Conformal welding for finitely connected regions (Q766035)

This interesting paper deals with a numerical method to solve the welding problem for multiply connected domains on the Riemann sphere \(\mathbb S^2\cong\overline{\mathbb C}\). Let \(D_1,\dots,D_n\) be \(n\) pairwise disjoint closed disks on \(\mathbb S^2\) and let \({h_j:\partial\mathbb D\to\partial D_j}\), \(j=1,\dots,n\), where \(\mathbb D:=\{z:|z|<1\}\), be given sense-preserving homeomorphisms. The (finitely connected) welding problem is to find conformal maps \[ f:\Omega=\mathbb S^2\setminus\bigcup_{j=1}^n D_j\overset{\text{into}}{\longrightarrow}\mathbb S^2\quad\text{and}\quad g_j:\mathbb D\overset{\text{into}}{\longrightarrow}\mathbb S^2,\quad j=1,\dots,n, \] that have continuous extensions \(\tilde f\) and \(\tilde g_j\) to \(\overline\Omega\) and \(\overline{\mathbb D}\), respectively, satisfying the relation \[ \tilde f(h_j(\zeta))=\tilde g_j(\zeta)\quad\text{for all }\zeta\in\partial\mathbb D, j=1,\dots,n . \] The geodesic zipper algorithm was invented in the 1980s to solve numerically a similar problem for simply connected domains, (see, e.g., [the author and \textit{S. Rohde}, SIAM J. Numer. Anal. 45, No. 6, 2577--2609 (2007; Zbl 1157.30006)]). A slight modification of this algorithm solves numerically the problem to find the Riemann map onto a given Jordan domain. The author describes a method which allows one to reduce the finitely connected welding problem to the welding problem for simply connected domains. To solve the welding problem for \(n\) disks~\(D_j\) (and, correspondingly, \(n\) homeomorphisms~\(h_j\)) it requires \(n\) simply connected conformal weldings (of the disk with its exterior) and \(n-1\) additional conformal maps. To ensure the existence and (essential) uniqueness of the solution to the simply connected welding problem, the homeomorphisms are assumed to be quasisymmetric (see, e.g., Section VII.3 in [\textit{J. B. Garnett} and the author, Harmonic measure. Cambridge: Cambridge University Press (2005; Zbl 1077.31001)] for details on quasisymmetric and quasiconformal maps). For numerical implementation the author suggests to use the geodesic zipper algorithm mentioned above and demonstrates its efficiency by an example with \(n=2\). To solve the inverse problem, i.e., the problem to find the disks \(D_j\) and the homeomorphisms \(h_j\) given the range \(\Omega\) of \(f\) (which is a multiply connected domain bounded by \(n\) pairwise disjoint Jordan curves), one should be able to find the conformal map of \(\Omega\) onto a circular domain (complement of pairwise disjoint closed disks). The latter problem can be solved by means of Koebe's iterative method (whose convergence was proved by \textit{D. Gaier} [Arch. Ration. Mech. Anal. 3, 149--178 (1959; Zbl 0088.28702)]). Koebe's method is based on composing Riemann maps of simply connected domains onto the unit disk. In order to implement this method numerically the author uses again the geodesic zipper algorithm. Two curious examples of numeric computation are given. As a byproduct of his argument the author proves that a conformal mapping of a finitely connected domain \(\Omega\) onto a circular domain can be represented as a finite composition of conformal maps of simply connected domains. If \(\Omega\) is bounded by \(n\) pairwise disjoint quasicircles, then it is enough to use \(n\) conformal maps of quasidisks. This result is interesting on it its own, although, as the author mentions, both in the paper and in the addendum, it is not new.