Conformal infinitesimal variations of Euclidean hypersurfaces (Q2118038)

Given a Euclidean hypersurface \(f \colon M^{n} \to \mathbb{R}^{n+1}\), a \textit{conformal variation} of \(f\) is a smooth map \(F \colon (-\epsilon, \epsilon) \times M^{n+1}\) such that \(F(0, \cdot) = f\) and \(F(t, \cdot)\) is a conformal immersion for all \(t \in (-\epsilon, \epsilon)\). This paper deals with a weaker class of variations, called \textit{conformal infinitesimal variations}, which are those that preserve angles ``up to first order''. The first main result of the paper, Theorem 1, is a classification of the hypersurfaces of dimension \(n \geq 5\) admitting nontrivial conformal infinitesimal variations; here \textit{nontrivial} means that the corresponding variational vector field (called \textit{infinitesimal bending}) does not come from a conformal Killing vector field of the ambient space. The classification uses the conformal Gauss parametrization and is based on a class of spherical surfaces associated to a second-order hyperbolic or elliptic PDE. It turns out that the family of hypersurfaces described by the theorem is much larger that those admitting nontrivial conformal variations, which were classified by \textit{E. Cartan} [Bull. Soc. Math. Fr. 45, 57--121 (1917; JFM 46.1129.02)] in 1917. The second main result of the paper is about \textit{conformally ruled} hypersurfaces, still of dimension \(n \geq 5\), to which Theorem 1 does not apply. Roughly speaking, it says that a conformally ruled hypersurface that is neither conformally surface-like nor conformally flat admits exactly as many conformal infinitesimal bendings as there are functions on an interval. It also reveals that any such bending is the variational vector field of a conformal variation. The paper is well written and builds on earlier work of the first and second author [Differ. Geom. Appl. 75, Article ID 101721, 21 p. (2021; Zbl 1466.53013)].