Let $X$ and $Y$ be topological spaces. If $f\colon X\to Y$ is a homeomorphism, then it induces an isomorphism $f_\sharp\colon\pi_1(X,x_0)\to \pi_1(Y,f(x_0))$. All good.

As far as I know, the result still holds if we weaken "homeomorphism" to "homotopy equivalence", but I am having trouble checking that. I know that $(f_2\circ f_1)_\sharp = (f_2)_\sharp\circ (f_1)_\sharp$, $({\rm id}_X)_\sharp = {\rm id}_{\pi_1(X,x_0)}$ and that $f_1\simeq f_2 ~ {\rm rel}\,\{x_0\}$ implies $(f_1)_\sharp = (f_2)_\sharp$.

If $f\colon X \to Y$ is a homotopy equivalence, then we have $g\colon Y\to X$ with $g\circ f \simeq {\rm id}_X$ and $f\circ g\simeq {\rm id}_Y$.

My problems are:

  • I can't guarantee that $g(f(x_0))=x_0$;
  • These homotopies don't need to be ${\rm rel}\,\{x_0\}$ and ${\rm rel}\,\{f(x_0)\}$, respectively.

If I have these two assumptions, I can make: $${\rm id}_{\pi_1(X,x_0)} = ({\rm id}_X)_\sharp = (g\circ f)_\sharp = g_\sharp\circ f_\sharp$$and repeat this for $f_\sharp\circ g_\sharp$.

How can I go around these problems?

  • $\begingroup$ This is proved somewhere in chapter 1 of Hatcher. Search "homotopy equivalence" and you'll surely find it. $\endgroup$
    – user98602
    Aug 15 '15 at 21:31

Let $H$ be a homotopy from $g\circ f$ to ${\rm id}_X$ and take a path $\gamma(t) = H(x_0, t)$, i.e. $\gamma(0) = g(f(x_o))$ and $\gamma(1) = x_0$. Consider

$\pi_1(X, x_0) \to \pi_1(Y, f(x_0)) \to \pi_1(X, g \circ f(x_0)) \to π_1(X, x_0)$,

where the first map is $f_{\sharp}$, the second $g_{\sharp}$ and the third is the one induced by $\gamma$. Hence $[\lambda] \in \pi_1(X, x_0)$ gets mapped to $[\gamma^{-1} g f \lambda \gamma]$. One immediately writes down (with the help of $H$) a homotopy from $\lambda$ to $g f \lambda$. Use it to write down a homotopy from $\lambda$ to $\gamma^{-1} g f \lambda \gamma$ concluding that the composition of those maps is indeed the identity.

  • $\begingroup$ Thanks, I get the idea. Writing $\widetilde{H}(s,t) = H(\lambda(s),t)$ gives $g\circ f \circ \lambda \simeq \lambda$, but this $\simeq$ is not ${\rm rel}\,\{0,1\}$, shouldn't this be a problem? Or am I doing something wrong? +1 of course $\endgroup$
    – Ivo Terek
    Aug 15 '15 at 22:12
  • $\begingroup$ The point is that you build a homotopy (without fixing endpoints) just to obtain that the composition of those maps is the identity, which is bijective. But if a composition of maps is bijective (i.e. injective + surjective) what can you say about the "compositioned" maps? $\endgroup$
    – M.U.
    Aug 16 '15 at 11:39
  • $\begingroup$ Yes, sure, but that's not my problem. My problem is: to prove that the composition is the identity, I must get $[\gamma^{-1}gf\lambda\gamma]=[\lambda]$, that is, $\gamma^{-1}gf\lambda\gamma\simeq\lambda ~{\rm rel}\,\{0,1\}$, and the suggestion gives $gf\lambda \simeq \lambda $, but not ${\rm rel}\,\{0, 1\}$ $\endgroup$
    – Ivo Terek
    Aug 16 '15 at 14:15
  • $\begingroup$ In my head, since the fundamental groups are classes from the relation $\simeq ~{\rm rel}\,\{0, 1\}$, we need to keep the endpoints fixed all the time $\endgroup$
    – Ivo Terek
    Aug 16 '15 at 14:19
  • $\begingroup$ Oh, wait, I think I got it. You're saying that even if $\widetilde{H}$ doesn't keep endpoints fixed, when I do the next homotopy, the composition will keep them fixed alright? $\endgroup$
    – Ivo Terek
    Aug 16 '15 at 14:21

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.