Kallin's lemma for rational convexity (Q991561)

A compact subset \(E\) of \(\mathbb C^n\) is said to be polynomially convex if \(K=\widehat{K}:=\{z\in\mathbb C^n:\;|p(z)|\leq\max_K|p|\) for every polynomial \(p\) in \(\mathbb C^n\}\). Similarly, \(K\) is said to be rationally convex if \(K=\widehat{K}_{\mathcal R}:=\{z\in\mathbb C^n: |g(z)|\leq\max_K|g|\) for every rational function \(g\) which is analytic in a neighbourhood of \(K\}\). We have \(K\subset \widehat{K}_{\mathcal R}\subset \widehat{K}\). If \(n=1\), by Runge's theorem, \(K=\widehat{K}\) iff \(\mathbb C\setminus K\) is connected. Note that union of two polynomially convex compact sets may even fail to be rationally convex [\textit{G. Stolzenberg}, Acta Math. 109, 259--289 (1963; Zbl 0122.08404)]. A powerful tool in verifying polynomial convexity of a finite union of polynomially convex sets is the well-known Kallin lemma [\textit{E. Kallin}, Function Algebras, Proc. Int. Symp. Tulane Univ. 1965, 149--152 (1966; Zbl 0142.10104); see also \textit{P. J. de Paepe}, Bull. Lond. Math. Soc. 33, No.~1, 1--10 (2001; Zbl 1041.32006)]. In this paper, the author establishes an analogue of Kallin's lemma for rational convexity which is the following Theorem. Suppose that 1) \(X_1\) and \(X_2\) are polynomially convex compact subsets of \(\mathbb C^n\); 2) \(Y_1\) and \(Y_2\) are polynomially convex compact subsets of \(\mathbb C\) such that \(Y_2\) is a homeomorphic image of the unit interval [0,1], \(\partial Y_1\) is a continuous Jordan curve and \(E=Y_1\cap Y_2\) has one-dimensional Hausdorff measure zero; 3) \(p\) is a rational function with poles outside \(X_1\cup X_2\) such that \(p(X_1)=Y_1\) and \(p(X_2)=Y_2\); 4) \(p^{-1}(\lambda)\cap(X_1\cup X_2)\) is polynomially convex for every \(\lambda\in E\). Then \(X_1\cup X_2\) is rationally convex. This yields in particular the following nice Corollary. Let \(\varphi\) be a \(\mathcal C^1\)-smooth complex valued function defined on a neighbourhood \(U\) of \(0\in\mathbb C\). Assume that 1) \(\varphi(0)=\frac{\partial\varphi}{\partial z}(0)=\frac{\partial\varphi}{\partial\bar{z}}(0)=0\); 2) \(\{z: \Im \varphi(z)=0\}\) is countable. Let \(M=\{(z,\bar{z}+\varphi(z)): z\in U\}\). Then \(N:=\mathbb R^2\cup M\) is locally rationally convex at the origin, i.e., there exists a closed ball \(B\) centered at 0 such that \(B\cap N\) is rationally convex.