I'm preparing for an exam by looking at an old exam archive. I encountered this problem which I believe should be solvable using only material from James Munkres' Topology or possibly John Lee's Introduction to Smooth Manifolds as that is how these exams are designed:

"Let $\mathbb{R}\mathbb{P}^n$ denote the $n$-dimensional projective space where $n$ is an odd integer. Suppose that a continuous map $f:\mathbb{R}\mathbb{P}^n \to \mathbb{R}\mathbb{P}^n$ induces the trivial homomorphism

$$f_*:\pi_1(\mathbb{R}\mathbb{P}^n,x_0) \to \pi_1(\mathbb{R}\mathbb{P}^n,f(x_0)).$$

Show that $f$ has a fixed point."

I was thinking that perhaps $n$ being odd is important mainly because $\mathbb{R}\mathbb{P}^n$ is orientable in those cases. My attempt so far: lift $f$ to a map $\hat{f}:\mathbb{R}\mathbb{P}^n \to S^n$ which is possible since $f$ is trivial and so $f_*(\pi_1(\mathbb{R}\mathbb{P}^n)) \subset p_*(\pi_1(S^n))$ where $p:S^n \to \mathbb{R}\mathbb{P}^n$ is the usual quotient map. From there, I was considering $\hat{f} \circ p: S^n \to S^n$ and trying to say something about the degree of the map or using something like Brouwer Fixed Point Theorem (first extend $f$ because it is nullhomotopic).

So far, some friends have suggested using suspensions (as in Hatcher) or the Lefschetz fixed point theorem but I think these are too advanced and wouldn't score points on the exam. Any hints would be a huge help; thank you.

Edit: I've written up a possible solution below; let me know what you think.

  • $\begingroup$ Can you sketch a solution using Lefschetz fixed point theorem? Would you be using the cohomology ring of $\mathbf RP^n$ in $\mathbb Z$ coefficients? $\endgroup$ May 29 '18 at 2:30

I think I have a solution (to my own question) which uses sufficiently elementary theory. I owe some thanks to friends for it.

Suppose that $f$ has no fixed point. Since $f_*$ is trivial, as I said in the question details, it lifts to a map $\hat{f}:\mathbb{R}\mathbb{P}^n \to S^n$. If we let $g = \hat{f} \circ p$, then $g: S^n \to S^n$ is a map such that $p \circ g = f \circ p$.

Claim: If $f$ has no fixed point, neither does $g$ on the $n$-sphere.

  • For, if $g(x)=x$, then $p \circ g(x) = [x] = f \circ p(x) = f[x] = [x]$ (equivalence class $x \sim -x$). Then $f$ has a fixed point which is a contradiction.

Claim: $g$ is homotopic to the antipodal map $\alpha(x) = -x$.

  • Let $H:S^n \times I \to S^n$ be defined by

$$H(x,t) = \frac{(1-t)g(x) + t \alpha(x)}{\|(1-t)g(x) + t \alpha(x)\|}.$$

The denominator is never zero because for it to be zero, we must have that $(1-t)g(x) -tx = 0 \Rightarrow g(x) = \frac{t}{1-t}x$. Since $\|g(x)\| = 1 = \|x\|$, we need $t = 1/2$. But then $g(x) = x$ and $g$ has no fixed points; a contradiction. Thus, $H$ is a homotopy for $g$ to $\alpha$.

Claim: When $n$ is odd, $\mathbb{R}\mathbb{P}^n$ is orientable and the antipodal map on $S^n$ is homotopic to the identity map.

  • I won't prove the first claim; I think it's a standard fact. As for the second, it is proved in the usual proof for the Hairy Ball Theorem (See Problem 16-6 in Lee's Smooth Manifolds). It relies on the fact that when $n$ is odd, there exists a nonvanishing vector field $V$ on $S^n$. We can make a homotopy from $\alpha$ to $V$ and from $V$ to the identity map.

Thus, if $f$ has no fixed points, $g$ is homotopic to the identity. Now, the degree of a map is homotopy invariant and since $\text{id}$ has degree 1, $\text{deg} \,g= 1$. Now, since $f \circ p = p \circ g$, $\text{deg}(f \circ p) = (\text{deg} \, f)(\text{deg} \, p) = (\text{deg} \, p)(\text{deg} \, g) = \text{deg}(p \circ g)$. So $\text{deg} \, f = \text{deg} \, g = 1$. By the way, degree is well-defined here since $\mathbb{R}\mathbb{P}^n, S^n$ are compact, orientable manifolds of the same dimension.

Claim: If $f_* =0$, then $\text{deg} \, f$ is even.

  • We consider the map $f^*: H^1(\mathbb{R}\mathbb{P}^n, \mathbb{Z}_2) \to H^1(\mathbb{R}\mathbb{P}^n, \mathbb{Z}_2)$. Now, by the Hurewicz theorem, $H_1(X) = \text{Ab}(\pi_1(X))$. Since $\pi_1(\mathbb{R}\mathbb{P}^n) = \mathbb{Z}_2$ is already abelian, then $H^1(\mathbb{R}\mathbb{P}^n, \mathbb{Z}_2) = \text{Hom}(H_1(\mathbb{R}\mathbb{P}^n, \mathbb{Z}_2), \mathbb{Z}_2) = \text{Hom}(\mathbb{Z}_2,\mathbb{Z}_2) = \mathbb{Z}_2$. So $f^*: \mathbb{Z}_2 \to \mathbb{Z}_2$. I want to say from this that $f^* = 0$ because $f_* = 0$.

If this is correct, then the conclusion is: $\text{deg} \, f = 1$ but should also be even. This is a contradiction and thus, $f$ has a fixed point.

  • 1
    $\begingroup$ I'm not sure about your last claim. But you don't need it. You've proven that $\deg f = 1$. Now, argue that $f^\ast:H^1(\mathbb{R}P^n;\mathbb{Z}/2\mathbb{Z})\rightarrow H^1(\mathbb{R}P^n; \mathbb{Z}/2\mathbb{Z})$ is the $0$ map because $f_\ast$ is the $0$ map. Use the ring structure of $H^\ast(\mathbb{R}P^n;\mathbb{Z}/2\mathbb{Z})$ to conclude the mod $2$ degree of $f$ is $0$. In other words, the degree of $f$ must be even, contradicting the fact that $\deg f = 1$. $\endgroup$ May 30 '18 at 3:25
  • $\begingroup$ @JasonDeVito I'm having some trouble arguing $f^* = 0$ and also that the degree of $f$ is even by looking at the ring structure. $\endgroup$
    – inkievoyd
    May 30 '18 at 15:35
  • $\begingroup$ I've added an answer addressing your comment. Let me know if I can further help! $\endgroup$ May 30 '18 at 16:01

(This meant to answer a question in the comments)

Let $X:=\mathbb{R}P^n$ and let $G:=\mathbb{Z}/2\mathbb{Z}$, just to save a bit of typing.

Lemma 1: If $f_\ast:\pi_1(X)\rightarrow \pi_1(X)$ is the $0$ map, then so is $f_\ast:H_1(X)\rightarrow H_1(X)$.

Proof: $H_1$ is the abelianization of $\pi_1$, which is already abelian. Further, the abelianization is natural, so $f_\ast$ is $0$.

Lemma 2: If $f_\ast:H_1(X)\rightarrow H_1(X)$ is the $0$ map, then so is $f^\ast H^1(X;G)\rightarrow H^1(X;G)$.

Proof. From universal coefficients, we know that $H^1(X;G) \cong Hom(H_1(X), G)\oplus Ext(H_0(X),G)$. Since $H_0(X)$ is free, $Ext(H_0(X),G) = 0$. The the isomorphism $H^1(X;G)\rightarrow Hom(H_1(X),G)$ is natural.

Now, $f_\ast:H_1(X)\rightarrow H_1(X)$ induces a map $\hat{f}$ on $Hom(H_1(x),G)$ given by $\hat{f}(g) = g\circ f_\ast$. If $g\in Hom(H_1(X), G)$ is the unique nontrival map sending $1\in G$ to $1\in G$, then $\hat{f}(g)(1) = g(f_\ast(1)) = g(0) = 0$, so $\hat{f}(g) = 0$. It follows that $\hat{f}:Hom(H_1(X), G)\rightarrow Hom(H_1(X),G)$ is the $0$ map.

By naturality of the isomorphism $H^1(X;G)\rightarrow Hom(H_1(X),G)$, $f^\ast$ must be the $0$-map as well.

Lemma 3: Suppose $f:X\rightarrow X$ induces the $0$ map on $H^1(X;G)$. Then $f$ induces the map on $H^i(X;G)$ for any positive $i$.


Recall $H^\ast(X;G)\cong G[x]/x^{n+1}$, where $x\in H^1(X;G)$.

By assumption, $f^\ast(x) = 0$. Then $f^\ast(x^k) = f^\ast(x)^k = 0^k = 0$, so $f^\ast$ is the $0$ map on anything $x$ generates. But $x$ generates all positive degree cohomology.


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.