5
$\begingroup$

Let $G$ be a compact Lie group with invariant measure $\mu$. An odd function is a continuous function, $\phi:G\to \mathbb{C}$, such that $\int_{G} \phi d\mu=0$. An odd map is a continuous map, $f:G\to G$, such that $\phi \circ f$ is an odd function for all odd functions $\phi:G\to \mathbb{C}$.

For example, $f_{g}(h)=gh$ is an odd map.

Is it true to say that every odd map is not null homotopic?

A weak and indirect motivation: "Every odd map on sphere is not homotopic to a constant map"

Another motivation: The above $f_{g},s$ are not null homotopic.

$\endgroup$
5
  • $\begingroup$ Is this standard notation, or new? $\endgroup$
    – Igor Rivin
    Commented Oct 18, 2014 at 16:07
  • $\begingroup$ @IgorRivin I am not aware of a reference with this notation. So may be it is introduced, already? may be it is not? I am not sure. $\endgroup$ Commented Oct 18, 2014 at 16:10
  • 5
    $\begingroup$ Usually in our planet by odd function people mean a function $f$ such that $f(-x)=-f(x)$, while a function with $\int_X f d\mu=0$ is usually called "a function with vanishing integral"... $\endgroup$ Commented Oct 18, 2014 at 20:30
  • $\begingroup$ @PietroMajer Thanks for the comment. Yes, of course. But the motivation for "name" was the following: I translated the equation $f(x)+f(-x)=0$ in the integral language $\int_{Z_{2}} f(g.x)dg$ where $Z_{2}$ has the obvious antipodal action. This is my reason of "name". I used this terminology in the following question, too.(with the same motivation)(But may be it should be replaced with a better terminology): mathoverflow.net/questions/184400/… $\endgroup$ Commented Oct 18, 2014 at 20:39
  • 2
    $\begingroup$ Since the group $G$ is compact you can use the term "function with vanishing mean" which is self-describing. $\endgroup$ Commented Oct 19, 2014 at 9:30

1 Answer 1

10
$\begingroup$

Here's a counterexample: take $G=S^1=\{z\in\mathbb{C}:|z|=1\}$, $$ f(z)=\begin{cases}z^2:\mathrm{Im}(z)\geq 0,\\ \overline{z}^2:\mathrm{Im}(z)\leq 0. \end{cases} $$ This $f$ is nullhomotopic, but is an odd map because $\int_G \varphi\circ f=\int_G \varphi$ for all $\varphi:G\to\mathbb{C}$.


In the example above, $f$ is continuous, but not differentiable at $z=\pm 1$. The result is true for $C^1$ functions.

Suppose $f:G\to G$ is a continuously differentiable odd map, and for simplicity assume $G$ is connected. We have $\int_G \varphi\circ f\,d\mu=\int_G \varphi\,df_*(\mu)$, where $f_*\mu$ is the pushforward measure. Since $\int_G \varphi\,d\mu=0\Rightarrow\int_G\varphi\,df_*\mu=0$, we must have that $f_*\mu$ is a scalar multiple of $\mu$, and therefore $f_*\mu=\mu$ because $\mu$ has finite total measure.

For $g\in G$, let's define the Jacobian determinant of $f$ at $g$ to be the determinant of the map induced on $T_g(G)$ by $L_{gf(g)^{-1}}\circ f$, and let's write $J_f(g)$ for this determinant. If $g$ is a regular value of $f$, then there is a neighborhood $U$ of $g$ which is evenly covered by $f$, say $f^{-1}(U)=V_1\sqcup\ldots\sqcup V_k$, and on $U$ $$ f_*\mu = \sum_{i=1}^k\frac{1}{|J_f\circ (f|_{V_i})^{-1}|}\mu, $$ so that $f_*\mu=\mu$ implies that on $U$: $$ \sum_{i=1}^k\frac{1}{|J_f\circ (f|_{V_i})^{-1}|}=1 $$ on $U$.

This should imply that $J_f$ doesn't vanish anywhere, and since we assume $G$ is connected, $J_f$ must have constant sign. For any $g\in G$, we can compute the action of $f$ on the top homology group of $G$ locally, and is is multiplication by $\pm(\#f^{-1}(g))$ (where the $\pm$ is the sign of $J_f$). In particular this value is non-zero for some (hence all) $g\in G$. Since $f$ induces a non-zero map on the top homology of $G$, $f$ is not null-homotopic.

$\endgroup$
3
  • $\begingroup$ thank you very much for your interesting counter example. May be it would be interesting to find a compact Lie group for which every odd map is necessarily non nullhomotopic. $\endgroup$ Commented Oct 18, 2014 at 20:08
  • 2
    $\begingroup$ I don't understand this proof. Can you please explain why the Jacobian of $f$ never vanishes? $\endgroup$ Commented Oct 19, 2014 at 10:08
  • 2
    $\begingroup$ The argument shows that at every regular value, the sum over the inverse image of $f$ of $|J_f|^{-1}$ is 1. If $J_f$ vanishes somewhere but is not identically 0, then it should be possible to find some $g\in G$ where $|J_f|$ is arbitrarily small, with $f(g)$ a regular value. This would contradict $\sum |J_f|^{-1}=1$. It looks like there's something to check here, and we still need to rule out $J_f\equiv 0$. $\endgroup$ Commented Oct 19, 2014 at 14:13

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .