The width of ellipsoids (Q2321848): Difference between revisions

The paper is concerned with geodesic networks obtained on a \(2\)-dimensional Riemannian sphere via Almgren-Pitts minmax contructions, and with the corresponding \(k\)-widths (where \(k\) is the number of parameters in the minmax). Computations are carried out specifically for \(k=1, \dots, 8\). The case in which the ambient dimension \(n+1\) is equal to \(2\) is peculiar, when compared to higher-dimensional cases. Certain properties of the stationary varifolds obtained via the minmax for \(3\leq n+1\leq 7\) are not true in the case \(n=1\). One of these is embeddedness: For \(n=1\), geodesic networks, rather than embedded geodesics, arise in general from the minmax construction. In this paper, the author shows that, for any \(k\), the density of the geodesic network is everywhere given by an integer (in particular, a triple junction, where there is a point with density \(3/2\), is not permitted). A second aspect in which \(n=1\) differs considerably from the case \(3\leq n+1\leq 7\) is the multiplicity of the stationary varifold: the author proves that an unstable embedded closed geodesic with multiplicity \(2\) can arise as minmax critical varifold on an ellipsoid (sufficiently close to the standard sphere), for a minmax construction with \(k\) parameters, with \(4\leq k\leq 8\). This is in stark contrast with Marques-Neves' multiplicity-\(1\) conjecture, whose validity is established by \textit{X. Zhou} in [``On the multiplicity one conjecture in min-max theory'', Preprint, \url{arXiv:1901.01173}]: According to this, minmax varifolds (with any number of parameters) appear with unit multiplicity when \(3\leq n+1\leq 7\). An additional consequence of the arguments in this paper is an example of stationary varifold, that realizes the \(k\)-th width of an ellipsoid, for which the sum of the index and nullity is strictly smaller than \(k\) (as before, \(k\leq 8\)). Here, index and nullity are computed by using ambient deformations induced by vector fields. Once again, \(n=1\) is peculiar: Due to the possibility, for example, of self-intersecting curves, ambient deformations do not reflect all deformations that are encoded in the minmax procedure.