I'm currently working through a paper by Brezis on the topology of Sobolev spaces. Right now I am having trouble understanding the following note made by Brezis.
Let $M$ be a compact and smooth manifold and $K \subset M$. Then there exists a diffeomorphism $$\psi: B_1(0) \to \epsilon\text{-neighbourhood of } K$$ (where $B_1(0) \subset \mathbb{R}^n$ and $\dim M = n$).
Does it have to do with $K_\epsilon$ being an open subset of the smooth manifold $M$ and thus a smooth manifold itself? Does that already imply the existence of such diffeomorphism? I am a little confused right now.
I'd appreciate any help.
The author is not claiming that such a diffeomorphism always exists. For instance, if $K = M = S^1$ or $M = S^1\times S^1, K = S^1 \times \{0\}$, then it is clear that no neighborhood of $K$ can be homeomorphic to an open ball.