Linearization of vector fields near resonant hyperbolic rest points (Q1891315)

The author considers the problem of linearizing a vector field \(\dot x = X(x) = Ax + O_2 (x)\), \(x \in \mathbb{R}^n\), in a neighborhood of a hyperbolic fixed point, here \(x = 0\). That is, none of the eigenvalues \(\lambda_1, \ldots, \lambda_n\) of \(A\) falls onto the imaginary axis. In the analytic context, the Poincaré-Siegel theorem says that there exists an analytic linearization (that is, an analytic transformation conjugating the local flow of \(X\) to the local flow of \(\dot x = Ax)\) provided all eigenvalues are nonresonant in a strong sense. On the other hand, the later Hartman-Grobman theorem says that there always exists a topological linearization, without any further assumptions. Between these two extremes there is the question how smooth the linearization could be if there are nonresonance relations up to a finite order \(h\): \(\lambda_j \neq k_1 \lambda_1 + \cdots + k_n \lambda_n\), for \(1 \leq j \leq n\), \(2 \leq k_1 + \cdots + k_n \leq h\), and \(k_i \geq 0\). Such results were first obtained by S. Sternberg and later for example by G. R. Sell. The author obtains a new theorem of this kind and compares his result with the known ones. He shows that each of them is better than the other two for certain configurations of eigenvalues. He also discusses the question of the smoothness requirements on the vector field \(X\). Here his result improves on the known ones.