Achievable connectivities of Fatou components for a family of singular perturbations (Q2158212)

The authors study the connectivity of Fatou components for rational maps in a family of singular perturbations. Here, the connectivity of a domain \(D\subset\hat{\mathbb{C}}\) is by definition the number of connected components of its boundary. Their results extend previous ones obtained by the first author in [Discrete Contin. Dyn. Syst. 37, No. 7, 3567--3585 (2017; Zbl 1401.37055)], which extends the study of Fatou and Julia sets of a smaller family by \textit{C. McMullen} [in: Holomorphic functions and moduli. I. New York etc.: Springer-Verlag. 31--60 (1988; Zbl 0692.30035)]. Let \(f:\hat{\mathbb{C}}\to\hat{\mathbb{C}}\) be a rational map. Its Fatou set \(\mathcal{F}(f)\) is the set of points \(z\in\hat{\mathbb{C}}\), where the family of iterates \(\{f^n\}_{n\ge0}\) forms a normal family in some neighborhood of \(z\), and its Julia set \(\mathcal{J}(f)\) is the complement of \(\mathcal{F}(f)\) in \(\hat{\mathbb{C}}\). It is easy to see that \(\mathcal{F}(f)\) is open. A celebrated result of \textit{D. Sullivan} [Ann. Math. (2) 122, 401--418 (1985; Zbl 0589.30022)] states that any Fatou component (i.e., a connected component of \(\mathcal{F}(f)\)) is preperiodic; in other words, it eventually falls into a periodic cycle under iteration by \(f\). When the orbit of a Fatou component is actually periodic, using the classification of such components, the connectivity of the component is either \(1\), \(2\), or \(\infty\). On the other hand, strictly preperiodic Fatou components may have finite but arbitrarily large connectivities \textit{a priori}, since they are branched coverings of periodic ones under some iterates of \(f\). The major achievement of this paper and its predecessors is to construct a family of examples where arbitrarily large connectivities are achieved. As a matter of fact, in [J. Math. Anal. Appl. 462, No. 1, 36--56 (2018; Zbl 1391.37040)], the first author showed that arbitrarily large connectivities can be achieved for a fixed rational map. As the natural next step, in the paper under review, the authors determine what natural numbers can appear as connectivities for the family (in fact for a more general family than the one discussed in the previous work mentioned above). The basic idea is to consider singular perturbations of certain base family of rational maps. Consider the following family of rational maps of degree \(n+1\): \[ S_{n,a,Q}(z)=\frac{z^n(z-a)}{Q(z)}, \] where \(n\ge2\), \(a\) is a nonzero complex number, \(Q\) is a polynomial of degree at most \(n\) with \(Q(0)Q(a)\neq0\), and satisfying the following conditions: \begin{itemize} \item The point \(z=0\) is a superattracting fixed point of degree \(n\) of \(S_{n,a,Q}\); \item The fixed point \(z=\infty\) is attracting (not necessarily superattracting). This means that the coefficient before \(z^n\) in \(Q\) has modulus strictly less than \(1\); \item There are exactly two Fatou components: the immediate basins of attraction \(\mathcal{A}^\ast(0)\) and \(\mathcal{A}^\ast(\infty)\) of \(0\) and \(\infty\) respectively. \end{itemize} These conditions in particular imply that the Julia set is a quasicircle. Given a family \(S_{n,a,Q}\) satisfying the above three conditions, consider the following perturbation: \[ S_{n,d,\lambda}(z)=S_{n,a,Q}(z)+\frac{\lambda}{z^d} \] where \(d\ge2\) satisfies \(1/n+1/d<1\), and \(\lambda\) is a complex number of small modulus. The new family is called a singular perturbation of the original, as the addition of the term \(\lambda/z^d\) when \(\lambda\neq0\) drastically changes the dynamics near \(z=0\), while the dynamics near \(z=\infty\) remains stable for \(|\lambda|\) small. Interesting dynamics can occur under singular perturbation: \textit{C. McMullen} [loc. cit.] studied the case \(n=2\), \(d=3\), \(a=0\), \(Q(z)=1\), and showed that when \(|\lambda|\) is small but nonzero, the Julia set of \(S_{n,d,\lambda}\) is a Cantor set of circles. The authors show that for \(|\lambda|\) small, the following hold: \begin{itemize} \item The immediate basin of attraction \(\mathcal{A}^\ast_\lambda(\infty)\) of \(z=\infty\) for the perturbed map \(S_{n,d,\lambda}\) is a quasidisk, and contains \(n\) critical points, which are continuous extensions of the \(n\) critical points in the immediate basin of attraction \(\mathcal{A}^\ast_0(\infty)\) of \(z=\infty\) for the original map \(S_{n,a,Q}\); \item When \(\lambda\neq0\), the pole \(z=0\) belongs to a quasidisk Fatou component \(T_\lambda\), which is mapped onto \(\mathcal{A}^\ast_\lambda(\infty)\) under \(S_{n,d,\lambda}\); \item The perturbed map \(S_{n,d,\lambda}\) has \(n+d\) critical points near \(z=0\), and they belong to a doubly connected Fatou component \(A_\lambda\) which is mapped onto \(T_\lambda\) under \(S_{n,d,\lambda}\); \item There is an additional critical point \(\nu_\lambda\) of \(S_{n,d\lambda}\), which is the continuous extension of the critical point \(\nu_0\) of the unperturbed map \(S_{n,a,Q}\) contained in the basin of attraction of \(z=0\); \end{itemize} To increase the connectivity beyond \(2\), it is important to assume that \(\nu_\lambda\) belongs to an iterated preimage \(\mathcal{U}_\nu\) of \(A_\lambda\), and that \(\mathcal{U}_\nu\) surrounds \(0\). Let \(k\) be the minimal number of iterations of \(S_{n,d,\lambda}\) needed to map the critical point \(\nu_\lambda\) into the disk bounded by the outer boundary of the annulus \(A_\lambda\). Under all these technical assumptions, the authors show the following: Theorem. Let \(U\) be a Fatou component of connectivity \(\kappa>2\). Then there exist natural numbers \(i,j,l\) so that \(\kappa=(n+1)^id^jn^l+2\) and \(l\le jk\). Moreover, for any given natural numbers \(i,j,l\) so that \(l\le j(k-1)\), there exists a Fatou component \(U\) of connectivity \(\kappa=(n+1)^id^jn^l+2\). Theorem. For any given \(i,l\ge0\) and \(j>0\), there exists a parameter \(\lambda\) such that \(S_{n,d,\lambda}\) has a Fatou component of connectivity \(\kappa=(n+1)^id^jn^l+2\), and a Fatou component of connectivity \(\kappa=(n+1)^i+2\).