Slow-growing counterexamples to the strong Eremenko Conjecture

Andrew P. Brown
2024-05-15
Abstract:Let $f\colon\mathbb{C} \to\mathbb{C}$ be a transcendental entire function. In 1989, Eremenko asked the following question concerning the set $I(f)$ of points that tend to infinity under iteration: can every point of $I(f)$ be joined to $\infty$ by a curve in $I(f)$? This is known as the strong Eremenko conjecture and was disproved in 2011 by Rottenfußer, Rückert, Rempe and Schleicher. The function has relatively small infinite order: it can be chosen such that $\log \log \,\lvert f(z)\rvert = (\log \lvert z\rvert)^{1+o(1)}$ as $f(z)\to \infty$. Moreover, $f$ belongs to the \emph{Eremenko--Lyubich class $\mathcal{B}$}. Rottenfußer et al also show that the strong Eremenko conjecture does hold for any $f\in\mathcal{B}$ of finite order. We consider how slow a counterexample $f\in\mathcal{B}$ can grow. Suppose that $\Theta\colon [t_0,\infty)\to [0,\infty)$ satisfies $\Theta(t) \to 0$ and \[ (\log t)^{-\beta \Theta(\log t)}/\Theta(t) \to 0 \quad\text{ as $t\to \infty$} \] for some $0<\beta<1$, along with a certain regularity assumption. Then there exists a counterexample $f\in\mathcal{B}$ as above such that \[ \log \log \vert f(z)\vert = O ( (\log \vert z \vert)^{1 + \Theta( \log \vert z \vert )}) \] as $\vert f(z)\vert \to\infty$. The hypotheses are satisfied, in particular, for $\Theta(t) = 1/(\log \log t)^{\alpha}$, for any $\alpha>0$.
Dynamical Systems,Complex Variables
What problem does this paper attempt to address?