Skip to main content

Section 13.3 Uniqueness

We have constructed one \(K(G, n)\text{.}\) We now show that any two such spaces are homotopy equivalent.

Proof.

Define \(f\) on the \(0\)-skeleton by sending the wedge point to the basepoint of \(Y\text{.}\) For each generator \(\alpha\text{,}\) let \(i_\alpha \colon S^n_\alpha \hookrightarrow X\) be the inclusion and extend \(f\) over \(S^n_\alpha\) by a representative of \(\psi([i_\alpha]) \in \pi_n(Y)\text{.}\) This defines \(f\) on the wedge \(\bigvee_\alpha S^n_\alpha\text{.}\)
Write \(W = \bigvee_\alpha S^n_\alpha\) and \(j \colon W \hookrightarrow X\text{.}\) The homomorphisms \(f_*\) and \(\psi \circ j_*\) from \(\pi_n(W)\) to \(\pi_n(Y)\) agree on the generators \([i_\alpha]\text{,}\) so they are equal. To extend over a cell \(e^{n+1}_\beta\) with attaching map \(\varphi_\beta \colon S^n \to W\text{,}\) we need \(f \circ \varphi_\beta\) to be null-homotopic. But
\begin{equation*} [f \circ \varphi_\beta] \;=\; f_*[\varphi_\beta] \;=\; \psi(j_*[\varphi_\beta]) \;=\; 0, \end{equation*}
since \(\varphi_\beta\) extends over the cell, so \(j_*[\varphi_\beta] = 0\) in \(\pi_n(X)\text{.}\) Hence \(f\) extends over \(X\text{.}\) The extension induces \(\psi\) on \(\pi_n\text{,}\) because it agrees with \(\psi\) after composing with \(j_*\text{,}\) and \(j_*\) is surjective since \((X, W)\) is \(n\)-connected (Lemma 9.2).

Proof.

It suffices to show that the CW complex \(K\) constructed in the proof of Theorem 13.10 is homotopy equivalent to every CW complex \(K'\) of type \(K(G, n)\text{.}\) In particular, \(K_{n+1} = X = \bigvee_\alpha S^n_\alpha \cup \bigcup_\beta e^{n+1}_\beta\text{.}\) By Lemma 13.12, the chosen isomorphism \(\pi_n(X) \cong \pi_n(K) \cong \pi_n(K')\) is realized by a map \(f \colon X \to K'\) inducing this isomorphism on \(\pi_n\text{.}\)
We extend \(f\) over the higher cells of \(K\) inductively. For each \((n+2)\)-cell of \(K\text{,}\) the composite of its attaching map with \(f\) lies in \(\pi_{n+1}(K') = 0\text{,}\) so it is null-homotopic and \(f\) extends over the cell. The same argument applies to every higher-dimensional cell, since \(\pi_i(K') = 0\) for all \(i > n\text{.}\) The resulting map \(f \colon K \to K'\) induces an isomorphism on \(\pi_n\) (by construction) and on every other homotopy group (both sides being zero), so \(f\) is a weak homotopy equivalence and hence, by Whitehead’s Theorem (Theorem 8.11), a homotopy equivalence.

Remark 13.14. Moore spaces.

There is a homology version of an Eilenberg–MacLane space, known as a Moore space. For an abelian group \(G\) and \(n \geq 1\) there exists a CW complex \(M(G, n)\) with
\begin{equation*} \widetilde{H}_k\bigl(M(G, n)\bigr) \cong \begin{cases} G & k = n, \\ 0 & \text{otherwise} \end{cases} \end{equation*}
(Exercise 13.19). Moore spaces can in fact be used to construct Eilenberg–MacLane spaces. For a based space \(X\text{,}\) let \(\mathrm{SP}^m(X) := X^{\times m}/\Sigma_m\) be the \(m\)th symmetric product, and let the infinite symmetric product \(\mathrm{SP}^\infty(X)\) be the colimit of the \(\mathrm{SP}^m(X)\) along the maps that add the basepoint as an extra coordinate. The Dold–Thom theorem says that for any connected CW complex \(X\) we have \(\pi_n\bigl(\mathrm{SP}^\infty(X)\bigr) \cong \widetilde{H}_n(X)\) for all \(n\text{.}\) In particular, \(\mathrm{SP}^\infty\bigl(M(G, n)\bigr)\) is a \(K(G, n)\text{.}\)