The root extraction problem (Q870098): Difference between revisions

Let \(G\) be the group of all germs of holomorphic diffeomorphisms \(f:(\mathbb{C},0)\to (\mathbb{C},0)\) about the origin of the complex plane \(\mathbb{C}\). The root extraction problem in \(G\) is the following: For a given \(f\in G\) and \(n\in \mathbb N\), is there any \(g\in G\) such that \(g^n=f\)? We denote, for \(\lambda\in\mathbb{C}\), by \(G_\lambda\) the set of all \(f\in G\) with \(\lambda =f'(0)\). If \(|\lambda|\neq 1\), it is known that \(f\) is linearizable, i.e., there is an \(h\in G\) such that \(h\cdot f\cdot h^{-1}= \lambda\cdot id\) and so we see that the root extraction problem is solvable for \(f\). If \(|\lambda|=1\) and if we put \(\lambda=\exp(2\pi ia)\) with \(a\in\mathbb{R}\) the solvability for \(f\) depends on \(a\). If \(a\) is irrational, \textit{J.-C. Yoccoz} [Astérisque 231 (1995; Zbl 0836.30001)] gives a necessary and sufficient condition (the so-called Brujno condition) for \(f\in G_\lambda\) to be linearizable. In this paper the author treats the case where \(\lambda\) is a root of unity \(\lambda=\exp(2\pi i\,p/q)\) with \(p\), \(q\in\mathbb{N}\) coprime. In this case it is known that \(f\) is linearizable if and only if \(f^q=1\), which is exceptional. In general, there is an integer \(k> 0\) such that \(f^q(z)=z+C\cdot z^{kq+1}+o(z^{kq+1})\) for some \(C \neq 0\). The author investigates a formal solution \(\widehat g\) satisfying \(f=\widehat g^N\) with \(\widehat g'(0)=\mu_j= \exp(2\pi(p+jq)/qN)\) for \(j= 0,1,\dots,N-1\). The existence of \(\widehat g\) depends on \(j\) and even if there exists \(\widehat g\), the convergence of \(\widehat g\) is very exceptional. The author asserts that conditions for the convergence of \(\widehat g\) can be given by means of the so-called Ecalle-Voronin modulus of \(f\). The author gives geometric explanation to the exceptionalness of the existence of an \(N\)th root by considering the unfolding of the diffeomorphism \(f\) in a family \(f_\varepsilon\). In the unfolding the author observes an accumulation of periodic points for \(f_\varepsilon\). Their presence is asserted to be an obstruction to the \(N\)th root extraction problem. The author finally makes the link between the root extraction problem and the problem of the section of a curvilinear angle in \(N\) equal parts in conformal geometry.