A corollary of the Kan and Thurston theorem states that the space $X$ (path connected) can be recovered, up to homotopy, by applying the fiber-wise $\mathbb{Z}$-completion functor of Bousfield–Kan (notation: $\dot{\mathbb{Z}}_\infty$) to the fibration $$K(P, 1)\to TX = K(G, 1)\to K(G/P, 1)$$ where $TX$ is a Kan–Thurston construction of $X$ (see references), $G/P = \pi_1(X)$, $P$ is a perfect group (i. e. $H_1(P; \mathbb{Z}) = 0$, trivial coefficients) and groups $P$, $G$ and $G/P$ form the exact sequence $$1\to P \to G \to G/P \to 1.$$

For understanding why $\dot{\mathbb{Z}}_\infty K(G, 1)\cong X$, I want to use the Whitehead theorem:

Let $f: X\to Y$ be a map of pointed connected $CW$-complexes such that $f$ induces an isomorphism on fundamental groups and on homology with any local coefficient system. Then $f$ is a homotopy equivalence.

As $P$ is perfect, the space $\mathbb{Z}_\infty K(P, 1)$ is simply-connected. Hence, from the long exact sequence of fibration $$\mathbb{Z}_\infty K(P, 1)\to \dot{\mathbb{Z}}_\infty K(G, 1)\to K(G/P, 1)$$ we have that the Kan–Thurston map $t: TX=K(G, 1)\to X$ induces an isomorphism $\pi_1(\dot{\mathbb{Z}}_\infty K(G, 1))\cong \pi_1(X)$. Moreover, by the main Kan–Thurston result, $t$ induces isomorphisms $H_\ast(TX; \mathcal{A}) \cong H_\ast(X; \mathcal{A})$.

Now consider a commutative diagram $\require{AMScd}$ \begin{CD} K(P, 1) @>>> K(G, 1) @>>> K(G/P, 1) \\ @VVV @VVV @| \\ \mathbb{Z}_\infty K(P, 1) @>>> \dot{\mathbb{Z}}_\infty K(G, 1) @>>> K(G/P, 1). \end{CD}

The morphism of fibrations in the diagram induces a morphism of spectral sequences. And by Zeeman's comparison theorem it is sufficient to prove isomorphism $H_\ast(\mathbb{Z}_\infty K(P, 1); \mathcal{A})\cong H_\ast(K(P, 1); \mathcal{A})$ for any local coefficient system $\mathcal{A}$.

It is known that the Bousfield–Kan $R$-completion ($R\subset \mathbb{Q}$ or $R = \mathbb{Z}/p$) of a space $Y$ with $R$-perfect fundamental group ($R\otimes H_1(\pi_1; \mathbb{Z}) = 0$) is $R$-good, that is, $H_\ast(R_\infty Y; R)\cong H_\ast(Y; R)$. It implies the desired isomorphism in the case of $\mathcal{A} = \mathbb{Z}$. But I don't know why $H_\ast(\mathbb{Z}_\infty K(P, 1); \mathcal{A})\cong H_\ast(P; \mathcal{A})$ for any local coefficient system $\mathcal{A}$. Moreover, it looks weird that $\mathcal{A}$ on $\mathbb{Z}_\infty K(P, 1)$ is trivial (thanks to $\pi_1(\mathbb{Z}_\infty K(P, 1)) = 0$) but we can't say the same about the local system on $K(P, 1)$….


  1. D. Kan and W. Thurston, Every connected space has the homology of a $K(\pi, 1)$, Topology Vol. 15. pp. 253–258, 1976.
  2. A. K. Bousfield and D. M. Kan, Homotopy limits, completions and localizations, Lecture Notes in Math. 304, Springer, 1972.

1 Answer 1


The Kan-Thurston construction depends not just on the homotopy type of $X$, but it depends very heavily on the exact choice of cell structure for $X$. However, it does have some naturality properties. In particular, if $\widetilde X$ is the universal covering space of $X$, then $T\widetilde X$ admits a free action of $G/P=\pi_1(X)$ such that the quotient is isomorphic to $TX$. In particular, if you pull back your first fibration along the universal covering map $\widetilde X\rightarrow X$, you get a fibration $K(P,1)\rightarrow T(\widetilde X)\rightarrow E(G/P)$, where the base space is contractible. I would use this fibration to deduce that $\mathbb{Z}_\infty K(P,1)$ is homotopy equivalent to $\widetilde X$.
Incidentally, any local coefficient system on $K(P,1)$ that is induced by pulling back a local coefficient system on $X$ is trivial, because $P$ is in the kernel of the map to $\pi_1(X)$.

  • $\begingroup$ Could you explain why $T(\widetilde{X})$ admits a free action of $\pi_1(X)$? $\endgroup$ Feb 18 at 15:13
  • $\begingroup$ The construction $Y\mapsto T(Y)$ can be designed in such a way that it is natural for maps of simplicial complexes that do not collapse any simplices; in particular for simplicial automorphisms. The free action of $\pi_1(X)$ on $\widetilde{X}$ thus induces an action of $\pi_1(X)$ on $T(\widetilde{X})$ too. $\endgroup$
    – IJL
    Feb 18 at 21:54

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.