Sharpness of Rrickman's Picard theorem in all dimensions (Q493512)

\textit{S. Rickman} [J. Anal. Math. 37, 100--117 (1980; Zbl 0451.30012)] showed that for \(K>1\) and \(n \geq 2\) there exists \(q=q(n,K) < \infty\) such that a non-constant \(K\)-quasiregular mapping \(f : \mathbb{R}^n \rightarrow \mathbb{R}^n\) omits at most \(q\) points. For another proof of this result, based on potential theoretic reasoning [\textit{J. L. Lewis}, Proc. Am. Math. Soc. 122, No. 1, 199--206 (1994; Zbl 0807.30010)]. For \(n=2\) the quasiregular version of the Picard theorem says that \(q(2,K) = 1\) for every \(K \geq 1\). In the other direction, \textit{S. Rickman} [Acta Math. 154, 195--242 (1985; Zbl 0617.30024)] showed that for \(n=3\) and for each finite set \(P\) of points in \(\mathbb{R}^3\) there exists a quasiregular mapping in \(\mathbb{R}^3\) omitting exactly the set \(P\). For \(q=1\) and all \(n \geq 3\) such a mapping was constructed by \textit{V. A. Zorich} [Math. USSR, Sb. 3, 389--403 (1969; Zbl 0184.10801)]. Rickman's proof relied on a delicate deformation theory. The authors complete the existence part by showing that for \(n \geq 3\) and a set \(P\) consisting of \(q \geq 2\) points, there exists a quasiregular mapping in \(\mathbb{R}^n\) omitting exactly the set \(P\). Although the construction is complicated it has a new feature: mappings of bounded length distortion (BLD) are used. An \(L\)-BLD mapping \(h : \mathbb{R}^n \rightarrow \mathbb{R}^n\) is a discrete and open mapping which satisfies \(l(\gamma)/L \leq l(h \circ \gamma) \leq L l(\gamma)\) for each curve \(\gamma\) in \(\mathbb{R}^n\); see [the reviewer and \textit{J. Väisälä}, Math. Ann. 282, No. 3, 423--443 (1988; Zbl 0632.35021)]. Here \(l(\gamma)\) is the length of \(\gamma\). A BLD mapping is quasiregular but BLD mappings form a much simpler class of mappings than quasiregular mappings. The authors show that a suitable change of metric produces a BLD mapping which has the properties needed in the quasiregular construction. More precisely, they prove that for \(n \geq 3\), \(p \geq 2\) and \(P = \{y_0,\dots, y_p \} \subset S^n\) there exists a surjective BLD mapping \(h : \mathbb{R}^n \rightarrow (M,g)\) where \(M = S^n \setminus P\) and \(g\) is a Riemannian metric on \(M\) such that \(B^n(y_i,r_i) \setminus \{y_i \} \) is isometric, in metric \(g\), to \(S^{n-1}(\delta) \times (0,\delta)\) for some \(\delta > 0\) and all \(0 \leq i \leq p\). From this result the aforementioned existence of the required quasiregular mapping easily follows. The construction for \(h\) uses several, well-illustrated, steps and gluing operations.