10

Maybe I should think a bit longer, but are there more than two homotopy classes of maps $\mathbb{RP}^2\rightarrow \mathbb{RP^2}$? I am interested in both based and unbased maps.

Thomas Rot
  • 10,399
  • Yes. See the second paragraph of http://math.stackexchange.com/questions/915897/homotopical-classes-of-mappings-mathbbcpn-to-mathbbcpm – Cheerful Parsnip Oct 29 '15 at 17:33
  • 2
    I don't see how either of these comments are relevant, to be honest. As far as I can tell they would only help to calculate homotopy classes of maps $\Bbb{RP}^2 \to \Bbb{RP}^m$, $m>2$. –  Oct 29 '15 at 17:35
  • @MikeMiller: actually I was really thinking about Piotr's answer for the complex case, where he claims the case $m=n$ is also covered. – Cheerful Parsnip Oct 29 '15 at 17:49
  • 4
    @GrumpyParsnip: This only works for $\Bbb{CP}^n$ because this has no $(2n+1)$-cells. Cellular approximation implies that if $X$ is a $k$-dimensional complex, $[X,Y] = [X,Y^{(k+1)}]$, where what I mean here is the $(k+1)$-skeleton of $Y$ - to get injectivity you need to be able to find a (relative) cellular approximation of a homotopy $X \times I \to Y$, so you need to work into the $(k+1)$-skeleton. Then the key point here is that $\Bbb{CP}^{\infty}$ has no $(2n+1)$-cells. The same technique in the real case would prove $[\Bbb{RP}^2,\Bbb{RP}^3] = [\Bbb{RP^2},\Bbb{RP}^\infty]$. –  Oct 29 '15 at 17:51
  • @MikeMiller: thanks, I was not thinking it through carefully enough. – Cheerful Parsnip Oct 29 '15 at 17:52

4 Answers4

7

There are at least $\Bbb Z$-many unbased homotopy classes of maps.

First, an odd map $S^2 \to S^2$ (that is, one such that $f(-x) = -f(x)$) descends to a map $\Bbb{RP}^2 \to \Bbb{RP}^2$. I claim that the degree of the map on the level of $S^2$ is a homotopy invariant of the map on the level of $\Bbb{RP}^2$. For pick a homotopy $f_t: \Bbb{RP}^2 \times I \to \Bbb{RP}^2$. By assumption $f_0$ came from an odd map $\tilde f_0: S^2 \to S^2$. I claim that there is a lift $\tilde f_t: S^2 \times I \to S^2$ such that all of the $\tilde f_t$ are odd.

This is pretty easy: just pick a lift! I claim that any lift is automatically odd. For if $\tilde f_t$ lifts $f_t$, we necessarily have $\tilde f_t(\{x,-x\}) = \{x,-x\}$; if this is sufficiently close to an odd map, then $\tilde f_t(x)$ must be close to $-x$; so $\tilde f_t(x)$ must actually be $-x$.

So any map that descends from an odd map $S^2 \to S^2$ is only homotopic to maps that descend from odd maps, and the homotopy class of the odd map is a homotopy invariant of the map $\Bbb{RP}^2 \to \Bbb{RP}^2$.

In particular, its degree is determined by the odd map up above. Now all you need to know is that there are odd maps of arbitrary odd degree. I rather believe I once proved this but I don't remember the construction right now. I'll edit it in if I remember it.

The same thing works for even maps - but for $S^2$ an even map must be degree 0. The lift of a map $\Bbb{RP}^2 \to \Bbb{RP}^2$ is either even or odd, so we've now classified all of them.

  • Based maps are probably more complicated. –  Oct 29 '15 at 18:31
  • This is great thanks! – Thomas Rot Oct 30 '15 at 09:43
  • 1
    One remark: I think you still would need to proof that any two odd maps that have to same degree must be homotopic through odd maps right? – Thomas Rot Oct 30 '15 at 09:49
  • @ThomasRot: Hm, you're right. I missed that. My guess is that it's true. –  Oct 30 '15 at 13:12
  • When you say that the degree of a self-map of $\mathbf P^2(\mathbf R)$ is well defined for self-maps having odd lift to $S^2$, you probably had in mind the absolute value of the topological degree of the lift. Indeed, the two lifts of a given self-map of $\mathbf P^2(\mathbf R)$ have opposite topological degree, as any one of them is equal to the composition of the other one with the antipodal map. Therefore, it would probably have been better to say that there are $\mathbf N$-many homotopy classes of self-maps of $\mathbf P^2(\mathbf R)$ having an odd lift. – Johannes Huisman Sep 19 '18 at 07:57
  • 1
    As a final comment one may observe that the monoid of homotopy classes of self-maps of $\mathbf P^2(\mathbf R)$ is isomorphic to the multiplicative commutative monoid $\mathbf{N}\mathrm{odd}\cup{0,2}$. The latter monoid is an extension of the usual multiplicative monoid $\mathbf{N}\mathrm{odd}$ of all odd natural numbers; the element $0$ is absorbing, the element $2$ is absorbing for all odd natural numbers, and $2\cdot 2=0$. – Johannes Huisman Sep 19 '18 at 08:50
4

Too long for a comment:

@MikeMiller: you seem to suggest that the maps with even lift are homotopically trivial on $\mathbf P^2(\mathbf R)$, which is not the case, I think: let $\pi$ be the quotient map $S^2\rightarrow \mathbf P^2(\mathbf R)$ and let $g\colon\mathbf P^2(\mathbf R)\rightarrow S^2$ be the map that contracts the $1$-cell to a point. Let $\tilde f$ be the composition $g\circ \pi$. It is clearly an even map from $S^2$ to itself. Allthough it is of degree $0$, it is not homotopically trivial between all even maps from $S^2$ to itself. Indeed, otherwise $g$ would be homotopically trivial as well, which it is not; $g$ is the unique homotopically nontrivial map from $\mathbf P^2(\mathbf R)$ to $S^2$. It follows that the map $f$ from $\mathbf P^2(\mathbf R)$ to itself induced by $\tilde f$, i.e., $f=\pi\circ g$, is homotopically nontrivial.

1

Just to supplement Mike's wonderful answer with the existence of the odd odd degree maps(this didn't fit into a comment).

See $S^2=\mathbb{R}/2\pi\mathbb{Z}\times [-1,1]/(x,\pm 1)\sim(y,\pm 1)$ (i.e. the suspension of $S^1$) . The map $f:S^2\rightarrow S^2$ being odd means

$$ f(\theta+\pi,-t)=(f_1(\theta,t)+\pi,-f_2(\theta,t)). $$

Then a degree $(2n+1)$ map is just $f(\theta,t)=((2n+1)\theta,-t)$. This is odd

$$ f(\theta+\pi,-t)=((2n+1)\theta+(2n+1)\pi,t)=((2n+1)\theta+\pi,t)=(f_1(\theta,t)+\pi,-f_2(\theta,t)). $$

Easier: Take $S^2\subset \mathbb C\times \mathbb R$ and define

$f(z,t)=(z^{2n+1},t)$. Then $f(-z,-t)=( (-z)^{2n+1},-t)=-(z^{2n+1},t)$.

Thomas Rot
  • 10,399
  • 1
    I'm a little confused. I think your definition of $S^2$ (which should start with $\Bbb R/2\pi \Bbb Z$ should probably be $(x,-1) \sim (y,-1)$ and the same with +1 - otherwise I think this gives you the torus. And then I think your map should be $f_n(\theta,t) = ((2n+1)\theta,-t)$, or else the map preserves the north and south pole. But other than the minor corrections this looks great. –  Nov 02 '15 at 23:46
  • @MikeMiller reading back I'm also confused... I did this correctly in my notes, but then typed it in mobile somewhere else... Thanks again! – Thomas Rot Nov 03 '15 at 08:30
  • 1
    @MikeMiller and ThomasRot : I do not understand why $((2n+1)\theta,t)$ would not be a good definition of an odd degree map on $S^2$. An odd map of $S^2$ into itself may very well preserve the North and South pole; e.g. the identity map on $S^2$ is perfectly odd! – Johannes Huisman Sep 12 '18 at 12:21
  • @JohannesHuisman: right – Thomas Rot Sep 12 '18 at 15:37
  • Thomas Rot's "easier" formula is good, but I think it should be normalized: $f(z,t)=(z^{2n+1},t)/|(z^{2n+1},t)|$. Of course, it has degree $2n+1$, although computations are a bit longer. – Jesus RS Nov 24 '19 at 10:02
0

Just an addition to this nice question. Let $f:\mathbb P^2(\mathbb R)\to\mathbb P^2(\mathbb R)$ have an even lifting $\tilde f:\mathbb S^2\to\mathbb S^2$. Then this lifting descends to a map $g:\mathbb P^2(\mathbb R)\to \mathbb S^2$. Now here we know that the homotopy class of $g$ is given by its degree mod $2$. If it is $0$, then $g$ is nullhomotopic; if it is $1$, then $g$ is homotopic to Huisman's mapping that can be realized as follows $$ g(x_0:x_1:x_2)\mapsto\tfrac{1}{\|x\|^2}(2x_0^2-\|x\|^2,2x_0x_1,2x_0x_2). $$ The even lift $\tilde f$ has always degree $0$, hence it is nullhomotopic. In case $g$ is not, we have an example of a homotopy of even maps (from $\tilde f$ to a constant map) which necessarily fails to be an even map (because it cannot descend to $\mathbb P^2(\mathbb R)$.

Jesus RS
  • 4,551