I was just about to write down my answer to an exercise in algebraic topology and I wanted to use the fact that $\pi_1(X)$ only depends on the $2$-skeleton of $X$ for any CW complex $X$. I am very sure we've had this fact in some of the topology lectures I've attended.

However, I'm not sure about what was the basic idea is here - please tell me if my thoughts are right, and if not, I'd like to see a short reasoning for this fact.

First we show that every map of CW complexes is homotopic to a cellular map. (I think the proof of this lemma was quite technical)

But now $\pi_n(X)$ is given by homotopy classes of maps $S^n \rightarrow X$, but a homotopy between such maps is just a map $S^n \times [0,1] \rightarrow X$, which by the lemma is homotopic to a map into the $(n+1)$-skeleton of $X$.

Is this the correct way to view the situation or is it flawed somewhere? I sadly don't remember when exactly we had proven said fact in a lecture, or I'd try to look it up in my scripts.

  • 1
    $\begingroup$ The intuition seems correct to me. Of course, it's sweeping few pages of technical details under the rug, as you mention yourself :) $\endgroup$ – Marek Jan 10 '14 at 14:26
  • $\begingroup$ Thanks for your comment. I was well aware that whatever the true "proof" was, as often when dealing with homotopies it would be quite lengthy. I just wanted to assure I knew the idea behind it. $\endgroup$ – Louis Jan 10 '14 at 14:35
  • 2
    $\begingroup$ math.cornell.edu/~hatcher/AT/AT.pdf Corollary 4.12 in the Hatcher's book should do the trick. $\endgroup$ – Marek Jan 10 '14 at 14:36

First of all, $\pi_n(X)$ is the set of pointed homotopy classes of maps; the basepoint matters. There's also a problem in your argument. When you replace $H : S^n\times [0, 1] \to X$ by the cellular map $H' : S^n\times[0, 1] \to X$, you get a homotopy from two potentially different maps, i.e. $H$ is a homotopy from $f$ to $g$, but $H'|_{S^n\times\{0\}}$ may not be $f$, it may have changed, so $H'$ is not necessarily a homotopy between $f$ and $g$.

In order to rectify this, let's recall the cellular approximation theorem.

Cellular Approximation Theorem: Let $X$ and $Y$ be CW complexes. Every continuous map $f : X \to Y$ is homotopic to a cellular map $f'$. Moreover, if $A$ is a subcomplex of $X$ such that $f|_A$ is cellular, then one can choose $f'$ such that $f'|_A = f|_A$.

Fix basepoints $p_0 \in S^n$, $x_0 \in X$. Choose a CW complex structure on $S^n$ with zero-skeleton $\{p_0\}$ and a CW complex structure on $X$ with $0$-skeleton containing $x_0$. Let $X^{(k)}$ denote the $k$-skeleton of $X$.

The class $\alpha \in \pi_n(X, x_0)$ is represented by a map $f : S^n \to X$ such that $f(p_0) = x_0$. As $f|_{\{p_0\}}$ is cellular, there is a cellular map $f' : S^n \to X^{(n)} \subseteq X$ such that $f$ is homotopic to $f'$, and $f'_{\{p_0\}} = f|_{\{p_0\}}$, i.e. $f'(p_0) = f(p_0) = x_0$. As $f'(p_0) = x_0$ and $f'$ is homotopic to $f$, $f'$ also represents the class $\alpha$.

If $g : S^n \to X$ is another representative of the class $\alpha$, then by the same argument, $g$ is homotopic to a cellular map $g' : S^n \to X^{(n)} \subseteq X$ with $g'(p_0) = x_0$, so $g'$ also represents the class $\alpha$.

As $f'$ and $g'$ both represent the class $\alpha$, there is a homotopy between them, i.e. a continuous map $H : S^n\times [0, 1] \to X$ such that $H(p, 0) = f'(p)$, $H(p, 1) = g'(p)$, and $H(p_0, t) = x_0$ for all $t \in [0, 1]$. Let $A = S^n\times\{0, 1\}\cup\{p_0\}\times [0, 1]$. As $H|_A$ is already cellular, $H$ is homotopic to a cellular map $H' : S^n\times [0, 1] \to X$ with $H'|_A = H|_A$, i.e. $H'(p, 0) = H(p, 0) = f'(p)$, $H'(p, 1) = H(p, 1) = g'(p)$, and $H'(p_0, t) = H(p_0, t) = x_0$. Therefore $H'$ is a cellular map and a homotopy between $f'$ and $g'$ (as pointed maps).

As $f'$, $g'$, and $H'$ are cellular, $f'(S^n) \subseteq X^{(n)} \subseteq X^{(n+1)}$, $g'(S^n) \subseteq X^{(n)} \subseteq X^{(n+1)}$, and $H'(S^n\times[0,1]) \subseteq X^{(n+1)}$. We obtain a class $\alpha' \in \pi_n(X^{(n+1)}, x_0)$ represented by $f'$ (and $g'$). Conversely, given $\beta' \in \pi_n(X^{(n+1)}, x_0)$, define $\beta = i_*\beta' \in \pi_n(X, x_0)$ where $i : X^{(n+1)} \to X$ is the inclusion map, i.e. if $\beta'$ is represented by $h$, then $\beta$ is represented by $i\circ h$. These two correspondences are inverses of each other, so we see that $\pi_n(X, x_0) = \pi_n(X^{(n+1)}, x_0)$; in particular, the fundamental group of $X$ only depends on its $2$-skeleton $X^{(2)}$.

More generally, the same argument shows that if $X$ is a $k$-dimensional CW complex and $Y$ is a CW complex, then $[X, Y] = [X, Y^{(k+1)}]$. If $X$ and $Y$ are pointed, then $[X, Y]_{\bullet} = [X, Y^{(k+1)}]_{\bullet}$.


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.