Computation of analytic capacity and applications to the subadditivity problem (Q380801)

This is a very interesting paper which introduces new ideas for the long-standing problem of the subadditivity of analytic capacity. The analytic capacity of a compact set \(K\) is defined as \[ \gamma(K):=\sup\big\{|f'(\infty)|:f\in H^\infty(\Omega),\|f\|_\infty\leq 1\big\} \] where \(\Omega\) is the complement of \(K\) with respect to the extended complex plane \(\mathbb C_\infty\). This was defined by \textit{L. V. Ahlfors} [Duke Math. J. 14, 1--11 (1947; Zbl 0030.03001)] with the purpose of geometrically characterizing so-called removable sets. \textit{P. R. Garabedian} [Trans. Am. Math. Soc. 67, 1--35 (1949; Zbl 0035.05402)] later showed that, in the case where \(\Omega\) is finitely connected with analytic boundary, \(\gamma(K)\) can also be characterized as the infimum of a family of integrals. There are just a few examples of sets \(K\) for which the analytic capacity \(\gamma(K)\) is known exactly. In this paper, the authors show that, for \(K\) with complement \(\Omega\) finitely connected and with \(C^\infty\) boundary, the analytic capacity is the solution of two extremal problems, namely \[ \gamma(K)=\min\left\{\frac{1}{2\pi}\int_{\partial\Omega}|g(z)|^2|dz|:g\in A(\Omega),\;g(\infty)=1\right\} \] and \[ \gamma(K)=\max\left\{2\text{Re}\,h'(\infty)-\frac{1}{2\pi}\int_{\partial\Omega}|h(z)|^2|dz|:h\in A(\Omega),\;h(\infty)=0\right\}, \] where \(A(\Omega)\) denotes the set of functions holomorphic on \(\Omega\) that extend continuously to the boundary. This is very interesting in that it gives rise to an algorithm for calculating numerically the analytic capacity of a set, by evaluating the extrema over well-chosen finitely generated subspaces of \(A(\Omega)\). The method is shown to converge, and is illustrated in several examples. For numerical reasons, the method is particularly fast when \(K\) is a union of discs. The method is also extended to sets where \(\Omega\) has piecewise analytic boundary. Here, tools from the theory of Smirnov classes have to be used to prove convergence. With this new possibility of evaluating the analytic capacity, the authors are able to take a new look at the old problem of the subadditivity of analytic capacitiy: is it true that, if \(E\) and \(F\) are compact, then \(\gamma(E\cup F)\leq \gamma(E)+\gamma(F)\)? It was shown by \textit{X. Tolsa} [Acta Math. 190, No. 1, 105--149 (2003; Zbl 1060.30031)] that there is a constant \(C\) such that \(\gamma(E\cup F)\leq C(\gamma(E)+\gamma(F))\) for all such \(E\) and \(F\), but whether one can take \(C=1\) is still an open problem. In view of this problem, the authors prove the very interesting result (Theorem~7.1) that in order to prove the subadditivity, it is enough to do so for the case where \(E\) and \(F\) are disjoint, and are finite unions of disjoint closed discs, all with the same radius. This formulation is particularly amenable to their method for calculating capacity. Such calculations have led the authors to a conjecture, namely that, for \(E\) and \(F\) disjoint unions of discs with fixed centers and increasing radius \(r\), the ratio \(\gamma(E\cup F)/(\gamma(E)+\gamma(F))\) is a decreasing function of \(r\). If proved, this conjecture would imply that the analytic capacity is, indeed, subadditive.