Schroeder's equation in several variables. (Q1811451)

When \(\varphi\) is an analytic map of the unit disk \(\mathbb D\subset\mathbb C\) into itself, with \(\varphi(0) = 0\) and \(\lambda =\varphi'(0)\) satisfying \(0 < | \lambda| < 1\), G. Koenigs in 1884 (see JFM 16.0376.01 and JFM 16.0376.02) gave an essentially unique solution \(f\) to Schroeder's functional equation \(f\circ\varphi=\lambda f\) that is analytic in \(\mathbb D\). The authors construct global analytic solutions in the unit ball \(\mathbb D\subset\mathbb C^N\) for \(N>1\) without requiring any consideration of difficult issues of convergence of formal power series, by using the theory of compact composition operators on certain Hilbert spaces of functions analytic on \(\mathbb D\). This enables to give necessary and sufficient conditions, under mild hypotheses on \(\varphi\), for the solution of a natural form of Schroeder's functional equation in the ball. The global solutions are found to exist under no less general conditions than the formal power series solutions found by earlier researchers. The more subtle issue of the existence of an analytic solution in the ball that is locally univalent near the origin becomes a matter of the diagonalizability of certain matrices whose size depends on particular arithmetic relationships among the eigenvalues of \(\varphi'(0)\). Various phenomena which may occur in the several variable setting that do not occur when \(N = 1\), are also discussed.