Epsilon-non-squeezing and \(C^0\)-rigidity of epsilon-symplectic embeddings (Q6042870)

The paper generalizes results on \(C^0\)-rigidity of symplectic embeddings to \(\epsilon\)-symplectic embeddings, and answers a question of \textit{M. Freedman} [private communication] motivated by applications to symplectic integrator methods and topological quantum computing. In particular, an \(\epsilon\)-generalization of Gromov's non-squeezing theorem is derived. Two independent proofs of \(C^0\)-rigidity are given, one based on symplectic capacities and the other on the shape invariant. The second proof can be adapted to \(\epsilon\)-contact embeddings. Given two symplectic manifolds \((M_1, \omega_1)\) and \((M_2, \omega_2)\) of the same dimension, and a Riemannian metric \(g\) on \(M_1\) (not necessarily compatible with \(\omega_1\)), the author calls an embedding \(\varphi: M_1\hookrightarrow M_2\) \(\epsilon\)-symplectic when \(\| \varphi^* \omega_2 - \omega_1 \| \le \epsilon\) for a \(g\)-induced norm on forms. The main result is that there are \(\delta = \delta (\omega_1, g) > 0\) and \(E = E (\omega_1, g, \epsilon) \ge 0\), with \(E \to 0^+\) as \(\epsilon \to 0^+\), such that if \(\epsilon < \delta\) and a sequence of \(\epsilon\)-symplectic embeddings converges uniformly on compact subsets to an embedding \(\varphi\), then \(\varphi\) is \(E\)-symplectic. In particular, if \(\epsilon_k \to 0^+\) and \(\varphi_k\) are \(\epsilon_k\)-symplectic embeddings converging uniformly on compact subsets to another embedding, then the limit is symplectic. For \(M_1=M_2=\mathbb{R}^{2n}\) with the standard symplectic form and metric, explicit bounds on \(\delta\) and \(E\) follow from the proof. The capacity proof is based on showing that embeddings are \(\epsilon\)-symplectic or \(\epsilon\)-antisymplectic if and only if they preserve the symplectic capacity of ellipsoids up to \(\epsilon\)-small errors. Much of the paper is dedicated to proving a linear version of this result. Specifically, a linear \(\epsilon\)-symplectic map \(\Phi \colon \mathbb{R}^{2 n} \to \mathbb{R}^{2 n}\) with \(0 \le \epsilon < 1 / \sqrt{2}\) (the constant \(1 / \sqrt{2}\) is not optimal) is \(\epsilon\sqrt{2}\)-non-squeezing and \(\epsilon\sqrt{2}\)-nonexpanding. And conversely, if a linear map is \(\epsilon\)-non-squeezing and \(\epsilon\)-nonexpanding then it is \(\epsilon'\)-symplectic or \(\epsilon'\)-antisymplectic with \(\epsilon'\) depending on \(\epsilon\) and going to \(0^+\) when \(\epsilon\) does. Finally, the generalization of Gromov's non-squeezing theorem states that if there is an \(\epsilon\)-symplectic embedding of a ball \(B_r\subset \mathbb{R}^{2 n}\) into a cylinder \(B_R\times\mathbb{R}^{2n-2}\subset \mathbb{R}^{2 n}\) with \(0 \le \epsilon < 1 / \sqrt{2}\), then \(R\ge(1 - \sqrt{2} \, \epsilon)^{\sqrt{2 n}}\,r\).