Is it possible to prove the Hairy Ball theorem via non-orientability of $P^2(\mathbb{R})$?

That is, the non-vanishing section $s \colon S^2 \to TS^2$ would induce (via “2-to-1” bundle $p \colon S^2 \to P^2(\mathbb{R})$) the non-vanishing “two-dimensional” field on $P^2(\mathbb{R})$, which could not exist? Drawing the simple picture hints that it could be possible, but I can't make the argument rigorous. Maybe the geometric intuition is misleading in this case? If yes, where is the obstruction?

(I hope the question, however amateur, is valuable enough.)

  • 1
    $\begingroup$ What do you mean by "two-dimensional field"? And I know what is a pull-back of section, but you seem to be using push-forward... $\endgroup$
    – Grigory M
    Jun 8, 2011 at 10:53

1 Answer 1


It's always hard to prove you can't do something, but I don't think you can do this. Let's try to make the proof and see where we get stuck.

Nonorientability of $\mathbb{RP}^2$ means that there is no nonvanishing section of $\bigwedge^2 T (\mathbb{RP}^2)$.

Suppose that $\sigma$ were a nonzero section of $T(S^2)$. For every point $x$ in $\mathbb{RP}^2$, there are two points $x_1$ and $x_2$ of $S^2$ lying above $x$. We can take the two vectors $\sigma(x_1)$ and $\sigma(x_2)$ and push them forward to tangent vectors at $x$; call the results $v_1$ and $v_2$. Note that there is no canonical way to number these; we just get an unordered pair of tangent vectors at $x$. We would like to build a section of $\bigwedge^2 T (\mathbb{RP}^2)$.

The only thing I see to do is look at $v_1 \wedge v_2$. But this doesn't work for two reasons (1) Since we don't have a well defined ordering of $v_1$ and $v_2$, the result is only defined up to sign. (2) Even if you could resolve the sign ambiguity, there is no particular reason that $v_1 \wedge v_2$ should be nonvanishing. Why couldn't you have a vector field $\sigma$ on $S^2$ that is everywhere nonvanishing but such that, at some pair of antipodal points $(x_1, x_2)$, the vectors $\sigma(x_1)$ and $\sigma(x_2)$ are parallel?


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.