Eilenberg-Mac Lane space



Homotopy theory



An Eilenberg–Mac Lane space is a connected topological space with nontrivial homotopy groups only in a single degree.


For GG a group, the Eilenberg–Mac Lane space K(G,1)K(G,1) is the image under the homotopy hypothesis Quillen equivalence :GrpdTop|-| : \infty Grpd \to Top of the one-object groupoid BG\mathbf{B}G whose hom-set is GG:

K(G,1)=BG. K(G,1) = | \mathbf{B} G | \,.

If GG is a group and n1n \geq 1 then an Eilenberg–Mac Lane space K(G,n)K(G,n) is a connected space with its only non-trivial homotopy group being GG in dimension nn, thus GG must necessarily be abelian for n2n \geq 2.

The construction of such a space can be given for n2n \geq 2 using the standard Dold-Kan correspondence between chain complexes and simplicial abelian groups: let C(G,n)C(G,n) be the chain complex which is GG in dimension nn and trivial elsewhere; the geometric realisation of the corresponding simplicial abelian group is then a K(G,n)K(G,n).

We can include the case n=1n=1 when GG may be nonabelian, by regarding C(G,n)C(G,n) as a crossed complex. Its classifying space B(C(G,n))B(C(G,n)) is then a K(G,n)K(G,n). (This also includes the case n=0n=0 when GG is just a set!) This method also allows for the construction of K(M,n;G,1)K(M,n;G,1) where GG is a group, or groupoid, and MM is a GG-module. This gives a space with π 1=G\pi_1 =G, π n=M\pi_n=M all other homotopy trivial, and with the given operation of π 1\pi_1 on π n\pi_n.

For AA an abelian group, the Eilenberg–Mac Lane space K(A,n)K(A,n) is the image of the ∞-groupoid B nA\mathbf{B}^n A that is the strict ω-groupoid given by the crossed complex [B nA][\mathbf{B}^n A] that is trivial everywhere except in degree nn, where it is AA:

[B nA] =([B nA] n+1[B nA] n[B nA] n1[B nA] 1[B nA] 0) =(*A***). \begin{aligned} [\mathbf{B}^n A] &= ( \cdots \to [\mathbf{B}^n A]_{n+1} \to [\mathbf{B}^nA]_{n} \to [\mathbf{B}^n A]_{n-1} \cdots \to [\mathbf{B}^n A]_{1} \stackrel{\to}{\to} [\mathbf{B}^n A]_{0}) \\ &= ( \cdots \to {*} \to A \to {*} \cdots \to {*} \stackrel{\to}{\to} {*}) \end{aligned} \,.


K(A,n)=B nA. K(A,n) = |\mathbf{B}^n A| \,.

Therefore Eilenberg–Mac Lane spaces constitute a spectrum: the Eilenberg–Mac Lane spectrum.

In general, if AA is an abelian topological group, then there exist a model for the classifying space A\mathcal{B}A which is an abelian topological group. Iterating this construction, one has a notion of nA\mathcal{B}^n A and a model for it which is an abelian topological group. If moreover AA is discrete, then A=BA=K(A,1)\mathcal{B}A=|\mathbf{B}A|=K(A,1), and one inductively sees that nA=B nA=K(A,n)\mathcal{B}^n A=|\mathbf{B}^n A|=K(A,n). Therefore one has a model for K(A,n)K(A,n) which is an abelian topological group.


One common use of Eilenberg–Mac Lane spaces is as coefficient objects for “ordinary” cohomology.

The nnth “ordinary” cohomology of a topological space XX with coefficients in GG (when n=1n=1) or AA (generally) is the collection of homotopy classes of maps from XX into K(G,1)K(G,1) or K(A,n)K(A,n), respectively:

H 1(X,G)=Ho Top(X,K(G,1))=Ho Grpd(X,BG) H^1(X,G) = Ho_{Top}(X, K(G,1)) = Ho_{\infty Grpd}(X, \mathbf{B} G)
H n(X,A)=Ho Top(X,K(A,n))=Ho Grpd(X,B nA). H^n(X,A) = Ho_{Top}(X, K(A,n)) = Ho_{\infty Grpd}(X, \mathbf{B}^n A) \,.

Here on the right Ho TopHo_{Top} and Ho GrpdHo_{\infty Grpd} denotes the homotopy category of the (∞,1)-categories of topological spaces and of ∞-groupoids, respectively.

Not only the set π 0Top(X,K(A,n))=Ho Top(X,K(A,n))\pi_0\mathbf{Top}(X, K(A,n))=Ho_{Top}(X, K(A,n)) is related to the cohomology of XX with coefficients in AA, but also the higher homotopy groups π iTop(X,K(A,n))\pi_i\mathbf{Top}(X, K(A,n)) are, and in the most obvious way: if XX is a connected CW-complex, then

H ni(X,A)=π iTop(X,K(A,n))=π iGrpd(X,B nA), H^{n-i}(X,A)=\pi_i\mathbf{Top}(X, K(A,n))=\pi_i\mathbf{\infty Grpd}(X, \mathbf{B}^n A),

for any choice of base point on the right hand sides. This fact, which appears to have first been remarked by Thom and Federer, is an immediate consequence of the natural homotopy equivalences

ΩH(X,Y)H(X,ΩY) \Omega\mathbf{H}(X,Y)\simeq \mathbf{H}(X,\Omega Y)


ΩK(A,n)K(A,n1) \Omega K(A,n)\simeq K(A,n-1)

one has in every (,1)(\infty,1)-topos, see loop space object. For GG a nonabelian group, Gottlieb proves the following nonabelian analogue of the above result: let XX be a finite dimensional connected CW-complex; for a fixed map f:XK(G,1)f:X\to K(G,1), let C fC_f be the centralizer in G=π 1K(G,1)G=\pi_1 K(G,1) of f *(π 1(X))f_*(\pi_1(X)). Then the connected component of ff in Top(X,K(G,1))\mathbf{Top}(X,K(G,1)) is a K(C f,1)K(C_f,1).

Notice that for GG a nonabelian group, H 1(X,G)H^1(X,G) is a simple (and the most familiar) example of nonabelian cohomology. Nonabelian cohomology in higher degrees is obtained by replacing here the coefficient \infty-groupoids of the simple for B nA\mathbf{B}^n A with more general \infty-groupoids.


The notion of Eilenberg–Mac Lane object makes sense in every (,1)(\infty,1)-topos, not just in L wheL_{whe}Top. See at Eilenberg-MacLane object.


Revised on April 16, 2014 03:31:18 by Urs Schreiber (