A symplectically non-squeezable small set and the regular coisotropic capacity (Q2444966): Difference between revisions

This paper is the third in a series by the authors considering the question ``How much symplectic geometry can a small subset of a symplectic manifold carry?'' By a ``small'' subset, the authors mean one with small Hausdorff dimension, and also one lying in a small radius ball in the standard symplectic \(\mathbb{R}^{2n}\). By ``carrying symplectic geometry'', the authors refer to the sizes of balls and cylinders in the standard symplectic \(\mathbb{R}^{2n}\) into which it can be symplectically embedded. On the one hand, Gromov's non-squeezing theorem showed that a ball carries enough symplectic geometry not to fit inside a cylinder of smaller radius. On the other hand, it is not difficult to show (as the authors do in this paper) that any Lagrangian submanifold carries so little symplectic geometry as to fit inside an arbitrarily small ball. The authors give results of a similar flavour, improving upon a result of Schlenk. The first main theorem (Theorem 1) shows that there exists, for any \(n \geq 2\), a ``small'' \(n\)-dimensional subset \(X\) of the standard symplectic \(\mathbb{R}^{2n}\) which carries enough symplectic geometry to be ``non-squeezable'', in the sense that it does not symplectically embed into the open cylinder \(Z^{2n} = B^2 \times \mathbb{R}^{2n-2}\). This \(X\) is ``small'' in the sense that it is compact, has Hausdorff dimension \(n\), lies in the closed \(2n\)-ball of radius \(\sqrt{2}\) (i.e., of Gromov width \(2\pi\)); and lies in \((\overline{B}^2)^n\) or \((\overline{B}^2)^{n-1} \times \mathbb{R}^2\), accordingly as \(n\) is even or odd. (Here \(B^2\) and \(\overline{B}^2\) denote the open and closed unit balls in \(\mathbb{R}^2\) respectively.) This small non-squeezable set \(X\) is constructed explicitly, as the union of a closed Lagrangian submanifold, together with the image of a smooth map \(S^2 \rightarrow \mathbb{R}^{2n}\). In their proposition 2, the authors give a result in the other direction, showing that certain ``smaller'' subsets of \(\mathbb{R}^{2n}\), in the sense of lying in a smaller ball (though they may actually have dimension greater than \(n\)), ``do not carry much symplectic geometry'': They embed into \(Z^{2n}\). In this way, theorem 1 is ``sharp up to a factor''. Precisely, every compact subset of the unit closed \(2n\)-ball with vanishing \((2n-1)\)-dimensional Hausdorff measure symplectically embeds into \(Z^{2n}\). The other main result of the paper considers the \textit{regular coisotropic capacities} \(A^d_{\text{coiso}}\) introduced by the authors in previous work [``Coisotropic displacement and small subsets of a symplectic manifold'', Math. Z. 271, No. 1, 415--445 (2012; Zbl 1258.53090)]. These capacities are useful as they give lower bounds on the displacement energy; they are related to the Lagrangian capacity of Cieliebak and Mohnke, and a capacity introduced by Geiges and Zehmisch. For each integer \(d\) satisfying \(n \leq d \leq 2n-1\), the \(d\)-th regular coisotropic capacity \(A^d_{\text{coiso}}\) of a symplectically aspherical manifold \((M, \omega)\) is the supremum, over certain coisotropic submanifolds \(N\) of \(M\), of the minimal symplectic area of 2-dimensional discs in \(M\) with boundary in an isotropic leaf of \(N\). The authors previously showed that each \(A^d_{\text{coiso}}\) is a symplectic capacity, although possibly not normalized. In their second main theorem of this paper (Theorem 3), the authors give lower bounds on the capacities of balls, showing \(A^n_{\text{coiso}} (B^{2n}) \geq \pi/2\), and \(A^d_{\text{coiso}} (B^{2n}) \geq \pi/3\) for \(d\) satisfying \(n+1 \leq d \leq 2n-3\). Combined with previous estimates, this shows that \(A^{2n-1}_{\text{coiso}}\) is normalized, and in general that \(A^d_{\text{coiso}}\) is normalized up to the factor 3.