How to prove that $SO(3) \neq S^2 \times S^1$ using Hairy Ball Theorem? In other words, if assuming $SO(3) = S^2 \times S^1$, how to construct a non-vanishing vector field on $S^2$?

  • $\begingroup$ Can you provide any more context or details? For instance do you know for a fact that "HBT $\implies SO(3)\neq S^2 \times S^1$" and you just need to prove it, or is this a conjecture that you're trying to verify? $\endgroup$ – William Mar 29 '20 at 14:33
  • 1
    $\begingroup$ Hi @William, this is an exercise from Introduction to Mechanics and Symmetry by Jerrold E. Marsden and Tudor S. Ratiu, Exercise 1.2-4. That's why I'm pretty sure this is correct. $\endgroup$ – hb12ah Mar 30 '20 at 0:27

I'm not exactly sure what you're looking for, but here's one approach.

Inside of the tangent bundle to $S^2$, $TS^2$, we can consider the vectors of unit length: $T^1S^2 =\{(p,v)\in TS^2: |v| = 1\}$. Then we have a map $\pi:T^1 S^2\rightarrow S^2$ given by $\pi(p,v) = p$, and this is a fiber bundle with fiber $S^1$.

Proposition 1: $T^1S^2\rightarrow S^2$ is bundle isomorphic to $\rho:SO(3)\rightarrow S^2$ with $\rho(A) = A_1$, the first column of $A$.

Proof: Given $(p,v)\in T^1 S^2$, note that $p\in S^2$ so $|p| = |v| = 1$. Further, $p\bot v$ since $v\in T_p S^2$. Thus, using the cross product on $\mathbb{R}^3$, the triple $\{p,v,p\times v\}$ is an oriented orthonormal basis of $\mathbb{R}^3$. So, we can create a map $\phi:T^1 S^2\rightarrow SO(3)$ by defining $\phi(p,v) = A$ where $A$ is the matrix whose columns are $A_1 = p, A_2 = v, A_3 = p\times v$. A simple calculation verifies that $\det (A) = 1$, so this really does land in $SO(3)$.

It's also easy to write down the inverse - $\phi^{-1}(A) = (A_1, A_2)$.

Finally, simply note that $\rho(\phi(p,v)) = p = \pi(p,v)$, so $\phi$ is a bundle map.$\square$

Note that under $\phi$, the projection map $T^1S^2\rightarrow S^2$ turns into $\rho:SO(3)\rightarrow S^2$ with $\rho(A) = A_1$.

Now, let $\psi:T^1 S^2\rightarrow S^1\times S^2$ be a diffeomorphism (or even just a homeomorphism). Transporting the bundle structure on $SO(3)$ to $\psi$, we see that the map $\alpha:\rho \circ \psi^{-1}:S^1\times S^2\rightarrow S^2$ is a fiber bundle map with fiber $S^1$. In particular, it is a fibration.

Proposition 2: The bundle $\alpha:S^1\times S^2\rightarrow S^2$ has a section. That is, there is a map $f:S^2\rightarrow S^1\times S^2$ for which $\alpha \circ f= Id_{S^2}$.

Proof: The last portion of the long exact sequence in homotopy groups associated to the bundle $S^1\rightarrow S^1\times S^2\xrightarrow{\alpha} S^2$ looks like $$0=\pi_2(S^1)\rightarrow \pi_2(S^1\times S^2)\xrightarrow{\alpha_\ast} \pi_2(S^2)\rightarrow \pi_1(S^1)\rightarrow \pi_1(S^1\times S^2)\rightarrow \pi_1(S^2)= 0$$ which is the same as $$0\rightarrow \mathbb{Z}\xrightarrow{\alpha_\ast} \mathbb{Z}\rightarrow \mathbb{Z}\rightarrow \mathbb{Z}\rightarrow 0.$$ The last self map of $\mathbb{Z}$ must be surjective by exactness. Since $\mathbb{Z}$ is Hopfian, this last self map is an isomorphism. Now exactness shows that $\alpha_\ast$ is an isomorphism as well.

Because $\alpha_\ast$ is an isomorphism, it maps generators to generators. So, let $g:S^2\rightarrow S^1\times S^2$ be a generator of $\pi_2(S^1\times S^2)$. Then the composition $S^2\xrightarrow{g} S^1\times S^2\xrightarrow{\alpha} S^2$, perhaps after replacing $\alpha$ with $a\circ \alpha$ with $a$ the antipodal map, has degree $1$, so is homotopic to the identity map.

Since $\alpha$ is a fibration, we may therefore apply the homotopy lifting property. This yields a homotopy $g_t$ of $g$ for which $g_0 = g$ and $g_1$ has the property that $\alpha\circ g_1 = Id_{S^2}$. Now simply take $f= g_1$.$\square$

Proposition 3: Assuming $SO(3)$ is homeomorphic to $S^1\times S^2$, there is a non-vanishing vector field on $S^2$.

Proof: Such a homeomorphism $\psi:SO(3)\rightarrow S^1\times S^2$ gives the map $\alpha = \rho \circ \psi^{-1}: S^1\times S^2\rightarrow S^2$ the structure of a fiber bundle over $S^2$ with fiber $S^1$. Applying proposition $2$, there is a map $f:S^2\rightarrow S^1\times S^2$ with $\alpha \circ f = Id_S^2$.

Then the map $\psi^{-1}\circ f: S^2\rightarrow SO(3)$ has the property that $\rho\circ(\psi^{-1}\circ f) = (\rho \circ \psi^{-1})\circ f = \alpha \circ f = Id_{S^2}$. That is, the bundle $SO(3)\rightarrow S^2$ has a section. Since this bundle is isomorphic to the bundle $T^1 S^2\rightarrow S^2$ (Proposition 1), this bundle must have a section. But a section is a unit length vector field on $S^2$, so is, in particular, a non-vanishing vector field. $\square$

  • 2
    $\begingroup$ Incidentally, I accidentally originally proved the converse: if $SO(3)\neq S^1\times S^2$, then $HBT$ is true. This proof is significantly easier: assuming $HBT$ is false, one can use the non-vanishing vector field to show $T^1 S^2 \cong S^1\times S^2$ by simply writing down a map. This, together with Proposition 1 above shows that $SO(3) = S^1\times S^2$. The contrapositive of all this gives the desired result. $\endgroup$ – Jason DeVito Mar 29 '20 at 18:43
  • $\begingroup$ Thanks, excellent solution! I have a few questions that I hope you could help clarify. 1) Regarding your comment, if we have a non-vanishing vector field on $S^2$, how do we construct a section of $T^1 S^2$. One thing I can think of is showing the norms of the vectors in the vector field have a nonzero lower bound since $S^2$ is compact. 2) After proving the bundle isomorphism between $T^1 S^2$ and $SO(3)$, you could treat them as if they are the same? I found it a bit confusing when you said "Transporting the bundle structure on 𝑆𝑂(3) to $\Psi$". $\endgroup$ – hb12ah Mar 30 '20 at 0:14
  • 1
    $\begingroup$ @hb12ah: For 1) you don't even need compactness. Give $M$ some background Riemannian metric. Then, if $V(p)$ is a non-vanishing vector field, then $\frac{V(p)}{|V(p)|}$ is a unit length vector field. (Different choices of Riemannian metrics lead $T^1 M$ being a different subset of $TM$, but all "copies" of $T^1 M$ are bundle isomorphic.) For 2) Yes, $T^1 S^2$ and $SO(3)$ are essentially the same. For the confusing part, I should have written "Transporting the bundle structure on $SO(3)$ by $\psi$." This is a general principle: if $A$ and $B$ are two objects and ..... $\endgroup$ – Jason DeVito Mar 30 '20 at 2:18
  • $\begingroup$ $\psi:A\rightarrow B$ is an isomorphism, then you can use $\psi$ to transport any "special structure" $A$ has (as long as its defined by maps into or out of $A$) to $B$. In this case, $SO(3)\rightarrow S^2$ is a bundle, so you can use $\psi$ to make $S^1\times S^2$ a bundle over $S^2$ by using the projection map $\rho\circ \psi^{-1}$. Then $\psi$ isn't just a diffeo, but a bundle isomorphism. But there are many more examples. For instance, if $(R,+,\cdot)$ is a ring and $(G,+)$ is a group with $f:R\rightarrow G$ an isomorphism from the additive group of $R$ to $G$, you can define a ... $\endgroup$ – Jason DeVito Mar 30 '20 at 2:23
  • 2
    $\begingroup$ If $V(p)$ is smooth, then $\frac{V(p)}{|V(p)|}$ is smooth. This is simply because the real function $f(x,y) = \frac{x}{y}$ (for $x,y\in\mathbb{R}$) is smooth wherever it is defined. For, in local coordinates, $V(p)$ is really just an $n$-tuple of $C^\infty$ functions. $\endgroup$ – Jason DeVito Mar 30 '20 at 2:27

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.