Real analytic parameter dependence of solutions of differential equations over Roumieu classes (Q2430136): Difference between revisions

The paper is concerned with partial differential operators \(P(D)\) on Roumieu type quasianalytic or non quasianalytic classes \(\mathcal{E}_{\{\omega\}}(\Omega)\) of ultradifferentiable functions depending on a real analytic parameter. Specifically, the author studies the problem if for every family \(f_\lambda\) of ultradifferentiable functions on \(\Omega\) depending on a real analytic parameter \(\lambda\in U\), the equation \(P(D)u_\lambda = f_\lambda\) can be solved with a family \(u_\lambda\) of the same kind. Since the problem is equivalent to the surjectivity of \(P(D)\otimes \text{id}_{A(U)}\), the techniques and results on surjectivity of tensorized maps \(P(D)\otimes \text{id}_E\) are used which were obtained in the closely related recent paper [Rev. Mat. Iberoam. 26, No. 1, 175--238 (2010; Zbl 1207.35050)] by the same author. In this way it is shown that the above problem always has a solution if and only if the kernel \(\text{ker} P(D)\) satisfies a linear topological invariant of \((P\Omega)\)-type. Using the Ehrenpreis-Palamodov fundamental principle this can be translated for convex \(\Omega\) into a certain Phragmen-Lindelöf condition valid on the characteristic variety of \(P(D)\). This section is the analytical core of the paper. It turns out that the problem is related to the existence of a continuous linear right inverse. In the case of real analytic functions (i.e. \({\mathcal E}_{\{\omega\}}(\Omega)=A(\Omega)\)) the problem has a solution for convex \(\Omega\) if and only if the principal part \(P_p(D)\) has a continuous linear right inverse on \(C^\infty(\Omega)\). The characterization thus is independent of the set \(U\) of parameters in this case. The conditions are then evaluated for operators in two variables, for homogeneous operators and for operators of order at most 2. Especially, the problem cannot always be solved for the Laplace operator and the heat equation while it is always solvable for the wave equation. The paper closes with a section containing several interesting open problems.