The radial growth of univalent functions (Q596226)

Let \(f(z)= z+ a_2z^2+\cdots\) be a function analytic and univalent in the unit disk, i.e. \(f\) belongs to the well-known class \(S\). Denote by \[ \underline\alpha(\theta)= \liminf_{r\to 1-}\,{\log| f(re^{i\theta}|\over \log(1/(1- r))} \] and \(I(f)= \{\theta\in [0, 2\pi): \underline\alpha(\theta)> 0\}\). It follows from a classical result of \textit{D. C. Spencer} [Trans. Am. Math. Soc. 48, 418--435 (1940; Zbl 0025.25902)] that the set \(I(f)\) is countable and \(\sum_{\theta\in I(f)}\underline\alpha(\theta)\leq 2\). The question was raised by N. Makarov (in a letter to the first author) of whether an analogue of the result is valid if \(\underline\alpha(\theta)\) is replaced by \[ \overline\alpha(\theta)= \limsup_{r\to 1-}\, {\log| f(re^{i\theta})|\over \log(1/(1- r))}. \] The main result of the paper is a negative answer to Makarov's question. Specifically, it is s hown that there exists a function \(f\) in \(S\) such that the set \(\{\theta: \overline\alpha(\theta)> 0\}\) is uncountably dense in every interval and its complement is of Baire category I. Such a function cannot be close-to-convex. Moreover, some new questions are raised in the paper.