The boundary behavior of Bloch functions and univalent functions (Q1117349): Difference between revisions

Let F(t) be a real valued function on the unit circle which belongs to the Zygmund class \(\Lambda^*\), and let E, \(J_ 1\), \(J_ 2\) be the sets where respectively \(F'(t)\) exists and is finite, \(F'(t)=+\infty\), and \(F'(t)=-\infty\). The authors prove that either E has positive Lebesgue measure, or else for every interval I, \[ m_ h(I\cap J_ 1)=m_ h(I\cap J_ 2)=\infty, \] where \(m_ h\) is the Hausdorff measure generated by the function \(h(t)=t \log 1/t\). This result is derived from a more general one about differentiability properties in Banach's class \(T_ 2\), which consists of functions which assume a.e. value in their range at most countably often. Now let \(f=u+iv\) be an analytic Bloch function in the unit disk. Then the radial integral of f belongs to \(\Lambda^*\), so the theorem stated above leads to one about the boundary behavior of f: Either f has a finite radial limit on a set of positive Lebesgue measure, or else each of u and v has radial limit \(+\infty\) on sets \(J_ 1(u)\), \(J_ 1(v)\) with \[ m_ h(J_ 1(u)\cap I)=m_ h(J_ 1(v)\cap I)=\infty, \] for every interval I. The same is true when \(+\infty\) is replaced by -\(\infty\). If g is univalent in the disk then log g\({}'\) is a Bloch function. It follows that \(g'\) has a finite radial limit, possibly zero, on a set of infinite h-measure. In particular, this set has Hausdorff dimension 1, and positive logarithmic capacity. Thus, the authors have given affirmative answers to questions raised by McMillan-Pommerenke (1971) and \textit{J. L. Fernández} and \textit{Ch. Pommerenke} [Holomorphic functions and moduli I, Proc. Workshop, Berkeley/Calif. 1986, Publ., Math. Sci. Res. Inst. 10, 75-78 (1988; Zbl 0652.30023), see question 1, p. 76]. N. G. Makarov has recently proved the stronger, essentially best possible result, that \(g'\) has a finite radial limit on a set of positive Hausdorff \(\phi\)-measure, where \[ \phi (t)=t(\log (1/t)\log \log \log (1/t))^{1/2}. \] The authors speculate that in their theorem about \(\Lambda_*\) perhaps h can also be replaced by \(\phi\).