4

I’m not sure if this is a dumb question but I can’t seem to get my way around it for several days now.

Is there a surjective smooth function from the 2-torus to the 2-sphere?

I’d be greatly happy if an example can be provided with some explanation.

On my own, I attempted to use Sard’s Theorem but I can’t seem to know how to draw a conclusion from there.

Thanks a lot in advance.

PS: This is my first question here and I’m sorry if the question fails some obvious expectations. I searched the site first but obtained the reverse question ( i.e., there being no such map from the sphere to the torus).

1 Answers1

4

The general result here is

Proposition: Suppose $M^n$ and $N^n$ are two connected closed manifolds of the same dimension. Then there is a smooth surjective map $f:M\rightarrow N$.

Here's one way of proving it.

Lemma 1. Let $B(0,1)\subseteq \mathbb{R}^n$ be the closed ball of radius $1$ centered at $0\in \mathbb{R}^n$. Then there is a radially symmetric smooth surjective map $g:B(0,1)\rightarrow B(0,1)$ with the property that $g$ maps the points of distance $\geq 3/4$ from $0$ to $0$.

Proof: Let $\psi$ be a smooth bump function which is supported on $[1/4, 3/4]$. The function $x\psi(x)$ is continuous, so achieves a maximum value $K$ on $[1/4,3/4]$.

Thinking of $B(0,1)$ in polar coordinates $(r,\omega)$ with $\omega\in S^{n-1}$, define $g(r,\omega) = ( \psi(r)r/K, \omega)$. Then $g$ is identically $0$ in a neighborhood of $0$, so is smooth at $0$ (which is always a concern when defining things in polar coordinates). In addition, $K$ is chosen so that $\psi(r)r/K\leq 1$ for any $r$ (so the image really does lie in $B(0,1)$, and that for some specific $r$, $\psi(r)r/K = 1$, so the image really is all of $B(0,1)$. $\square$

Lemma 2: There is a smooth surjective map $h:M^n\rightarrow B(0,1)$.

Proof: Given $p\in M$, let $U\subseteq M$ be an open neighborhood of $p$. By shrinking $U$ if necessary, we may assume that the closure of $U$ is diffeomorphic to the closed ball $B(0,1)$. Let $h_1:\overline{U}\rightarrow B(0,1)$ be such a diffeomorphism. Then the function $g\circ h_1$ (with $g$ from Lemma 1) is smooth on $U$ and is identically equal to $0\in B(0,1)$ for points near $\partial U$. In particular, we can extend $g\circ h_1$ to a map $h:M\rightarrow B(0,1)$ by defining $h(x) = \begin{cases}g(h_1(x)) & x\in U\\ 0 & x\notin U\end{cases}$ $.\square$

Lemma 3: There is a smooth surjective map $j:B(0,1)\rightarrow N$.

Proof: Picking a background Riemannian metric on $N$, because $N$ is closed and connected, we can rescale the metric so that the diameter of $N$ is strictly smaller than $1$. Further, because $N$ is closed, this metric is automatically complete, so by the Hopf-Rinow theorem, given any two points, there is a minimizing geodesic connecting them.

So, the exponential map $\exp_n:T_n N\rightarrow N$ is surjective when restricted to a ball $B$ of radius $1$ in $T_n N$. Of course, $T_n N$ is isometric (as an inner product space), there is a diffeomorphism $j_1:B(0,1)\rightarrow B$. Then $j :=\exp\circ j_1$ is the desired map. $\square$

Now, to prove the Proposition, just use $f = j\circ h.$ Since a composition of smooth surjective maps is smooth and surjective, we are done.

  • 1
    Some comments: We actually don't require much of $M$. It can be disconnected, non-compact, or have boundary, and everything still works. Also, if $N$ is not closed, then, of course, to have a surjective map $M$ must be not be closed. Lastly, if both $M$ and $N$ are not closed, I'm not positive what happens, but I think there is still a surjective map. I'd approach it by finding exhaustions of $M$ and $N$ by closed submanifolds. – Jason DeVito - on hiatus Oct 03 '20 at 20:55
  • According to your answer, I think this answer is wrong. Isn't? – C.F.G Oct 03 '20 at 21:29
  • Many thanks for this detailed response. Going to internalize it properly. For now, I think I am satisfied with it. It is interesting to note that it holds in such generality with even milder assumptions on the manifolds (as I had anticipated it would fail for non-closed manifolds)......: However, I would like to ask (as I was thinking about this a while before your answer): Would existence of such surjective smooth maps, perhaps in the case of oriented manifolds, follow because the degree of smooth maps must be necessarily be onto the integers? – Jack LeGrüß Oct 03 '20 at 21:30
  • 1
    @C.F.G: Yes, I agree with Jim Belk's comment on that answer. – Jason DeVito - on hiatus Oct 03 '20 at 22:57
  • 1
    @JackLeGrüß: The degree map is not necessarily onto. For example, as in the answer C.F.G. links to above, every map $S^2\rightarrow T^2$ has degree $0$. There are even more interesting examples. For example, an integer $d$ is a degree of a map $\mathbb{C}P^n\rightarrow \mathbb{C}P^n$ iff $d$ is a perfect $n$-th power. – Jason DeVito - on hiatus Oct 03 '20 at 23:10
  • @Jason: Thanks for the response. It has cleared up a few inaccuracies in my thinking about the problem. I have carefully read your detailed answer to my question and am now fully satisfied with it. – Jack LeGrüß Oct 04 '20 at 00:25