5
$\begingroup$

The following result is well-known.

Theorem. The antipodal map on $S^n$ is orientation preserving if and only if $n$ is odd.

Below I provide a proof in which there must be an error since I reach to a wrong conclusion. Can somebody please point out the error?

Let $a:S^n\to S^n$ be the antipodal map on $S^n$ and $A$ be the antipodal map on $\mathbf R^{n+1}$. Consider the following diagram:

enter image description here

Let $\Omega=dx_1\wedge \cdots \wedge dx_{n+1}$ be an orientation form on $\mathbf R^{n+1}$. Let $X$ be the vector field on $\mathbf R^{n+1}$ defined as \begin{equation*} X = \sum_{i=1}^{n+1} x_i \frac{\partial}{\partial x_i} \end{equation*} This vector field is nowhere tangent to $S^n$. The standard orientation on $S^n$ is determined by the contracting $\Omega$ using $X$ and restricting it to $S^n$. Thus the $n$-form $\omega=i^*(X\lrcorner \Omega)$ determines the standard orientation on $S^n$. Now we have $$ \begin{array}{rcl} a^*\omega &=& a^*(i^*(X\lrcorner \Omega))\\ &=& (i\circ a)^*(X\lrcorner \Omega)\\ &=& (A\circ i)^*(X\lrcorner \Omega)\\ &=& i^*(A^*(X\lrcorner \Omega)) \end{array} $$

Note that by the very definition of $A$, we have $A^*\eta=(-1)^{k} \eta$ for any $k$-form $\eta$ on $\mathbf R^{n+1}$.

Thus $A^*(X\lrcorner \Omega)=(-1)^{n}(X\lrcorner \Omega)$ and we get $a^*\omega=(-1)^n\omega$. Therefore $a$ is orientation preserving if $n$ is even and orientation reversing if $n$ is odd.

What is the mistake?

$\endgroup$
9
$\begingroup$

Your mistake is the following statement:

Note that by the very definition of $A$, we have $A^*\eta=(-1)^{k}\eta$ for any $k$-form $\eta$ on $\mathbf R^{n+1}$.

This is not true. Any $k$-form is a sum of terms like $f\,dx^{i_1}\wedge\dots\wedge dx^{i_k}$ where $f$ is some smooth function. If $\eta$ is such a form, then the pullback of $\eta$ by $A$ is $$ A^*\eta = (f\circ A) (-1)^k dx^{i_1}\wedge\dots\wedge dx^{i_k}. $$ This is equal to $(-1)^k\eta$ if and only if the coefficient function $f$ is invariant under $A$.

For example, consider the $1$-form $\eta = x^1\, dx^1$. Then $A^* \eta = \eta\ne (-1)^1\eta$.

One way to see what happens when you pull back $X\lrcorner \Omega$ is to note that $A_*X=X$ and $A^*\Omega = (-1)^{n+1}\Omega$. Thus we can compute what the pullback form does to an $n$-tuple of vector fields: \begin{align*} (A^* ( X\lrcorner\Omega)) \left( V_1,\dots,V_n\right) &= (X\lrcorner\Omega) \left( A_*V_1,\dots, A_*V_n\right)\\ &= \Omega \left( X, A_*V_1,\dots, A_*V_n\right)\\ &= \Omega \left( A_*X, A_*V_1,\dots, A_*V_n\right)\\ &= (A^*\Omega) \left( X, V_1,\dots, V_n\right)\\ &= (-1)^{n+1}\Omega \left( X, V_1,\dots, V_n\right)\\ &= (-1)^{n+1}(X\lrcorner\Omega) \left( V_1,\dots, V_n\right). \end{align*} It follows that $A^*(X\lrcorner \Omega) = (-1)^{n+1}X\lrcorner \Omega$.

$\endgroup$
  • $\begingroup$ Thank you Professor Lee. I was repeatedly making this mistake. $\endgroup$ – caffeinemachine Dec 7 '15 at 19:04

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.