Reducibility of Dupin submanifolds (Q819205)

A smooth normal vector field \(\eta\) of an isometric immersion \(f:M^n\to\mathbb{R}^N\) is called a principal normal with multiplicity \(s\geq 1\) if the tangent subspaces \(E_\eta= \ker(h_f-\langle,\rangle\eta)\) have constant dimension \(s\), where \(h_f: TM\times TM\to T^\perp_f M\) stands for the second fundamental form of \(f\) with values in the normal bundle. A principal normal \(\eta\) of multiplicity \(s\) is called Dupin if it is parallel in the normal connection of \(f\) along the distribution \(E_\eta\). A \(k\)-Dupin submanifold is given by \(f\) with flat normal bundle that have exactly \(k\) principal normals all of which are Dupin. It is called weakly reducible if it has a principal normal \(\eta\) for which the connullity distribution \(E^\perp_\eta\), that is, the orthogonal distribution to \(E_\eta\) in the tangent bundle \(TM\), is integrable. The main result is that any \(k\)-Dupin submanifold that is weakly reducible is the \(N\)-Ribaucour transform of a \(( k-1\))-Dupin submanifold determined by a commuting Codazzi tensor of Dupin type. As a main application, an explicit recursive procedure is derived to generate all holonomic Dupin submanifolds in terms of solutions of completely integrable systems of linear partial differential equations of first-order. A complete description of the weakly reducible \(4\)-Dupin submanifolds is given.