A crossed module consists of a homomorphism of groups \(\partial\colon M\rightarrow G\) together with an action \((g,m)\mapsto\, {^gm}\) of \(G\) on \(M\) satisfying
\(\partial(^gm) = gmg^{-1}\)
\(^{\partial m}m' = mm'm^{-1}\)
for \(g\in G\), \(m,m'\in M\).
A crossed module \(\partial\colon M\rightarrow G\) is equivalent to a cat\(^1\)-group \((H,s,t)\) (see 6.9) where \(H=M \rtimes G\), \(s(m,g) = (1,g)\), \(t(m,g)=(1,(\partial m)g)\). A cat\(^1\)-group is, in turn, equivalent to a simplicial group with Moore complex has length \(1\). The simplicial group is constructed by considering the cat\(^1\)-group as a category and taking its nerve. Alternatively, the simplicial group can be constructed by viewing the crossed module as a crossed complex and using a nonabelian version of the Dold-Kan theorem.
The following example concerns the crossed module
\(\partial\colon G\rightarrow Aut(G), g\mapsto (x\mapsto gxg^{-1})\)
associated to the dihedral group \(G\) of order \(16\). This crossed module represents, up to homotopy type, a connected space \(X\) with \(\pi_iX=0\) for \(i\ge 3\), \(\pi_2X=Z(G)\), \(\pi_1X = Aut(G)/Inn(G)\). The space \(X\) can be represented, up to homotopy, by a simplicial group. That simplicial group is used in the example to compute
\(H_1(X,\mathbb Z)= \mathbb Z_2 \oplus \mathbb Z_2\),
\(H_2(X,\mathbb Z)= \mathbb Z_2 \),
\(H_3(X,\mathbb Z)= \mathbb Z_2 \oplus \mathbb Z_2 \oplus \mathbb Z_2\),
\(H_4(X,\mathbb Z)= \mathbb Z_2 \oplus \mathbb Z_2 \oplus \mathbb Z_2\),
\(H_5(X,\mathbb Z)= \mathbb Z_2 \oplus \mathbb Z_2 \oplus \mathbb Z_2 \oplus \mathbb Z_2\oplus \mathbb Z_2\oplus \mathbb Z_2\).
gap> C:=AutomorphismGroupAsCatOneGroup(DihedralGroup(16)); Cat-1-group with underlying group Group( [ f1, f2, f3, f4, f5, f6, f7, f8, f9 ] ) . gap> Size(C); 512 gap> Q:=QuasiIsomorph(C); Cat-1-group with underlying group Group( [ f9, f8, f1, f2*f3, f5 ] ) . gap> Size(Q); 32 gap> N:=NerveOfCatOneGroup(Q,6); Simplicial group of length 6 gap> K:=ChainComplexOfSimplicialGroup(N); Chain complex of length 6 in characteristic 0 . gap> Homology(K,1); [ 2, 2 ] gap> Homology(K,2); [ 2 ] gap> Homology(K,3); [ 2, 2, 2 ] gap> Homology(K,4); [ 2, 2, 2 ] gap> Homology(K,5); [ 2, 2, 2, 2, 2, 2 ]
The following example concerns the Eilenberg-MacLane space \(X=K(\mathbb Z_3,3)\) which is a path-connected space with \(\pi_3X=\mathbb Z_3\), \(\pi_iX=0\) for \(3\ne i\ge 1\). This space is represented by a simplicial group, and perturbation techniques are used to compute
\(H_7(X,\mathbb Z)=\mathbb Z_3 \oplus \mathbb Z_3\).
gap> A:=AbelianGroup([3]);;AbelianInvariants(A); [ 3 ] gap> K:=EilenbergMacLaneSimplicialGroup(A,3,8); Simplicial group of length 8 gap> C:=ChainComplex(K); Chain complex of length 8 in characteristic 0 . gap> Homology(C,7); [ 3, 3 ]
For integer \(n>1\) and abelian group \(A\) the Eilenberg-MacLane space \(K(A,n)\) is better represented as a simplicial free abelian group. (The reason is that the functorial bar resolution of a group can be replaced in computations by the smaller functorial Chevalley-Eilenberg complex of the group when the group is free abelian, obviating the need for perturbation techniques. When \(A\) has torision we can replace it with an inclusion of free abelian groups \(A_1 \hookrightarrow A_0\) with \(A\cong A_0/A_1\) and again invoke the Chevalley-Eilenberg complex. The current implementation unfortunately handles only free abelian \(A\) but the easy extension to non-free \(A\) is planned for a future release.)
The following commands compute the integral homology \(H_n(K(\mathbb Z,3),\mathbb Z)\) for \( 0\le n \le 16\). (Note that one typically needs fewer than \(n\) terms of the Eilenberg-MacLance space to compute its \(n\)-th homology -- an error is printed if too few terms of the space are available for a given computation.)
gap> A:=AbelianPcpGroup([0]);; #infinite cyclic group gap> K:=EilenbergMacLaneSimplicialFreeAbelianGroup(A,3,14); Simplicial free abelian group of length 14 gap> for n in [0..16] do > Print("Degree ",n," integral homology of K is ",Homology(K,n),"\n"); > od; Degree 0 integral homology of K is [ 0 ] Degree 1 integral homology of K is [ ] Degree 2 integral homology of K is [ ] Degree 3 integral homology of K is [ 0 ] Degree 4 integral homology of K is [ ] Degree 5 integral homology of K is [ 2 ] Degree 6 integral homology of K is [ ] Degree 7 integral homology of K is [ 3 ] Degree 8 integral homology of K is [ 2 ] Degree 9 integral homology of K is [ 2 ] Degree 10 integral homology of K is [ 3 ] Degree 11 integral homology of K is [ 5, 2 ] Degree 12 integral homology of K is [ 2 ] Degree 13 integral homology of K is [ ] Degree 14 integral homology of K is [ 10, 2 ] Degree 15 integral homology of K is [ 7, 6 ] Degree 16 integral homology of K is [ ]
For an \(n\)-connected pointed space \(X\) the Freudenthal Suspension Theorem states that the map \(X \rightarrow \Omega(\Sigma X)\) induces a map \(\pi_k(X) \rightarrow \pi_k(\Omega(\Sigma X))\) which is an isomorphism for \(k\le 2n\) and epimorphism for \(k=2n+1\). Thus the Eilenberg-MacLane space \(K(A,n+1)\) can be constructed from the suspension \(\Sigma K(A,n)\) by attaching cells in dimensions \(\ge 2n+1\). In particular, there is an isomorphism \( H_{k-1}(K(A,n),\mathbb Z) \rightarrow H_k(K(A,n+1),\mathbb Z)\) for \(k\le 2n\) and epimorphism for \(k=2n+1\).
For instance, \( H_{k-1}(K(\mathbb Z,3),\mathbb Z) \cong H_k(K(\mathbb Z,4),\mathbb Z) \) for \(k\le 6\) and \( H_6(K(\mathbb Z,3),\mathbb Z) \twoheadrightarrow H_7(K(\mathbb Z,4),\mathbb Z) \). This assertion is seen in the following session.
gap> A:=AbelianPcpGroup([0]);; #infinite cyclic group gap> K:=EilenbergMacLaneSimplicialFreeAbelianGroup(A,4,11); Simplicial free abelian group of length 11 gap> for n in [0..13] do > Print("Degree ",n," integral homology of K is ",Homology(K,n),"\n"); > od; Degree 0 integral homology of K is [ 0 ] Degree 1 integral homology of K is [ ] Degree 2 integral homology of K is [ ] Degree 3 integral homology of K is [ ] Degree 4 integral homology of K is [ 0 ] Degree 5 integral homology of K is [ ] Degree 6 integral homology of K is [ 2 ] Degree 7 integral homology of K is [ ] Degree 8 integral homology of K is [ 3, 0 ] Degree 9 integral homology of K is [ ] Degree 10 integral homology of K is [ 2, 2 ] Degree 11 integral homology of K is [ ] Degree 12 integral homology of K is [ 5, 12, 0 ] Degree 13 integral homology of K is [ 2 ]
The cup product is not implemented for the cohomology ring \(H^\ast(K(\pi,n),\mathbb Z)\). Standard theoretical spectral sequence arguments have to be applied to obtain basic information relating to the ring structure. To illustrate this the following commands compute \(H^n(K(\mathbb Z,2),\mathbb Z)\) for the first few values of \(n\).
gap> K:=EilenbergMacLaneSimplicialFreeAbelianGroup(A,2,10);; gap> List([0..10],k->Cohomology(K,k)); [ [ 0 ], [ ], [ 0 ], [ ], [ 0 ], [ ], [ 0 ], [ ], [ 0 ], [ ], [ 0 ] ]
There is a fibration sequence \(K(\pi,n) \hookrightarrow \ast \twoheadrightarrow K(\pi,n+1)\) in which \(\ast\) denotes a contractible space. For \(n=1, \pi=\mathbb Z\) the terms of the \(E_2\) page of the Serre integral cohomology spectral sequence for this fibration are
\(E_2^{pq}= H^p( K(\mathbb Z,2), H^q(K(\mathbb Z,1),\mathbb Z) )\) .
Since \(K(\mathbb Z,1)\) can be taken to be the circle \(S^1\) we know that it has non-trivial cohomology in degrees \(0\) and \(1\) only. The first few terms of the \(E_2\) page are given in the following table.
| \(1\) | \(\mathbb Z\) | \(0\) | \(\mathbb Z\) | \(0\) | \(\mathbb Z\) | \(0\) | \(\mathbb Z\) | \(0\) | \(\mathbb Z\) | \(0\) | \(\mathbb Z\) |
| \(0\) | \(\mathbb Z\) | \(0\) | \(\mathbb Z\) | \(0\) | \(\mathbb Z\) | \(0\) | \(\mathbb Z\) | \(0\) | \(\mathbb Z\) | \(0\) | \(\mathbb Z\) |
| \(q/p\) | \(0\) | \(1\) | \(2\) | \(3\) | \(4\) | \(5\) | \(6\) | \(7\) | \(8\) | \(9\) | \(10\) |
Let \(x\) denote the generator of \(H^1(K(\mathbb Z,1),\mathbb Z)\) and \(y\) denote the generator of \(H^2(K(\mathbb Z,2),\mathbb Z)\). Since \(\ast\) has zero cohomology in degrees \(\ge 1\) we see that the differential must restrict to an isomorphism \(d_2\colon E_2^{0,1} \rightarrow E_2^{2,0}\) with \(d_2(x)=y\). Then we see that the differential must restrict to an isomorphism \(d_2\colon E_2^{2,1} \rightarrow E_2^{4,0}\) defined on the generator \(xy\) of \(E_2^{2,1}\) by
\[d_2(xy) = d_2(x)y + (-1)^{{\rm deg}(x)}xd_2(y) =y^2\ . \]
Hence \(E_2^{4,0} \cong H^4(K(\mathbb Z,2),\mathbb Z)\) is generated by \(y^2\). The argument extends to show that \(H^6(K(\mathbb Z,2),\mathbb Z)\) is generated by \(y^3\), \(H^8(K(\mathbb Z,2),\mathbb Z)\) is generated by \(y^4\), and so on.
In fact, to obtain a complete description of the ring \(H^\ast(K(\mathbb Z,2),\mathbb Z)\) in this fashion there is no benefit to using computer methods at all. We only need to know the cohomology ring \(H^\ast(K(\mathbb Z,1),\mathbb Z) =H^\ast(S^1,\mathbb Z)\) and the single cohomology group \(H^2(K(\mathbb Z,2),\mathbb Z)\).
A similar approach can be attempted for \(H^\ast(K(\mathbb Z,3),\mathbb Z)\) using the fibration sequence \(K(\mathbb Z,2) \hookrightarrow \ast \twoheadrightarrow K(\mathbb Z,3)\) and, as explained in Chapter 5 of [Hat01], yields the computation of the group \(H^i(K(\mathbb Z,3),\mathbb Z)\) for \(4\le i\le 13\). The method does not directly yield \(H^3(K(\mathbb Z,3),\mathbb Z)\) and breaks down in degree \(14\) yielding only that \(H^{14}(K(\mathbb Z,3),\mathbb Z) = 0 {\rm ~or~} \mathbb Z_3\). The following commands provide \(H^3(K(\mathbb Z,3),\mathbb Z)= \mathbb Z\) and \(H^{14}(K(\mathbb Z,3),\mathbb Z) =0\).
gap> A:=AbelianPcpGroup([0]);; gap> K:=EilenbergMacLaneSimplicialFreeAbelianGroup(A,3,15);; gap> Cohomology(K,3); [ 0 ] gap> Cohomology(K,14); [ ]
However, the implementation of these commands is currently a bit naive, and computationally inefficient, since they do not currently employ any homological perturbation techniques.
The Hurewicz Theorem immediately gives
\[\pi_n(S^n)\cong \mathbb Z ~~~ (n\ge 1)\]
and
\[\pi_k(S^n)=0 ~~~ (k\le n-1).\]
As a CW-complex the Eilenberg-MacLane space \(K=K(\mathbb Z,n)\) can be obtained from an \(n\)-sphere \(S^n=e^0\cup e^n\) by attaching cells in dimensions \(\ge n+2\) so as to kill the higher homotopy groups of \(S^n\). From the inclusion \(\iota\colon S^n\hookrightarrow K(\mathbb Z,n)\) we can form the mapping cone \(X=C(\iota)\). The long exact homotopy sequence
\( \cdots \rightarrow \pi_{k+1}K \rightarrow \pi_{k+1}(K,S^n) \rightarrow \pi_{k} S^n \rightarrow \pi_kK \rightarrow \pi_k(K,S^n) \rightarrow \cdots\)
implies that \(\pi_k(K,S^n)=0\) for \(0 \le k\le n+1\) and \(\pi_{n+2}(K,S^n)\cong \pi_{n+1}(S^n)\). The relative Hurewicz Theorem gives an isomorphism \(\pi_{n+2}(K,S^n) \cong H_{n+2}(K,S^n,\mathbb Z)\). The long exact homology sequence
\( \cdots H_{n+2}(S^n,\mathbb Z) \rightarrow H_{n+2}(K,\mathbb Z) \rightarrow H_{n+2}(K,S^n, \mathbb Z) \rightarrow H_{n+1}(S^n,\mathbb Z) \rightarrow \cdots\)
arising from the cofibration \(S^n \hookrightarrow K \twoheadrightarrow X\) implies that \(\pi_{n+1}(S^n)\cong \pi_{n+2}(K,S^n) \cong H_{n+2}(K,S^n,\mathbb Z) \cong H_{n+2}(K,\mathbb Z)\). From the GAP computations in 12.3 and the Freudenthal Suspension Theorem we find:
\[ \pi_3S^2 \cong \mathbb Z, ~~~~~~ \pi_{n+1}(S^n)\cong \mathbb Z_2~~~(n\ge 3).\]
The Hopf fibration \(S^3\rightarrow S^2\) has fibre \(S^1 = K(\mathbb Z,1)\). It can be constructed by viewing \(S^3\) as all pairs \((z_1,z_2)\in \mathbb C^2\) with \(|z_1|^2+|z_2|^2=1\) and viewing \(S^2\) as \(\mathbb C\cup \infty\); the map sends \((z_1,z_2)\mapsto z_1/z_2\). The homotopy exact sequence of the Hopf fibration yields \(\pi_k(S^3) \cong \pi_k(S^2)\) for \(k\ge 3\), and in particular
\[\pi_4(S^2) \cong \pi_4(S^3) \cong \mathbb Z_2\ .\]
It will require further techniques (such as the Postnikov tower argument in Section 12.7 below) to establish that \(\pi_5(S^3) \cong \mathbb Z_2\). Once we have this isomorphism for \(\pi_5(S^3)\), the generalized Hopf fibration \(S^3 \hookrightarrow S^7 \twoheadrightarrow S^4\) comes into play. This fibration is contructed as for the classical fibration, but using pairs \((z_1,z_2)\) of quaternions rather than pairs of complex numbers. The Hurewicz Theorem gives \(\pi_3(S^7)=0\); the fibre \(S^3\) is thus homotopic to a point in \(S^7\) and the inclusion of the fibre induces the zero homomorphism \(\pi_k(S^3) \stackrel{0}{\longrightarrow} \pi_k(S^7) ~~(k\ge 1)\). The exact homotopy sequence of the generalized Hopf fibration then gives \(\pi_k(S^4)\cong \pi_k(S^7)\oplus \pi_{k-1}(S^3)\). On taking \(k=6\) we obtain \(\pi_6(S^4)\cong \pi_5(S^3) \cong \mathbb Z_2\). Freudenthal suspension then gives
\[\pi_{n+2}(S^n)\cong \mathbb Z_2,~~~(n\ge 2).\]
For any group \(G\) we consider the homotopy groups \(\pi_n(\Sigma K(G,1))\) of the suspension \(\Sigma K(G,1)\) of the Eilenberg-MacLance space \(K(G,1)\). On taking \(G=\mathbb Z\), and observing that \(S^2 = \Sigma K(\mathbb Z,1)\), we specialize to the homotopy groups of the \(2\)-sphere \(S^2\).
By construction,
\[\pi_1(\Sigma K(G,1))=0\ .\]
The Hurewicz Theorem gives
\[\pi_2(\Sigma K(G,1)) \cong G_{ab}\]
via the isomorphisms \(\pi_2(\Sigma K(G,1)) \cong H_2(\Sigma K(G,1),\mathbb Z) \cong H_1(K(G,1),\mathbb Z) \cong G_{ab}\). R. Brown and J.-L. Loday [BL87] obtained the formulae
\[\pi_3(\Sigma K(G,1)) \cong \ker (G\otimes G \rightarrow G, x\otimes y\mapsto [x,y]) \ ,\]
\[\pi_4(\Sigma^2 K(G,1)) \cong \ker (G\, {\widetilde \otimes}\, G \rightarrow G, x\, {\widetilde \otimes}\, y\mapsto [x,y]) \]
involving the nonabelian tensor square and nonabelian symmetric square of the group \(G\). The following commands use the nonabelian tensor and symmetric product to compute the third and fourth homotopy groups for \(G =Syl_2(M_{12})\) the Sylow \(2\)-subgroup of the Mathieu group \(M_{12}\).
gap> G:=SylowSubgroup(MathieuGroup(12),2);; gap> ThirdHomotopyGroupOfSuspensionB(G); [ 2, 2, 2, 2, 2, 2, 2, 2, 2 ] gap> gap> FourthHomotopyGroupOfDoubleSuspensionB(G); [ 2, 2, 2, 2, 2, 2 ]
A Postnikov system for the sphere \(S^3\) consists of a sequence of fibrations \(\cdots X_3\stackrel{p_3}{\rightarrow} X_2\stackrel{p_2}{\rightarrow} X_1\stackrel{p_1}{\rightarrow} \ast\) and a sequence of maps \(\phi_n\colon S^3 \rightarrow X_n\) such that
\(p_n \circ \phi_n =\phi_{n-1}\)
The map \(\phi_n\colon S^3 \rightarrow X_n\) induces an isomorphism \(\pi_k(S^3)\rightarrow \pi_k(X_n)\) for all \(k\le n\)
\(\pi_k(X_n)=0\) for \(k > n\)
and consequently each fibration \(p_n\) has fibre an Eilenberg-MacLane space \(K(\pi_n(S^3),n)\).
The space \(X_n\) is obtained from \(S^3\) by adding cells in dimensions \(\ge n+2\) and thus
\(H_k(X_n,\mathbb Z)=H_k(S^3,\mathbb Z)\) for \(k\le n+1\).
So in particular \(X_1=X_2=\ast, X_3=K(\mathbb Z,3)\) and we have a fibration sequence \(K(\pi_4(S^3),4) \hookrightarrow X_4 \twoheadrightarrow K(\mathbb Z,3)\). The terms in the \(E_2\) page of the Serre integral cohomology spectral sequence of this fibration are
\(E_2^{p,q}=H^p(\,K(\mathbb Z,3),\,H_q(K(\mathbb Z_2,4),\mathbb Z)\,)\).
The first few terms in the \(E_2\) page can be computed using the commands of Sections 12.2 and 12.3 and recorded as follows.
| \(8\) | \(\mathbb Z_2\) | \(0\) | \(0\) | |||||||
| \(7\) | \(\mathbb Z_2\) | \(0\) | \(0\) | |||||||
| \(6\) | \(0\) | \(0\) | \(0\) | |||||||
| \(5\) | \(\pi_4(S^3)\) | \(0\) | \(0\) | \(\pi_4(S^3)\) | \(0\) | \(0\) | \(0\) | \(\) | ||
| \(4\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | ||||
| \(3\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | ||||
| \(2\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | ||
| \(1\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | ||
| \(0\) | \(\mathbb Z\) | \(0\) | \(0\) | \(\mathbb Z\) | \(0\) | \(0\) | \(\mathbb Z_2\) | \(0\) | \(\mathbb Z_3\) | \(\mathbb Z_2\) |
| \(q/p\) | \(0\) | \(1\) | \(2\) | \(3\) | \(4\) | \(5\) | \(6\) | \(7\) | \(8\) | \(9\) |
Since we know that \(H^5(X_4,\mathbb Z) =0\), the differentials in the spectral sequence must restrict to an isomorphism \(E_2^{0,5}=\pi_4(S^3) \stackrel{\cong}{\longrightarrow} E_2^{6,0}=\mathbb Z_2\). This provides an alternative derivation of \(\pi_4(S^3) \cong \mathbb Z_2\). We can also immediately deduce that \(H^6(X_4,\mathbb Z)=0\). Let \(x\) be the generator of \(E_2^{0,5}\) and \(y\) the generator of \(E_2^{3,0}\). Then the generator \(xy\) of \(E_2^{3,5}\) gets mapped to a non-zero element \(d_7(xy)=d_7(x)y -xd_7(y)\). Hence the term \(E_2^{0,7}=\mathbb Z_2\) must get mapped to zero in \(E_2^{3,5}\). It follows that \(H^7(X_4,\mathbb Z)=\mathbb Z_2\).
The integral cohomology of Eilenberg-MacLane spaces yields the following information on the \(E_2\) page \(E_2^{p,q}=H_p(\,X_4,\,H^q(K(\pi_5S^3,5),\mathbb Z)\,)\) for the fibration \(K(\pi_5(S^3),5) \hookrightarrow X_5 \twoheadrightarrow X_4\).
| \(6\) | \(\pi_5(S^3)\) | \(0\) | \(0\) | \(\pi_5(S^3)\) | \(0\) | \(0\) | ||
| \(5\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | |
| \(4\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | |
| \(3\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | |
| \(2\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | |
| \(1\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | |
| \(0\) | \(\mathbb Z\) | \(0\) | \(0\) | \(\mathbb Z\) | \(0\) | \(0\) | \(0\) | \(H^7(X_4,\mathbb Z)\) |
| \(q/p\) | \(0\) | \(1\) | \(2\) | \(3\) | \(4\) | \(5\) | \(6\) | \(7\) |
Since we know that \(H^6(X_5,\mathbb Z)=0\), the differentials in the spectral sequence must restrict to an isomorphism \(E_2^{0,6}=\pi_5(S^3) \stackrel{\cong}{\longrightarrow} E_2^{7,0}=H^7(X_4,\mathbb Z)\). We can conclude the desired result:
\[\pi_5(S^3) = \mathbb Z_2\ .\]
\(~~~\)
Note that the fibration \(X_4 \twoheadrightarrow K(\mathbb Z,3)\) is determined by a cohomology class \(\kappa \in H^5(K(\mathbb Z,3), \mathbb Z_2) = \mathbb Z_2\). If \(\kappa=0\) then we'd have \(X_4 =K(\mathbb Z_2,4)\times K(\mathbb Z,3)\) and, as the following commands show, we'd then have \(H_4(X_4,\mathbb Z)=\mathbb Z_2\).
gap> K:=EilenbergMacLaneSimplicialGroup(AbelianPcpGroup([0]),3,7);; gap> L:=EilenbergMacLaneSimplicialGroup(CyclicGroup(2),4,7);; gap> CK:=ChainComplex(K);; gap> CL:=ChainComplex(L);; gap> T:=TensorProduct(CK,CL);; gap> Homology(T,4); [ 2 ]
Since we know that \(H_4(X_4,\mathbb Z)=0\) we can conclude that the Postnikov invariant \(\kappa\) is the non-zero class in \(H^5(K(\mathbb Z,3), \mathbb Z_2) = \mathbb Z_2\).
Consider the suspension \(X=\Sigma K(G,1)\) of a classifying space of a group \(G\) once again. This space has a Postnikov system in which \(X_1 = \ast\), \(X_2= K(G_{ab},2)\). We have a fibration sequence \(K(\pi_3 X, 3) \hookrightarrow X_3 \twoheadrightarrow K(G_{ab},2)\). The corresponding integral cohomology Serre spectral sequence has \(E_2\) page with terms
\(E_2^{p,q}=H^p(\,K(G_{ab},2), H^q(K(\pi_3 X,3)),\mathbb Z)\, )\).
As an example, for the Alternating group \(G=A_4\) of order \(12\) the following commands of Section 12.6 compute \(G_{ab} = \mathbb Z_3\) and \(\pi_3 X = \mathbb Z_6\).
gap> AbelianInvariants(G); [ 3 ] gap> ThirdHomotopyGroupOfSuspensionB(G); [ 2, 3 ]
The first terms of the \(E_2\) page can be calculated using the commands of Sections 12.2 and 12.3.
| \(7\) | \(\mathbb Z_2 \) | \(0\) | \(\) | \(\) | \(\) | \(\) | ||
| \(6\) | \(\mathbb Z_2\) | \(0\) | \(0\) | \(0\) | \(\) | \(\) | ||
| \(5\) | \(0\) | \(0\) | \(0\) | \(0\) | \(\) | \(\) | \(\) | |
| \(4\) | \(\mathbb Z_6\) | \(0\) | \(0\) | \(\mathbb Z_3\) | \(\) | \(\) | \(\) | |
| \(3\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(\) | |
| \(2\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | |
| \(1\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | |
| \(0\) | \(\mathbb Z\) | \(0\) | \(0\) | \(\mathbb Z_3\) | \(0\) | \(\mathbb Z_3\) | \(0\) | \(\mathbb Z_9\) |
| \(q/p\) | \(0\) | \(1\) | \(2\) | \(3\) | \(4\) | \(5\) | \(6\) | \(7\) |
We know that \(H^1(X_3,\mathbb Z)=0\), \(H^2(X_3,\mathbb Z)=H^1(G,\mathbb Z) =0\), \(H^3(X_3,\mathbb Z)=H^2(G,\mathbb Z) =\mathbb Z_3\), and that \(H^4(X_3,\mathbb Z)\) is a subgroup of \(H^3(G,\mathbb Z) = \mathbb Z_2\). It follows that the differential induces a surjection \(E_2^{0,4}=\mathbb Z_6 \twoheadrightarrow E_2^{5,0}=\mathbb Z_3\). Consequently \(H^4(X_3,\mathbb Z)=\mathbb Z_2\) and \(H^5(X_3,\mathbb Z)=0\) and \(H^6(X_3,\mathbb Z)=\mathbb Z_2\).
The \(E_2\) page for the fibration \(K(\pi_4 X,4) \hookrightarrow X_4 \twoheadrightarrow X_3\) contains the following terms.
| \(5\) | \(\pi_4 X\) | \(0\) | \(0\) | \(\) | \(\) | \(\) | \(\) |
| \(4\) | \(0\) | \(0\) | \(0\) | \(0\) | \(\) | \(\) | \(\) |
| \(3\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(\) |
| \(2\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | |
| \(1\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) |
| \(0\) | \(\mathbb Z\) | \(0\) | \(0\) | \(\mathbb Z_3\) | \(\mathbb Z_2\) | \(0\) | \(\mathbb Z_2\) |
| \(q/p\) | \(0\) | \(1\) | \(2\) | \(3\) | \(4\) | \(5\) | \(6\) |
We know that \(H^5(X_4,\mathbb Z)\) is a subgroup of \(H^4(G,\mathbb Z)=\mathbb Z_6\), and hence that there is a homomorphisms \(\pi_4X \rightarrow \mathbb Z_2\) whose kernel is a subgroup of \(\mathbb Z_6\). If follows that \(|\pi_4 X|\le 12\).
A 2-type is a CW-complex \(X\) whose homotopy groups are trivial in dimensions \(n=0 \) and \(n>2\). As explained in 6.9 the homotopy type of such a space can be captured algebraically by a cat\(^1\)-group \(G\). Let \(X\), \(Y\) be \(2\)-tytpes represented by cat\(^1\)-groups \(G\), \(H\). If \(X\) and \(Y\) are homotopy equivalent then there exists a sequence of morphisms of cat\(^1\)-groups
\[G \rightarrow K_1 \rightarrow K_2 \leftarrow K_3 \rightarrow \cdots \rightarrow K_n \leftarrow H\]
in which each morphism induces isomorphisms of homotopy groups. When such a sequence exists we say that \(G\) is quasi-isomorphic to \(H\). We have the following result.
Theorem. The \(2\)-types \(X\) and \(Y\) are homotopy equivalent if and only if the associated cat\(^1\)-groups \(G\) and \(H\) are quasi-isomorphic.
The following commands produce a list \(L\) of all of the \(62\) non-isomorphic cat\(^1\)-groups whose underlying group has order \(16\).
gap> L:=[];; gap> for G in AllSmallGroups(16) do > Append(L,CatOneGroupsByGroup(G)); > od; gap> Length(L); 62
The next commands use the first and second homotopy groups to prove that the list \(L\) contains at least \(37\) distinct quasi-isomorphism types.
gap> Invariants:=function(G) > local inv; > inv:=[]; > inv[1]:=IdGroup(HomotopyGroup(G,1)); > inv[2]:=IdGroup(HomotopyGroup(G,2)); > return inv; > end;; gap> C:=Classify(L,Invariants);; gap> Length(C);
The following additional commands use second and third integral homology in conjunction with the first two homotopy groups to prove that the list \(L\) contains at least \(49\) distinct quasi-isomorphism types.
gap> Invariants2:=function(G) > local inv; > inv:=[]; > inv[1]:=Homology(G,2); > inv[2]:=Homology(G,3); > return inv; > end;; gap> C:=RefineClassification(C,Invariants2);; gap> Length(C); 49
The following commands show that the above list \(L\) contains at most \(51\) distinct quasi-isomorphism types.
gap> Q:=List(L,QuasiIsomorph);; gap> M:=[];; gap> for q in Q do > bool:=true;; > for m in M do > if not IsomorphismCatOneGroups(m,q)=fail then bool:=false; break; fi; > od; > if bool then Add(M,q); fi; > od; gap> Length(M); 51
Let us define the order of a cat\(^1\)-group to be the order of its underlying group. The function IdQuasiCatOneGroup(C) inputs a cat\(^1\)-group \(C\) of "low order" and returns an integer pair \([n,k]\) that uniquely idenifies the quasi-isomorphism type of \(C\). The integer \(n\) is the order of a smallest cat\(^1\)-group quasi-isomorphic to \(C\). The integer \(k\) identifies a particular cat\(^1\)-group of order \(n\).
The following commands use this function to show that there are precisely \(49\) distinct quasi-isomorphism types of cat\(^1\)-groups of order \(16\).
gap> L:=[];; gap> for G in AllSmallGroups(16) do > Append(L,CatOneGroupsByGroup(G)); > od; gap> M:=List(L,IdQuasiCatOneGroup); [ [ 16, 1 ], [ 16, 2 ], [ 16, 3 ], [ 16, 4 ], [ 16, 5 ], [ 4, 4 ], [ 1, 1 ], [ 16, 6 ], [ 16, 7 ], [ 16, 8 ], [ 16, 9 ], [ 16, 10 ], [ 16, 11 ], [ 16, 9 ], [ 16, 12 ], [ 16, 13 ], [ 16, 14 ], [ 16, 15 ], [ 4, 1 ], [ 4, 2 ], [ 16, 16 ], [ 16, 17 ], [ 16, 18 ], [ 16, 19 ], [ 16, 20 ], [ 16, 21 ], [ 16, 22 ], [ 16, 23 ], [ 16, 24 ], [ 16, 25 ], [ 16, 26 ], [ 16, 27 ], [ 16, 28 ], [ 4, 3 ], [ 4, 1 ], [ 4, 4 ], [ 4, 4 ], [ 4, 2 ], [ 4, 5 ], [ 16, 29 ], [ 16, 30 ], [ 16, 31 ], [ 16, 32 ], [ 16, 33 ], [ 16, 34 ], [ 4, 3 ], [ 4, 4 ], [ 4, 4 ], [ 16, 35 ], [ 16, 36 ], [ 4, 3 ], [ 16, 37 ], [ 16, 38 ], [ 16, 39 ], [ 16, 40 ], [ 16, 41 ], [ 16, 42 ], [ 16, 43 ], [ 4, 3 ], [ 4, 4 ], [ 1, 1 ], [ 4, 5 ] ] gap> Length(SSortedList(M)); 49
The next example first identifies the order and the identity number of the cat\(^1\)-group \(C\) corresponding to the crossed module (see 12.1)
\[\iota\colon G \longrightarrow Aut(G), g \mapsto (x\mapsto gxg^{-1})\]
for the dihedral group \(G\) of order \(10\). It then realizes a smallest possible cat\(^1\)-group \(D\) of this quasi-isomorphism type.
gap> C:=AutomorphismGroupAsCatOneGroup(DihedralGroup(10)); Cat-1-group with underlying group Group( [ f1, f2, f3, f4, f5 ] ) . gap> Order(C); 200 gap> IdCatOneGroup(C); [ 200, 42, 4 ] gap> gap> IdQuasiCatOneGroup(C); [ 2, 1 ] gap> D:=SmallCatOneGroup(2,1); Cat-1-group with underlying group Group( [ f1 ] ) .
The following commands construct the crossed module \(\partial \colon G\otimes G \rightarrow G\) involving the nonabelian tensor square of the dihedral group $G$ of order \(10\), identify it as being number \(71\) in the list of crossed modules of order \(100\), create a quasi-isomorphic crossed module of order \(4\), and finally construct the corresponding cat\(^1\)-group of order \(100\).
gap> G:=DihedralGroup(10);; gap> T:=NonabelianTensorSquareAsCrossedModule(G); Crossed module with group homomorphism GroupHomomorphismByImages( Group( [ f3*f1*f3^-1*f1^-1, f3*f2*f3^-1*f2^-1 ] ), Group( [ f1, f2 ] ), [ f3*f1*f3^-1*f1^-1, f3*f2*f3^-1*f2^-1 ], [ <identity> of ..., f2^3 ] ) gap> IdCrossedModule(T); [ 100, 71 ] gap> Q:=QuasiIsomorph(T); Crossed module with group homomorphism Pcgs([ f2 ]) -> [ <identity> of ... ] gap> Order(Q); 4 gap> C:=CatOneGroupByCrossedModule(T); Cat-1-group with underlying group Group( [ F1, F2, F1 ] ) .
generated by GAPDoc2HTML