of dB-subspaces associated to \((W_t)_{t\in D^c}\). Denote by \(s_+\) the number \[ s_+=\sup\limits_{t\in D^c}\text{et}W_t=\sup\limits_{t\in M_{reg}}\text{mt}\mathcal{B}_t,\tag{1} \] and for \(s\in(0,\infty)\) by \(\nu(s)\) the number \(\nu(s)=\text{ind}\_\mathcal H(f^s)\). Then we have: (i) If \(s<s_+\), then \(\Delta(f^s)<\infty\) and \((2)\;\nu(s)+\Delta(f^s)\leq k \,. \) (ii) If \(s=s_+\) and the supremum in (1) is attained, then \(\Delta(f^s)<\infty\) and (2) holds. item[(iii)] If \(s=s_+\) and the supremum in (1) is not attained, then \(\Delta(f^s)>0\) and \(\nu(s)=k\). item[(iv)] If \(s>s_+\), then \(\Delta(f^s)=\infty\) and \(\nu(s)=k\). In the cases (i) and (ii) we have \(\Delta(f^s)=\text{ind}_0\mathcal{B}_{t\_(s)}\).