Here a $M,N$ are topological manifolds and $\mathcal{A}$ and $\mathcal{B}$ are atlases. The brackets $[]$ denote the formation of the equivalence class of atlases.

Let $(M,[\mathcal{A}])$ and $(N,[\mathcal{B}])$ smooth manifolds exotic to each other ($M$ and $N$ homeomorphic, lets say $h(M)=N$, but not diffeomorphic with the smooth structures).

I wondered if the following statements are true.

  1. $f\in C^\infty (M,[\mathcal{A}])$ is not equivalent to $f\circ h \in C^\infty (N,[\mathcal{B}])$
  2. There exists an $f\in C^\infty (M,[\mathcal{A}])$ such that there is no $g\in C^\infty (N,[\mathcal{B}])$ such that $f=g\circ h$ and the other way around: There exists an $g\in C^\infty (N,[\mathcal{B}])$ such that there is no $f\in C^\infty (M,[\mathcal{A}])$ such that $g=f\circ h^{-1}$.
  3. For all $f\in C^\infty (M,[\mathcal{A}])$ there is no $g\in C^\infty (N,[\mathcal{B}])$ such that $f=g\circ h$.

Or in more transparent version with atlases dropped from notation and $M=N$ as topological spaces, but still not diffeomorphic.

  1. $C^\infty M\neq C^\infty N$
  2. $C^\infty M\not\subset C^\infty N$ and $C^\infty N\not\subset C^\infty M$
  3. $C^\infty M\cap C^\infty N=\emptyset$

Thanks in advance and maybe the diffoelogy characterization of smoothness is helpful.

Kind regards


(corrected the error)

  • $\begingroup$ I'm guessing it's supposed to be an intersection in the second 3. $\endgroup$ – Glougloubarbaki Feb 10 '14 at 18:31
  • $\begingroup$ 1. is true, so we cannot have both $C^\infty M \subset C^\infty N$ and $C^\infty N \subset C^\infty M$, so by symmetry we have 2. 3. is wrong because constant functions are always smooth $\endgroup$ – Glougloubarbaki Feb 10 '14 at 18:33
  • $\begingroup$ That was actually my intuition: The constant functions are always smooth. but how can one proof it? $\endgroup$ – w_w Feb 11 '14 at 13:09
  • 2
    $\begingroup$ @user127698 Work in coordinates. $\endgroup$ – Neal Feb 11 '14 at 13:19
  • $\begingroup$ Thanks. My question bordered on a stupid question in this case. But as long as you do not know it is better to ask. $\endgroup$ – w_w Feb 12 '14 at 6:48
  1. True, essentially by the definition.

  2. False in general. Edit: There are examples of homeomorphic but not diffeomorphic manifolds $M, N$ so that there exists a smooth homeomorphism $$ h: M\to N. $$ (Milnor's exotic spheres satisfy this.) Note that, of course, the inverse of such $h$ is not smooth. Therefore, the pull-back via $h$ defines an embedding $$ h^*: C^\infty(N)\to C^\infty(M), \quad h^*(\varphi)= \varphi \circ h. $$ Alternatively, if you take a homeomorphism $h$ whose inverse is smooth, you obtain the embedding
    $$ h_*: C^\infty(M)\to C^\infty(N). $$

  3. Always false, take $f$ which is constant.

Note that I was addressing here question in the first three numbered items, the last 3 numbered items make no sense to me.

2nd Edit: One more thing, which answers a separate question by Jason (in comments). An isomorphism of ${\mathbb R}$-algebras $$ C^\infty(N)\to C^\infty(M) $$ will imply diffeomorphism of the corresponding manifolds $M$ and $N$ (no need to assume a priori that they are homeomorphic). This is because there is a natural bijection $$ C^\infty(M,N)\to Hom(C^\infty(N), C^\infty(M))$$ see theorem 2.3 in the book

Navarro Gonzalez and Sancho de Salas, "$C^\infty$-differentiable spaces", Springer Lecture Notes in Math. vol. 1824.

  • 2
    $\begingroup$ Can you give an example where the sets of smooth functions on $M$ is contained in the smooth functions on $N$, that is $C^{\infty}M\subset C^{\infty}N$? $\endgroup$ – w_w Feb 11 '14 at 13:08
  • $\begingroup$ While 1. is true by definition, I think a much more interesting question is whether or not $C^\infty(M)$ and $C^\infty(N)$ can be isomorphic (as $\mathbb{R}$-algebra). I know for compact manifolds, if $C^\infty(M)\cong C^\infty(N)$, then any isomorphism can actually be used to construct a diffeomorphism between $M$ and $N$ (this is an exercise in Milnor/Stasheff's Characteristic classes book), but I actually don't know what happens when $M$ and $N$ are both non-compact. $\endgroup$ – Jason DeVito Feb 11 '14 at 14:10
  • $\begingroup$ @user127698: See the edit. $\endgroup$ – Moishe Kohan Feb 11 '14 at 14:12
  • $\begingroup$ @JasonDeVito: This is a good question, see an edit to my answer. $\endgroup$ – Moishe Kohan Feb 11 '14 at 15:36
  • 1
    $\begingroup$ @user127698: This is definitely true for compact manifolds, it is most likely to be true for noncompact manifolds too, but I have to check this. $\endgroup$ – Moishe Kohan Feb 13 '14 at 19:48

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.