Removable sets for Lipschitz harmonic functions in the plane (Q1573715)

This important paper contains the long and technical proof of a deep result for planar harmonic functions. A compact set \(E\) in the complex plane \(\mathbb C\) is removable for Lipschitz harmonic functions if whenever \(U\) is an open set containing \(E\) and \(u:U\to \mathbb R\) is a Lipschitz function which is harmonic in \(U\setminus E\), then \(u\) is harmonic in \(U\). Let \(H^1\) denote the one-dimensional Hausdorff measure (length). The authors prove that, under the assumption that \(H^1(E)<\infty\), \(E\) is removable for Lipschitz harmonic functions if and only if \(H^1(R\cap \Gamma)=0\) for every rectifiable curve \(\Gamma\). The same statement but with ``bounded analytic functions'' in the place of ``Lipschitz harmonic functions''was an old and famous conjecture until its recent proof by \textit{G. David} [Rev. Mat. Iberoam. 14, No. 2, 369-479 (1998; Zbl 0913.30012)]. Actually, this paper of David was written after the paper under review and it partially relies on the results and techniques of the present paper. By an old result of Painlevé, if \(H^1(E)=0\), then \(E\) is removable. Also, Calderón and others proved that if \(H^1(E)>0\) and \(E\) lies on a rectifiable curve then \(E\) is not removable. So, one direction of the theorem was known to be true. To prove the other direction the authors make use of a recently discovered tool: Melnikov's curvature of a measure. \textit{P. Mattila, M. S. Melnikov}, and \textit{J. Verdera} [Ann. Math. 144, II. Ser. No. 1, 127-136 (1996; Zbl 0897.42007)] studied the case when \(E\) is regular using a relation between Melnikov's curvature and the Cauchy kernel. The paper under review uses this relation as well as a purely geometric-measure-theoretic result of G. David and \textit{J. C. Léger} [Ann. Math. (2) 149, No. 3, 831-869 (1999; Zbl 0966.28003)] which connects Melnikov's curvature and rectifiability. The problem is thus reduced to the construction of a measure with certain properties. Almost all of the paper consists of delicate and technical arguments which lead to such a measure. The construction uses a special system of dyadic cubes and stopping time arguments. It is known that if \(E\) has Hausdorff dimension bigger that \(1\), then \(E\) is not removable. So, for a complete geometric characterization of removable sets, the only remaining case is when \(H^1(E)=\infty\) and \(E\) has dimension \(1\).