I want to prove that the antipodal map from $S^n$ to $S^n$ is homotopic to the identity map if $n$ is odd. (I know it's actually true if and only if)

If I consider the map $ H(x,t) = (1-2t)x $, why doesn't this work? I don't really see how this is any different if $n$ is odd or even. Is it continuous? Again, I don't see why it isn't



As Henning or Rasmus have explained to you, $H(x,t) = (1-2t)x$ fails to land within $S^n$ for $t\neq 0,1$. The usual trick to arrange that, namely redefine

$$ H(x,t) = \dfrac{(1-2t)x}{\| (1-2t)x \|} \ , $$

doesn't work either because this new $H$ is not continuous for $t= \frac{1}{2}$.

If you want an specific homotopy from $\mathrm{id}_{S^{2n-1}}$ to $-\mathrm{id}_{S^{2n-1}}$, here is one.

Let's start with $n=1$. Let's think $S^1$ as the set of complex numbers $z \in \mathbb{C}$ of modulus 1: $\vert z \vert = 1$. Consider the map:

$$ H: S^1 \times I \longrightarrow S^1 $$

defined by

$$ H(z,t) = e^{i\pi t} z \ . $$

This map is plainly continuous, its image lies on the sphere $\vert H(z,t) \vert = \vert e^{i\pi t}\vert \cdot \vert z \vert = \vert z \vert = 1$ and $H(z,0) = z$ and $H(z,1) = -z$. Hence, it may be the homotopy you were looking for, right?

For $n \geq 1$, exactly the same map works. Now you think your odd sphere $S^{2n-1}$ as being the subset of complex $n$-tuples $z = (z_1, \dots , z_n) \in \mathbb{C}^n$ of modulus 1, $\vert z \vert = \sqrt{\vert z_1\vert^2 + \dots + \vert z_n\vert^2 } = 1$. And you consider the same map:

$$ H: S^{2n-1} \times I \longrightarrow S^{2n-1} \ . $$


$$ H(z,t) = e^{i\pi t} z = (e^{i\pi t} z_1 , \dots ,e^{i\pi t} z_n ) \ . $$

The same proof we did for $n=1$ applies now and shows that $H$ is the desired homotopy.

Notice that, since we had to rely on complex numbers, this does not work for even spheres.

Finally, if you prefer, another way to write the map $H$ for $n=1$, now thinking $S^1$ as the set of points $(x,y) \in \mathbb{R}^2$ such that $\| (x,y) \| = 1$, could be

$$ H(x,y,t) = \begin{pmatrix} \cos (\pi t) & -\sin (\pi t) \\ \sin (\pi t) & \cos (\pi t) \end{pmatrix} \begin{pmatrix} x \\ y \end{pmatrix} \ . $$

That is, we are rotating $(x,y)$ more and more, as $t$ increases, till we arrive at $(-x,-y)$. Notice that you need to be in $\mathbb{R}^2$ to do this: you couldn't rotate anything if we were talking about the "even" sphere $S^0 = \left\{ -1, +1\right\} \subset \mathbb{R}$ and clearly there is no homotopy from $\mathrm{id}_{S^0}$ to $-\mathrm{id}_{S^0}$; that is, a path inside $S^0$, joining $-1$ and $+1$.

  • 1
    $\begingroup$ Thanks, but I don't understand what you mean by "Notice that, since we had to rely on complex numbers, this does not work for even spheres." $\endgroup$ – TRY Oct 7 '11 at 21:47
  • 1
    $\begingroup$ @TRY: try to do the same thing for an even sphere and see what happens! $\endgroup$ – Mariano Suárez-Álvarez Oct 7 '11 at 21:54
  • $\begingroup$ @TRY. I think Mariano's answer is a great one. :-) $\endgroup$ – Agustí Roig Oct 8 '11 at 1:12
  • 1
    $\begingroup$ @TRY. More precisely: where does an even sphere $S^{2n}$ -for instance, $S^0$, or $S^2$- "live"? Is it possible to construct a map like $H$ in those "ambient spaces"? $\endgroup$ – Agustí Roig Oct 8 '11 at 3:15

This follows easily from some algebraic topology, but you can:

1) prove that $GL(n,\mathbb{C})$ is path-connected (by hands)

2) $U(n)$ is path-connected, because they are homotopy-equivalent (equivalence being the Gram-Schmidt orthogonalisation.)

Every element of $U(2n)$ restricts to a map of $S^{2n-1}$. But both identity and antipodal maps lie in $U(n)$, so they are connected by a path. This will be your required homotopy.

  • $\begingroup$ Nice. But I think it suffices that for $n$ odd, both the antipodal and the identity lie in the connected $\mathrm{SL}_n(\mathbb{R})$. Also, I am curious about what is the "some algebraic topology" explicitly. $\endgroup$ – Anonymous Coward Jul 8 '13 at 17:06

Your $H(x,t) = (1-2t)x$ fails to land within $S^n$ if $t$ is strictly between $0$ and $1$.

Hint: $-I\in SO(k)$ iff $k$ is even.


Here is why your approach does not work:

If $t$ is not $0$ or $1$, then $(1-2t)$ will have modulus less than $1$. Hence $(1-2t)x$ does not belong to the sphere anymore. But if $H$ was a homotopy of maps to the sphere, it would take values on the sphere.


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.