Dilatation of maps between spheres. (Q701254)

Let \(f:S^m\to S^n\) be a \(C^1\)-map and \(df\) its differential. The dilatation of \(f\) is given by \(\delta (f)=\sup\{ | df(V)| ;| V| =1\}\) [see \textit{R. Olivier}, Invent. Math. 1, 380--390 (1966; Zbl 0146.19802)]. It is natural to measure geometrical complexity of \(f\) by its dilatation. In the above mentioned paper, R. Olivier proved that the Hopf fibrations \(S^{15}\to S^8\), \(S^7\to S^4\) and \(S^3\to S^2\) have dilatation exactly 2. In the paper under review, using the Berstein-Szegő theorem, the author generalizes this result and proves that \(\delta (f)\leq k\), for any polynomial map \(f:S^m\to S^n\) of algebraic degree \(k\). R. Olivier used also the following definition: \[ \delta (n,k)=\inf\{\delta (f);f:S^m\to S^n\, \text{ differentiable, }\deg(f)=k\}, \] where deg\((f)\) denotes the Brouwer degree of the map \(f\), and conjectured that \(\delta (2,k)=k\), for \(k\geq 0\). This conjecture is false. The present author shows that \(\delta (2,4)\leq 2\sqrt{2}\). Moreover, he constructs a family of differentiable maps \(S^3\to S^2\) of Brouwer degree 4, whose dilation approaches 2. This implies \(\delta (3,4)=2\). On the other hand, \textit{J. Roitberg} [Adv. Math. 20, 280--284 (1976; Zbl 0324.55017)] conjectured that \(\delta (\sigma\circ\tau)=\delta (\sigma)\delta (\tau)=2\delta (\sigma)\), at least if \(\sigma\) is one of the Hopf classes. A counterexample is provided in the paper under review.