Marchenko-Ostrovski mappings for periodic Jacobi matrices (Q1033744)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Marchenko-Ostrovski mappings for periodic Jacobi matrices
scientific article

    Statements

    Marchenko-Ostrovski mappings for periodic Jacobi matrices (English)
    0 references
    10 November 2009
    0 references
    Let \(L\) be a Jacobi matrix defined on \(l^{2} (\mathbb{Z})\) by \[ (L y)_{n} = a_{n - 1} y_{n} + a_{n} y_{n + 1} + b_{n} y_{n}, \quad n \in \mathbb{Z}, \] where \(a_{n} = e^{x_{n}} > 0\), \(x_{n}, b_{n} \in \mathbb{R}\) such that \((x_{n})\) and \((b_{n})\) are \(N\)-periodic sequences. Let \(x = (x_{n})_{1}^{N}\), \(b = (b_{n})_{1}^{N},\) \(p = (x,b) \in \mathcal{H}^{2}\), where \(\mathcal{H} = \{b \in \mathbb{R}^{n} : \sum_{1}^{N} b_{n} = 0 \}\). Denote by \(\varphi = (\varphi_{n} (\lambda,p))_{n \in \mathbb{Z}}\) and \(\vartheta = (\vartheta_{n} (\lambda, p))_{n \in \mathbb{Z}}\) the fundamental solutions of the equation \(L y = \lambda y, (\lambda, n) \in \mathbb{C} \times \mathbb{Z},\) with the initial conditions \(\varphi_{0} \equiv \vartheta_{1} = 0\) and \(\varphi_{1} \equiv \vartheta_{0} = 1\). The function \(\Delta (\lambda,p) = \varphi_{N + 1} (\lambda,p) + \vartheta_{N} (\lambda, p)\) is referred to as the Lyapunov function for the operator \(L\). The spectrum of \(L\) is absolutely continuous and consists of \(N\) intervals \(\sigma_{n} = [\lambda_{n - 1}^{+}, \lambda_{n}^{-}]\), \(n = 1,\dots,N,\) where \(\lambda_{N}^{+} = \lambda_{0}^{+} < \lambda_{1}^{-} \leq \lambda_{1}^{+} < \cdots < \lambda_{N - 1}^{-} \leq \lambda_{N - 1}^{+} < \lambda_{N}^{-}\). For each \(n = 1,\dots,N\), there is a unique value \(\lambda_{n} \in [\lambda_{n}^{-}, \lambda_{n}^{+}]\) such that \(\Delta^{'} (\lambda_{n}, P) = 0\), \(\Delta^{''} (\lambda_{n}, P) \neq 0\) and \((- 1)^{S_{n}} \Delta (\lambda_{n}, P) \geq 2,\) where \(S_{n} = N - n\) and \((^{'})\) designates \(\partial/\partial \lambda\). Let \(\mu_{n} = \mu_{n} (p)\) be the zeros of \(\vartheta_{N + 1} (\lambda,p)\). The Marchenko-Ostrovski mapping \(h : \mathcal{H}^{2} \longrightarrow \mathbb{R}^{2 N - 2}\) is defined by \(h (p) = (h_{n}(p))_{1}^{N - 1},\) where \(h_{n} = (h_{1 n}, h_{2 n}) \in \mathbb{R}^{2},\) \(h_{1 n} = \log [(-1)^{s_{n}} \vartheta_{N} (\mu_{n})]\), \(h_{2 n} = |\,|h_{n}|^{2} - h_{1 n}^{2}|^{1/2} \text{sign}(\lambda_{n} - \mu_{n})\) and \(2 \cosh |h_{n}| = (- 1)^{s_{n}} \Delta (\lambda_{n} (p), p)\), \(p \in \mathcal{H}^{2}\). The authors prove (as their main result) that the Marchenko-Ostrovski mapping \(h\) is a real-analytic isomorphism between \(\mathcal{H}^{2}\) and \(\mathbb{R}^{2 N - 2}\) and, moreover, the estimates \[ \frac{1}{4} \,e^{2 h_{+}} < c^{2} < \sum_{1}^{N} (b_{n}^{2} + 2 a_{n}^{2}) < 4 N c^{2} < 32 N c^{2 h_{+}} \] hold true, where \(c = 1/2 (\lambda_{N}^{-} - \lambda_{0}^{+})\) and \(h_{+} = \max |h_{n}|\).
    0 references
    0 references
    0 references
    0 references
    0 references
    Jacobi matrix
    0 references
    Lyapunov function
    0 references
    spectrum
    0 references
    Marchenko-Ostrovski mapping
    0 references
    0 references
    0 references
    0 references
    0 references