Let $M$ be a connected and complete Riemannian manifold of positive dimension, $k$ be a positive integer, and let $\mathfrak{X}^k_c$ be the set of class $C^k$-vector fields on $M$ of compact support. It is well known that $Diff_c^k(M)$, the set of class $C^k$ compactly supported homeomorphisms on $M$ is an infinite-dimensional Lie group with Lie algebra $\mathfrak{X}_c^k$; with exponential map $$ Exp:\mathfrak{X}_c^k\ni V \mapsto \operatorname{Solution}\left[\frac{d}{dt} \phi_t(x) = V(\phi_t(p)),\, \phi_0=1_M\right] \in Diff_c^k(M) $$ and that this map is continuous but not $1-1$ nor onto. My question is, are the conditions on $M$'s geometry which ensure that $M$ is $1-1$?
$\begingroup$
$\endgroup$
7
-
2$\begingroup$ The Riemannian metric is surely irrelevant, as it has no role in the definition of compact sets, vector fields, or the exponential map of a vector field. Every manifold admits a complete Riemannian metric, so this is not a special property of the manifold. $\endgroup$– Ben McKayCommented Jan 7, 2022 at 9:36
-
$\begingroup$ Yes, I imagined it only has to do with $M$'s topological properties. I expected if $M$ is contractable then it should be okay, but I'm not certain if this is true nor if its too strong of a condition.. $\endgroup$– ABIMCommented Jan 7, 2022 at 9:37
-
$\begingroup$ The image of Exp should consists of the path-connected component of the identity in the diffeomorphism group: on the one hand the vector field X gives you a flow connecting the identity to Exp(X), on the other hand (up to checking details) a path from Exp(X) to the identity should be realised by a $C^k$-path in the diffeomorphism group (one has to check that this is still true in this infinite-dimensional manifold), which should correspond to a flow by $C^k$-diffeomorphisms, whose time derivative should give a $C^k$-vector field. $\endgroup$– ThiKuCommented Jan 7, 2022 at 9:57
-
3$\begingroup$ The map is never surjective, see this question $\endgroup$– Saúl RMCommented Jan 7, 2022 at 10:29
-
1$\begingroup$ Contractibility is not sufficient, the exponential map is not one-to-one on $\mathbb{R}^n$ whenever $n\ge 2$. Simply consider the vector field of rotations aruond a codimension-$2$ subspace: it generates a group isomorphic to $\mathbb{S}^1$. $\endgroup$– Benoît KloecknerCommented Jan 7, 2022 at 12:59
|
Show 2 more comments