Isotopy problems for saddle surfaces (Q973103)

The investigations in this paper arose in connection with A. D. Alexandrov's uniqueness conjecture for smooth convex surfaces in \(\mathbb R^3\) put forth after it has been proved for analytic surfaces [\textit{A. D. Alexandrov}, C. R. (Dokl.) Acad. Sci. URSS, n. Ser. 22, 99--102 (1939; Zbl 0020.40202 and JFM 65.0828.03)]. The conjecture states that if for a smooth convex body \(K \subset \mathbb R^3\) there is a constant \(C\) so that for every point of \(\partial K\) one has \(R_1 \leq C \leq R_2,\) where \(R_1, \, R_2\) are the principal curvature radii of \(\partial K,\) then \(K\) is a ball. A counterexample to the smooth version of the conjecture was found by \textit{Y. Martinez-Maure} [C. R. Acad. Sci., Paris, Sér. I, Math. 332, No.~1, 41--44 (2001; Zbl 1008.53002)], the construction of which required the introduction of hyperbolic virtual polytopes, hyperbolic hérissons and associated saddle surfaces. The counter-example to the conjecture is given as the Minkowski sum of a hyperbolic hérisson and a sufficiently large ball. If \(h: \mathbb R^3 \to \mathbb R \) is a continuous positively homogeneous \(C^2\)-function, the hérisson \(H\) with the support function \(h\) is the surface obtained as the envelope of the family of planes \(\{e_h(\xi) \}_{\xi \in S^2},\) where \(e_h(\xi)\) is defined by \(\langle \xi ,x \rangle = h(\xi).\) One can also associate a certain polytopal surface to \(h\) (the virtual polytope with the support function \(h\)). A function \(h\) is said to be hyperbolic if its affine graph is a saddle surface for every normal direction \(\xi \in S^2.\) A surface \(F \subset \mathbb R^3\) is called a saddle surface if no plane intersects \(F\) locally at just one point. Likewise, a surface \(F \subset \mathbb R^3\) is called a spherically saddle surface if no great two-dimensional sphere intersects \(F\) locally at just one point. An hérisson (or a virtual polytope) is hyperbolic if its support function is hyperbolic. The author proves that a smooth saddle surface in \(S^3\) (not coinciding with a great sphere) has at least four inflection arches. Further, each hyperbolic hérisson generates an arrangement of at least four disjoint oriented great semicircles on the unit sphere \(S^2\) and there is a natural one-to-one correspondence of semicircles of the arrangement to horns of the hérisson as well as to the inflection arches of the graph of the support function \(h_H.\) A point \(P\) on a surface is called a horn if there exists a plane through \(P\) which intersects the surface locally just at one point \(P\). The author further constructs two hyperbolic polytopes with four horns and smoothens them to obtain two non-isotopic hyperbolic hérissons which are smooth saddle surfaces (except for the horns). One of them appeared in the work of Martinez-Maure and the other one is a new example.