Nonselfadjoint spectral problems for linear pencils \(N- \lambda P\) of ordinary differential operators with \(\lambda\)-linear boundary conditions: Completeness results (Q1814824)

Boundary eigenvalue problems of the form \[ N(y) - \lambda P(y) = 0,\;U_j(y) -\lambda V_j(y) = 0,\;j=1, \dots, n \tag{1} \] are considered. Here \(N\) and \(P\) are ordinary linear differential operators on \([0,1]\) of orders \(n>p \geq 0\), and \(U_j(y)\), \(V_j (y)\) are nontrivial linear combinations of \(y\) and its derivatives of order up to and including \(l_j\leq n-1\) and \(h_j\leq p-1\), respectively, at the endpoints 0 and 1. Due to the occurrence of \(\lambda\) in the boundary conditions, (1) is written as \({\mathbf K} y- \lambda {\mathbf H} y=0\), where \({\mathbf K}\) and \({\mathbf H}\) are operators from \(L_2(0,1)\) to \(L_2(0,1) \times \mathbb{C}^n\). Assuming that (1) has a nonempty resolvent set one may assume that \({\mathbf K}\) is invertible, and also the compact operator \({\mathbf A} = {\mathbf H} {\mathbf K}^{-1}\) is considered. If (1) is normal of order 1, then it is shown in Proposition 4.1 that the maximal length of the Jordan chains of \({\mathbf A}^*\) does not exceed \(r(l)+1\), where \(r(l) = [{n-p+ l+ \max \{0,p-1 - \kappa_{p+1}\} \over n-p}]\) and \(\kappa_j = \max \{l_j,n - p+h_j\}\) with \(\kappa_1 \leq \kappa_2 \leq \cdots \leq \kappa_n\). The number \(r(l)\) plays a crucial role in the main result of this paper, Theorem 4.3, which states that the system of eigenfunctions and associated functions of (1) is complete in the spaces \({\mathcal W}^k_{{\mathcal U}, r(l)+1}\) for \(k=0, \dots,n\), where these spaces are defined as follows: For a suitably chosen basis \(\{\varphi_j\}^q_{j=1}\) of the null space of \((A^*)^r\), \(r\) a positive integer, define the linear forms \[ U_j(y): = \bigl({\mathbf K} (y), \varphi_j \bigr),\;y \in W^n_2(0,1),\;j=1, \dots, q. \] Theses forms can be written as \(U_j(y): = \widehat U_j(y)+ \int^1_0 y(x) \overline {u_j (x)} dx\), where \(\widehat U_j (y)\) is a two-point boundary condition. The order of \(U_j\) is defined as the order of \(\widehat U_j\) with \(\text{ord} U_j= -1\), if \(\widehat U_j=0\). Normalizing the system \(\{U_j\}^q_1\) an equivalent system of linear forms \(\{{\mathcal U}_j\}_1^m\) is obtained, and one defines \[ {\mathcal W}^k_U: =\bigl\{y \in W^k_2 (0,1) : {\mathcal U}_j(y) = 0 \text{ if ord} {\mathcal U_j} \leq k-1\}, \quad k = 0, \dots,n. \] In Section 5 several examples are considered, including Petterson-König's rod.