Nevanlinna domains with large boundaries (Q2317980)

In this paper the authors give a complete solution to the following problem posed in the early 2000s: how large, in the sense of dimension, can be the boundaries of Nevanlinna domains? We recall that a bounded simply connected domain $G\subset\mathbb{C}$ is a Nevanlinna domain if there exist two functions $u,v\in H^{\infty}(G)$ with $v\ne 0$ such that the equality \[ \bar{z}=\frac{u(z)}{v(z)} \] holds on $\partial G$ in the sense of conformal mappings. This means the equality of angular boundary values \[ \overline{f(\zeta)}= \frac{(u\circ f)(\zeta)} {(v\circ f)(\zeta)} \] for almost all $\zeta\in \mathbb{T}$, where $f$ is a conformal mapping from $\mathbb{D}$ to $G$, and $f(\zeta)$ is the angular boundary value of $f\in H^{\infty}(\mathbb{D})$ at the point $\zeta\in \mathbb{T}$. Nevanlinna domains play a crucial role in recent progress in problems of uniform approximation of functions on compact sets in $\mathbb{C}$ by polynomial solutions of elliptic equations with constant complex coefficients. The answer to the posed question is the following: \begin{itemize} \item For every $\beta\in [1,2]$ there exists a Nevanlinna domain $G$ such that $\dim_{H}(\partial G)=\beta$, where $\dim_{H}$ stands for the Hausdorff dimension of sets. \end{itemize} Constructing Nevanlinna domains with irregular boundaries is a rather difficult and delicate problem. It is of interest to consider a quantitative version of the problem of the existence of Nevanlinna domains with non-rectifiable boundaries. In particular one studies the question of how the length of the boundary of the Nevanlinna domain $R(\mathbb{D})$ grows in relation to the degree of the rational function $R$. Given a positive integer $n$, let denote by $\mathcal{R}_{n}$ the set of all rational functions of degree at most $n$ and by $\mathcal{R}\mathcal{U}_{n}$ the set of all functions in $\mathcal{R}_{n}$ without poles in $\overline{\mathbb{D}}$ and univalent on $\mathbb{D}$. Finally, let $\mathcal{R}\mathcal{U}_{n,1}$ be the set of all functions $R\in\mathcal{R}\mathcal{U}_{n}$ such that $\|R\|_{\infty,\mathbb{T}}\le 1$. Set \[ \gamma_{0}=\limsup_{n\to\infty}\sup_{R\in\mathcal{R}\mathcal{U}_{n,1}}\frac{\log\ell(R)}{\log n}\text{ where }\ell(R)=\frac{1}{2\pi}\int_{\mathbb{T}} |R'(\zeta)||d\zeta|. \] With these notations the following result is proved: \begin{itemize} \item For some absolute constant $\alpha>0$ and for every $n\ge 1$ we have \[ \alpha\sqrt{n}\le \sup_{R\in\mathcal{R}\mathcal{U}_{n,1}}\ell (R)\le 6\pi\sqrt{n}, \] so that $\gamma_{0}=1/2$. \end{itemize} The new result here is the lower estimate that is obtained by constructing a snake-like domain $R(\mathbb{D})$ with long boundary. Given a bounded simply connected domain $G$ consider the set $\partial_{a}G\subset \partial G$ which consists of all points of $\partial G$ being accessible from $G$ by some curve. It is known that the equality \[ \partial_{a}G=\{f(\zeta)\cdot \zeta \in \mathcal{F}(f)\} \] holds, where $f$ is some conformal mapping from the unit disc $\mathbb{D}$ onto $G$ and $\mathcal{F}(f)$ it is the set of all points $\zeta\in\mathbb{T}$ where the angular boundary value $f(\zeta)$ exist. The definition of Nevanlinna domains imposes conditions only on the accessible part $\partial_{a}G$ of their boundaries. So it is more accurate to pose the question about the existence of Nevanlinna domains with large accessible boundaries. Let $f$ be a map defined on the unit disc by \[ f(z)=\sum^{\infty}_{n=1}\frac{c_{n}}{1-\bar{a}_{n}z}, \tag{1} \] where $(a_{n})_{n\ge 1}$ is some interpolating Blaschke sequence, that is, satisfying the condition \[ \inf_{n\in \mathbb{N}}\prod^{\infty}_{\begin{subarray}{c} k=1\\ k\ne n\end{subarray}}\left| \frac{a_{n}-a_{k}}{1-a_{n}\bar{a}_{k}}\right|>0, \] and $(c_{n})_{n\ge 1}$ is an appropriately chosen sequence of coefficients. The main result of the paper is the following: \begin{itemize} \item For every $\beta\in [1,2]$ there exists a function $f$ of the form (1) univalent in $\mathbb{D}$ and such that the Nevanlinna domain $G=f(\mathbb{D})$ satisfies the property $\dim_{H}(\partial_{a}G)=\beta$. \end{itemize} If $\theta\in H^{\infty}(\mathbb{D})$ is an inner function and $H^{2}=H^{2}(\mathbb{D})$ is the standard Hardy space, define the so-called model space \[ K_{\theta}:=(\theta H^{2})^{\perp}=H^{2}\ominus \theta H^{2}. \] The function $f$ in the main result belongs to the space $K_{B}$ for some appropriately chosen Blachke product $B$. It is possible to construct similar examples working with univalent functions from the space $K_{S}$, where $S$ is some singular inner function. The simplest example of such a space $K_{S}$ is the Paley-Wiener space $\mathcal{P}W^{\infty}_{[0,1]}$, the Fourier image of $L^{2}[0,1]$, considered as a space of functions analytic in the upper half-plane $\mathbb{C}_{+}$. Then the authors prove the following result: \begin{itemize} \item For every $\beta\in [1,2]$ there exists a univalent function $f$ belonging to the space $\mathcal{P}W^{\infty}_{[0,1]}$ such that the Nevanlinna domain $G=f(\mathbb{C}_{+})$ satisfies the property $\dim_{H}(\partial G)=\beta$. \end{itemize}