Let $f: S^n \to S^n$ be a continuous morphism between $n$-spheres. One knows (for example using Freudenthal's suspension thm) that for all $n \in \mathbb{N}$ holds $\pi_n(S^n) \cong \mathbb{Z}$.

Therefore we can define the degree $deg(f) \in \mathbb{Z}$ of $f$ in following unique way such the diagram below commutates:

$$ \require{AMScd} \begin{CD} \pi_n(S^n) @>{f_*} >> \pi_n(S^n) \\ @VV \cong V @VV \cong V \\ \mathbb{Z} @>{\cdot deg(f)}>> \mathbb{Z} \end{CD} $$

Remark: the isomorphisms $\pi_n(S^n) \cong \mathbb{Z}$ are choosen compatible in the way that the fixed generator of $i_n \in \pi_n(S^n)$ is wlog in the left and right vertical maps is maped to $1$.

Therefore the map $\mathbb{Z} \xrightarrow{\text{deg(f)}} \mathbb{Z} $ is given as multiplication map $z \to deg(f) \cdot z$

Consider from now on as $f_n$ the concatenation of canonical maps

$$f_n: S^n \xrightarrow{\text{p}} \mathbb{PR}^n \xrightarrow{\text{q}} \mathbb{PR}^n/\mathbb{PR}^{n-1} \cong S^n$$

Here $p$ comes from double cover $S^n \to \mathbb{PR}^n$ and $q$ is the quotient arising from the CW / pushout structure of $\mathbb{PR}^n \cong \mathbb{PR}^{n-1} \cup_j D^n$

where $j: S^{n-1} \to \mathbb{PR}^{n-1}$ is the attatching map (the same as $p$ but for lower power $n-1$).

One can calculate using homology groups of $S^n$ and $\mathbb{PR}^n$ that

\begin{equation} deg(f_n) = \begin{cases} 2 & \text{if n odd} \\ 0 & \ \text{if n even} \end{cases} \end{equation}

My question is how one can visulize / understand intuitively that $deg(f_2) =0$ therefore that $f_2:S^2 \to S^2$ is null homotopic?

If we consider the case $f_1: S^1 \to S^1$ then one can intuitively relize that $deg(f_1) =2$ since by construction of $f_n$ and the identification $\mathbb{PR}^2 = S^1 /(x \sim -x)$ the map $f_1$ makes $S^1$ to run two times around itself.

But where is the the crux of the matter is why this argument fails for $f_2$?

Here I have drawn (please don't critizise my drawing talent :) ) the situation for n=1:

enter image description here

But for $S^2$ it seems since $deg(f_2)=0$ that $f_2$ can be contracted to constant map. I suppose that this has do with the properties of boundary $S^1$ but I can’t really find an intuitive argument.

Can anybody help me to visualize the intuition behind this phenomena?

  • $\begingroup$ This is an awesome question! +1 $\endgroup$ Sep 19 '18 at 23:23
  • $\begingroup$ Very tangentially related, but a while back I asked essentially the same question if you use an odd dimensional sphere as the domain and use $\mathbb{C}P^n$ everywhere else. See math.stackexchange.com/questions/308318/…. $\endgroup$ Sep 20 '18 at 13:57

For this post, I am thinking of $\mathbb{R}P^n$ modeled by the northern hemisphere of $S^n$ with antipodal equatorial points identified. Also, I am just going to write $f$ instead of $f_n$.

Let $N,S\subseteq S^n$ denote the (closed) northern and southern hemispheres. Let $n\in N$, $s\in S$ be the north and south pole. (For definiteness, $n = (0,0,...,0,1)$ and $s = -n$.

What is $f|_N$? Well, $p|_N$ is the identity (except for at the equator), while $q$ wraps the hemisphere all the way around the sphere. But points on the equatorial $S^{n-1}\subseteq N$ get mapped to $s$.

In other words, we may think of $f|_N$ in hyper polar coordinates as $f(\vec{\theta}, \phi) = 2\phi$. To be clear, $\phi\in [0,\pi]$ measures the angle from $(0,...,1)$ to $x\in S^{n-1}$ and $\vec{\theta} = (\theta_1,....\theta_{n-1})$ with $\theta_1\in[0,2\pi]$, but every other $\phi_i\in[0,\pi],$ is the collection of angle parameters on $S^{n-1}$. (When $n = 2$, $\vec{\theta}$ is the usual $\theta$ in spherical coordinates.)

What is $f|_S$? Well, we first use the antipodal map $a$ to move all these points into the northern hemisphere, then copy $f|_S$. So, $f|_S = f|_N \circ a$. In terms of coordinates, $a(\vec{\theta}, \phi) = (\underline{\vec{\theta}}, \pi-\phi)$ where $\underline{\vec{\theta}} = (-\theta_1, \pi-\theta_2,...,\pi-\theta_{n-1})$.

(Since the anitpodal map has degree $(-1)^{n+1}$, so far this just reproduces your proof that $\deg(f_n) = 1 + (-1)^{n+1}$.)

Thus, we may describe $f$ via $f(\vec{\theta},\phi) = \begin{cases}(\vec{\theta}, 2\phi) & \phi\in[0,\pi/2] \\ (\underline{\vec{\theta}} , 2(\pi-\phi)) & \phi\in[\pi/2,\pi] \end{cases}.$

As a sanity check, this formula clearly gives a continuous $f$ away from $\phi = \pi/2$. But $\lim_{\phi\rightarrow \pi/2^-} (\vec{\theta}, 2\phi) = s = \lim_{\phi\rightarrow \pi/2^+}(\underline{\vec{\theta}}, 2(\pi-\phi))$, so this formula describes a continuous function.

Let's assume $n $ is even. In this language, your question is to find a homotopy between $f$ and a constant map. We will write down this homotopy as a composition of two homotopies. The first homotopy uses the fact that the antipodal map $\vec{\theta}\mapsto \underline{\vec{\theta}}$ is homotopic to the identity because $S^{n-1}$ is an odd dimensional sphere. Suppose $F(\vec{\theta},t)$ is such a homotopy. (For $n=2$, you can use $F(\theta,t) = \theta + t$ with $t\in[0,\pi]$)

We claim that $f_t:=\begin{cases}(\vec{\theta}, 2\phi) & \phi\in[0,\pi/2] \\ F(\underline{\vec{\theta}},t) , 2(\pi-\phi)) & \phi\in[\pi/2,\pi] \end{cases}$ is continuous.

In fact, the argument is just as it was for $f$ above: this is clearly continuous away from $\phi = \pi/2$. And at $\phi = \pi/2$, both formulas limit to $s$, so it is continuous everywhere.

At the end of this homotopy, we get a new map $f_1 = \begin{cases}(\vec{\theta}, 2\phi) & \phi\in[0,\pi/2] \\ (\vec{\theta} , 2(\pi-\phi)) & \phi\in[\pi/2,\pi] \end{cases}.$

Intuitively, this map wraps $N$ all the way around $S^n$ with decreasing lattitude mapping to even more southern lattitudes, then wraps $S$ around the sphere with decreasing lattitude moving northern.

Let's use a final homotopy to get the constant map.

The homotopy here is $G(\vec{\theta}, \phi, t) = \begin{cases}(\vec{\theta}, 2t\phi) & \phi\in[0,\pi/2] \\ (\vec{\theta} , 2t(\pi-\phi)) & \phi\in[\pi/2,\pi] \end{cases}.$ Once again, we need to check that this is continous, and again, this is obvious away from $\phi = \pi/2$.

But when $\phi = \pi/2$, both the top and bottom map send $(\vec{\theta},\phi)$ to $(\vec{\theta}, \pi t)$, so $G$ is continuous. Finally, simply note that $G(\vec{\theta}, \phi, 0) = (\vec{\theta},0) = n$, so is constant.

  • $\begingroup$ Thank you for this excellent and detailed answer. One note: In the part where you decribed how $f$ acts on restrictions to $N$ and $S$, do you mean maybe $deg(f_n) = 1 + (-1)^{n+1}$? $a$ should there be the antipodal map on $S^n$, so this gives $n+1$ reflections. Or did I misunderstood the action of $a$? $\endgroup$
    – KarlPeter
    Sep 20 '18 at 0:19
  • $\begingroup$ @KarlPeter: You got it right. Editting now... $\endgroup$ Sep 20 '18 at 13:55

Consider $\Bbb Z/2$-invariant cell structure on $S^n$ (two hemispheres, ..., two hemiequators, two points), call it $X$. Composition $X \to \Bbb RP^n/\Bbb RP^{n-1}$ contracts $(n-1)$-skeleton, therefore factors through $X/X_{(n-1)} = S^n \vee S^n$. Those spheres map with opposite orientations if $n$ even and with same if $n$ odd, because corresponding top cells of $X$ did so mapping into $\Bbb RP^n$.


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.