In Allen Hatchers short exposition of the Madsen-Weiss Theorem he defines the topology on the space $\mathcal{C}^n$ of smooth oriented properly embedded $d$-dimensional submanifolds of $\mathbb{R}^n$ on page 6 with the help of the

"standard $C^\infty$ topology on the space of d-dimensional smooth compact submanifolds of B that are properly embedded (meaning that the intersection of the submanifold with $\partial B$ is the boundary of the submanifold, and this is a transverse intersection)",

where $B$ is a closed ball centered in the origin.

I am searching for a reference or an explicit definition of the "standard topology" Hatcher is referring to.


It's in Hirsch's "Differential Topology" textbook. Specifically, given a compact manifold $M$, the weak $C^k$-topology on the set of $C^k$-smooth embeddings $Emb(M,\mathbb R^n)$ is the one that requires uniform closeness of not only the maps themselves, but all derivatives up to order $k$. For $C^\infty$ mapping spaces you demand uniform agreement up to some arbitrary order $k$ -- it is not a normable space anymore so you think of the topology as being induced by a countable sequence of (semi)norms, one for each $k$.

That's the space of embeddings. So the space of submanifolds of $\mathbb R^n$ diffeomorphic to $M$ is the space

$$Emb(M,\mathbb R^n) / Diff(M) $$

where $Diff(M)$ is given the same "weak" topology as in Hirsch. $Diff(M)$ is the space of $C^k$-diffeomorphisms of $M$. Notice two embeddings $M \to \mathbb R^n$ have the same image if and only if they differ by a diffeomorphism of $M$.

Then the space of all $m$-dimensional manifolds in $\mathbb R^n$ is the disjoint union:

$$\sqcup_M Emb(M, \mathbb R^n) / Diff(M)$$

where you take the disjoint union over all diffeomorphism types of $m$-dimensional manifolds $M$.

It's a jazzed-up version of the Whitney embedding theorem that states $Emb(M,\mathbb R^n)$ is highly-connected for $n$ much larger than $m$.

  • $\begingroup$ Does everything you say also work with manifolds with boundary? Is the space of submanifolds of a closed ball Hatcher talks about then induced by the subspace topology of the space you described? $\endgroup$ – Madeleine Jan 31 '15 at 19:36
  • $\begingroup$ Yes, it also works although the argument is twice as thick in a sense. Usually you want relative embeddings $(M,\partial M) \to (D^n, \partial D^n)$. It's the same argument in spirit, just more longwinded. $\endgroup$ – Ryan Budney Jan 31 '15 at 20:06
  • $\begingroup$ I thought that topologies induced by sequences of metrics were always metrizable. $\hspace{1.46 in}$ $\endgroup$ – user5810 Jan 31 '15 at 20:21
  • $\begingroup$ Sorry this might be offtopic wrt OP but this answer reminded me of something which I probably should better formulate as a separate question but maybe somebody can tell me here as well. By analogy with moduli spaces, does also in this setting exist a compactification of the above space whose points could be interpreted as manifolds with controllable singularities and which would tie together these connected components? $\endgroup$ – მამუკა ჯიბლაძე Feb 1 '15 at 16:08

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.