A new approach to the theory of linear problems for systems of differential partial equations. III (Q1589126): Difference between revisions

For Parts I and II, cf. Russ. Math. 43, No. 1, 22--32 (1999); translation from Izv. Vyssh. Uchebn. Zaved., Mat. 1999, No. 1, 25--35 (1999; Zbl 1049.35071) and Russ. Math. 43, No. 7, 27--38 (1999); translation from Izv. Vyssh. Uchebn. Zaved., Mat. 1999, No. 7, 30--41 (1999; Zbl 1049.35072)]. With the notations of the former parts, here the authors consider the so-called \((b-a)\)-periodic problem in which the basic space \(H=L_m^2(a,b)\) and the weight function \(\phi=\{e(j)\exp (2\pi ikcdot(x-a)/(b-a)), j=1,\dots,m, k\in\mathbb Z^n\}\). Assume that \(C_{\alpha}(x)\) as a multiplication operator is a measurable \(m\times m\) matrix. Then Theorem 5.2. Assume that \(C_{\alpha}(x)\in \mathbb C^{\{\alpha\}}_{m,m}(b-a)\), and \(q(z)=\sum_{\alpha \in\Phi}(iz)^{\alpha}\) is a scalar polynomial satisfying some conditions; then the following two statements are equivalent: (1) there exist \(t\in\mathbb R\), \(d_1(t), d_2>0, \alpha_0\in\mathbb C\), such that for any \(x\in [a,b], z\in\mathbb C^m, k\in Z^n\), we have \[ \bigg|\bigg(\alpha_0tE+\sum_{\alpha\in\Phi}\bigg( \frac{2\pi ik}{b-a}\bigg)^{\alpha}\bigg)z\bigg|^2\geq |z|^2(d_1(t)+d_2|q(k)^2|)\tag{5.17} \] in which \(d_1(t)\rightarrow +\infty\) when \(|t|\rightarrow +\infty\); (2) there exist \(t\in\mathbb R\), \(d_3(t), d_4, d_5>0, \alpha_0\in C\), such that for any \(h(x)\in C_m^{\infty}\) we have \[ \begin{multlined} \|(P(x,\partial/\partial x)+\alpha_0tE)h(x)\|\geq\\ d_3(t)\|h(x)\|+ d_4\|q(-i\partial/\partial x)h(x)\|\geq\\d_3(t) \|h(x)\|+d_5\sum_{\alpha \in \Phi} \sum_{\beta < \alpha} \|h^{\beta}(x)\|\end{multlined}\tag{5.18} \] in which \(d_3(t)\rightarrow +\infty\) when \(|t| \rightarrow +\infty\). Theorem 5.3: Under the assumptions described in Theorem 5.2, (5.17) and given sufficiently small \(\epsilon\), then (1) the \(\phi\)-problem for the equation (7) \((P(x,\partial/\partial x)-\lambda E)u=f(x)\) (see Part I [loc. cit.]) has only point spectrum; (2) a \(\phi\)-solution for (7) (if it exists) belongs to \(W^{\widetilde {\Phi},2}_m (b-a)\), where \(\widetilde {\Phi}=\bigcup_{\alpha \in \Phi}\{\gamma\colon 0 \leq \gamma \leq \alpha\}\); (3) if besides, for all \(\alpha \in \Phi\) we have \(C_{\alpha}(x) \in C^{\widetilde {\gamma}}_{m,m}(b-a), f(x) \in W^{\{\widetilde{\gamma}\},2}_m (b-a)\), and the multi-index \(\gamma\), which is independent of \(\alpha\), satisfies the condition \((\text{sgn}\, \widetilde{\gamma}_1,\dots,\text{sgn}\,\widetilde{\gamma}_n)\leq \beta\) for some \(\beta \in \Phi\), then the \(\phi\)-solution (if it exists) belongs to \(W^{\widetilde{\Phi}+ \widetilde{\gamma},2}_m (b-a)\), where \(\widetilde{\Phi}+\widetilde{\gamma} =\bigcup _{\alpha \in \widetilde {\Phi}}\{\alpha + \widetilde{\gamma}\}\). In Section 3 the authors consider the problem of when the functions \(\phi_p, p\in\mathbb N\), are restrictions of functions from the class \(W^{\Phi,2}_m (b-a)\).