I'm having trouble reconciling my intuitive notion of "functions close to $f$" with the seemingly very technical definition of the weak and strong topologies on $C^r(M,N)$ in Hirsch's Differential Topology.

Hirsch's definition of the weak topology:

Let $f \in C^r(M,N)$. Let $\varphi: U \to \mathbf R^m$ and $\psi: V \to \mathbf R^n$ be charts on $M$ and $N$, respectively. Let $K$ be a compact subset of $U$ such that $f(K) \subseteq V$. Let $\varepsilon \in (0,\infty]$. Define a weak subbasic neighborhood $$\mathcal N^r(f;(\varphi,U), (\psi,V), K, \varepsilon)$$ to be $$\{ g \in C^r(M,N) \mid g(K) \subseteq V \text{ and } \|D^k(\psi f \varphi^{-1})(x) - D^k(\psi g \varphi^{-1})(x) \| < \varepsilon \text{ for all } x \in \varphi(K) \text{ and all } k = 0, \ldots, r\}.$$ The weak topology on $C^r(M,N)$ is the one generated by all such sets.

I won't bother typing out the definition of the strong topology. These definitions look pretty opaque to me. I have no idea how anyone came up with this definition, I have a very vague picture of what an "$\varepsilon$-neighborhood of $f$" looks like, and I'm not even sure my picture is right.

A picture that is clear to me is the following special case. Let's pretend we don't know about the weak/strong topologies. Certainly in any reasonable universe we would like to define a path of homeomorphisms (resp., diffeomorphisms), i.e, a map $[0,1] \to \operatorname{Homeo}(M)$ (resp., $(-\varepsilon,1+\varepsilon) \to \operatorname{Diff}(M)$), to be continuous (resp., smooth) iff the induced map $M \times [0,1] \to M$ is continuous (resp., $M \times (-\varepsilon,1+\varepsilon) \to M$ is smooth).

Let's call this the "isotopy topology." More precisely, let $I := (-\varepsilon,1+\varepsilon)$, let $M$ be a $C^r$-manifold, and consider a set map $I \to \operatorname{Diff}^r(M)$. Declare such a map to be "continuous" iff the induced map $M \times I \to M$ is $C^r$. Then endow $\operatorname{Diff}^r(M)$ with the finest topology that makes all "continuous" maps continuous (see final topology).

Question 1: $\operatorname{Diff}^r(M)$ inherits the weak/strong topology as a subspace of $C^r(M)$. How does the isotopy topology compare to either of these topologies? (One would certainly hope that the isotopy topology is finer than both the weak and the strong topologies.)

Question 2: Is there a reference for this kind of stuff?

Soft question: Why would anyone not define this to be the natural topology on $\operatorname{Diff}^r(M)$? This makes it clear (at least to me) that an $\varepsilon$-neighborhood of a diffeomorphism is the set of diffeomorphisms that are "a small isotopy away," that "tangent vectors" are vector fields, etc.

Not-really-a-question: I'm sure this generalizes easily to maps in $C^r(M,N)$ that are not necessarily homeomorphisms/diffeomorphisms (can define $\varepsilon$-neighborhood to be functions that are "a small homotopy away," etc.) I just haven't given it too much thought.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.