Let Y denote a finite regular CW-complex. Let widetilde Y denote its universal covering space. The covering space inherits a regular CW-structure which can be computed and stored using the datatype of a π_1Y-equivariant CW-complex. The cellular chain complex C_∗widetilde Y of widetilde Y can be computed and stored as an equivariant chain complex. Given an admissible discrete vector field on Y, we can endow Y with a smaller non-regular CW-structre whose cells correspond to the critical cells in the vector field. This smaller CW-structure leads to a more efficient chain complex C_∗ widetilde Y involving one free generator for each critical cell in the vector field.
The following commands construct a 6-dimensional regular CW-complex Y≃ S^1 × S^1× S^1 homotopy equivalent to a product of three circles.
gap> A:=[[1,1,1],[1,0,1],[1,1,1]];; gap> S:=PureCubicalComplex(A);; gap> T:=DirectProduct(S,S,S);; gap> Y:=RegularCWComplex(T);; Regular CW-complex of dimension 6 gap> Size(Y); 110592
The CW-somplex Y has 110592 cells. The next commands construct a free π_1Y-equivariant chain complex C_∗widetilde Y homotopy equivalent to the chain complex of the universal cover of Y. The chain complex C_∗widetilde Y has just 8 free generators.
gap> Y:=ContractedComplex(Y);; gap> CU:=ChainComplexOfUniversalCover(Y);; gap> List([0..Dimension(Y)],n->CU!.dimension(n)); [ 1, 3, 3, 1 ]
The next commands construct a subgroup H < π_1Y of index 50 and the chain complex C_∗widetilde Y⊗_ ZH Z which is homotopy equivalent to the cellular chain complex C_∗widetilde Y_H of the 50-fold cover widetilde Y_H of Y corresponding to H.
gap> L:=LowIndexSubgroupsFpGroup(CU!.group,50);; gap> H:=L[Length(L)-1];; gap> Index(CU!.group,H); 50 gap> D:=TensorWithIntegersOverSubgroup(CU,H); Chain complex of length 3 in characteristic 0 . gap> List([0..3],D!.dimension); [ 50, 150, 150, 50 ]
General theory implies that the 50-fold covering space widetilde Y_H should again be homotopy equivalent to a product of three circles. In keeping with this, the following commands verify that widetilde Y_H has the same integral homology as S^1× S^1× S^1.
gap> Homology(D,0); [ 0 ] gap> Homology(D,1); [ 0, 0, 0 ] gap> Homology(D,2); [ 0, 0, 0 ] gap> Homology(D,3); [ 0 ]
We'll contruct two spaces Y,W with isomorphic fundamental groups and isomorphic intergal homology, and use the integral homology of finite covering spaces to establsh that the two spaces have distinct homotopy types.
By spinning a link K ⊂ R^3 about a plane P⊂ R^3 with P∩ K=∅, we obtain a collection Sp(K)⊂ R^4 of knotted tori. The following commands produce the two tori obtained by spinning the Hopf link K and show that the space Y= R^4∖ Sp(K) = Sp( R^3∖ K) is connected with fundamental group π_1Y = Z× Z and homology groups H_0(Y)= Z, H_1(Y)= Z^2, H_2(Y)= Z^4, H_3(Y, Z)= Z^2. The space Y is only constructed up to homotopy, and for this reason is 3-dimensional.
gap> Hopf:=PureCubicalLink("Hopf"); Pure cubical link. gap> Y:=SpunAboutInitialHyperplane(PureComplexComplement(Hopf)); Regular CW-complex of dimension 3 gap> Homology(Y,0); [ 0 ] gap> Homology(Y,1); [ 0, 0 ] gap> Homology(Y,2); [ 0, 0, 0, 0 ] gap> Homology(Y,3); [ 0, 0 ] gap> Homology(Y,4); [ ] gap> GY:=FundamentalGroup(Y);; gap> GeneratorsOfGroup(GY); [ f2, f3 ] gap> RelatorsOfFpGroup(GY); [ f3^-1*f2^-1*f3*f2 ]
An alternative embedding of two tori L⊂ R^4 can be obtained by applying the 'tube map' of Shin Satoh to a welded Hopf link [Sat00]. The following commands construct the complement W= R^4∖ L of this alternative embedding and show that W has the same fundamental group and integral homology as Y above.
gap> L:=HopfSatohSurface(); Pure cubical complex of dimension 4. gap> W:=ContractedComplex(RegularCWComplex(PureComplexComplement(L))); Regular CW-complex of dimension 3 gap> Homology(W,0); [ 0 ] gap> Homology(W,1); [ 0, 0 ] gap> Homology(W,2); [ 0, 0, 0, 0 ] gap> Homology(W,3); [ 0, 0 ] gap> Homology(W,4); [ ] gap> GW:=FundamentalGroup(W);; gap> GeneratorsOfGroup(GW); [ f1, f2 ] gap> RelatorsOfFpGroup(GW); [ f1^-1*f2^-1*f1*f2 ]
Despite having the same fundamental group and integral homology groups, the above two spaces Y and W were shown by Kauffman and Martins [KFM08] to be not homotopy equivalent. Their technique involves the fundamental crossed module derived from the first three dimensions of the universal cover of a space, and counts the representations of this fundamental crossed module into a given finite crossed module. This homotopy inequivalence is recovered by the following commands which involves the 5-fold covers of the spaces.
gap> CY:=ChainComplexOfUniversalCover(Y); Equivariant chain complex of dimension 3 gap> LY:=LowIndexSubgroups(CY!.group,5);; gap> invY:=List(LY,g->Homology(TensorWithIntegersOverSubgroup(CY,g),2));; gap> CW:=ChainComplexOfUniversalCover(W); Equivariant chain complex of dimension 3 gap> LW:=LowIndexSubgroups(CW!.group,5);; gap> invW:=List(LW,g->Homology(TensorWithIntegersOverSubgroup(CW,g),2));; gap> SSortedList(invY)=SSortedList(invW); false
The π_1Y-equivariant cellular chain complex C_∗widetilde Y of the universal cover widetilde Y of a regular CW-complex Y can be used to compute the homology H_n(Y,A) and cohomology H^n(Y,A) of Y with local coefficients in a Zπ_1Y-module A. To illustrate this we consister the space Y arising as the complement of the trefoil knot, with fundamental group π_1Y = ⟨ x,y : xyx=yxy ⟩. We take A= Z to be the integers with non-trivial π_1Y-action given by x.1=-1, y.1=-1. We then compute
beginarraylcl H_0(Y,A) &= & Z_2 , H_1(Y,A) &= & Z_3 , H_2(Y,A) &= & Z .endarray
gap> K:=PureCubicalKnot(3,1);; gap> Y:=PureComplexComplement(K);; gap> Y:=ContractedComplex(Y);; gap> Y:=RegularCWComplex(Y);; gap> Y:=SimplifiedComplex(Y);; gap> C:=ChainComplexOfUniversalCover(Y);; gap> G:=C!.group;; gap> GeneratorsOfGroup(G); [ f1, f2 ] gap> RelatorsOfFpGroup(G); [ f2^-1*f1^-1*f2^-1*f1*f2*f1, f1^-1*f2^-1*f1^-1*f2*f1*f2 ] gap> hom:=GroupHomomorphismByImages(G,Group([[-1]]),[G.1,G.2],[[[-1]],[[-1]]]);; gap> A:=function(x); return Determinant(Image(hom,x)); end;; gap> D:=TensorWithTwistedIntegers(C,A); #Here the function A represents gap> #the integers with twisted action of G. Chain complex of length 3 in characteristic 0 . gap> Homology(D,0); [ 2 ] gap> Homology(D,1); [ 3 ] gap> Homology(D,2); [ 0 ]
The granny knot is the sum of the trefoil knot and its mirror image. The reef knot is the sum of two identical copies of the trefoil knot. The following commands show that the degree 1 homology homomorphisms
H_1(p^-1(B), Z) → H_1(widetilde X_H, Z)
distinguish between the homeomorphism types of the complements X⊂ R^3 of the granny knot and the reef knot, where B⊂ X is the knot boundary, and where p: widetilde X_H → X is the covering map corresponding to the finite index subgroup H < π_1X. More precisely, p^-1(B) is in general a union of path components
p^-1(B) = B_1 ∪ B_2 ∪ ⋯ ∪ B_t .
The function FirstHomologyCoveringCokernels(f,c) inputs an integer c and the inclusion f: B↪ X of a knot boundary B into the knot complement X. The function returns the ordered list of the lists of abelian invariants of cokernels
coker( H_1(p^-1(B_i), Z) → H_1(widetilde X_H, Z) )
arising from subgroups H < π_1X of index c. To distinguish between the granny and reef knots we use index c=6.
gap> K:=PureCubicalKnot(3,1);; gap> L:=ReflectedCubicalKnot(K);; gap> granny:=KnotSum(K,L);; gap> reef:=KnotSum(K,K);; gap> fg:=KnotComplementWithBoundary(ArcPresentation(granny));; gap> fr:=KnotComplementWithBoundary(ArcPresentation(reef));; gap> a:=FirstHomologyCoveringCokernels(fg,6);; gap> b:=FirstHomologyCoveringCokernels(fr,6);; gap> a=b; false
If p:widetilde Y → Y is the universal covering map, then the fundamental group of widetilde Y is trivial and the Hurewicz homomorphism π_2widetilde Y→ H_2(widetilde Y, Z) from the second homotopy group of widetilde Y to the second integral homology of widetilde Y is an isomorphism. Furthermore, the map p induces an isomorphism π_2widetilde Y → π_2Y. Thus H_2(widetilde Y, Z) is isomorphic to the second homotopy group π_2Y.
If the fundamental group of Y happens to be finite, then in principle we can calculate H_2(widetilde Y, Z) ≅ π_2Y. We illustrate this computation for Y equal to the real projective plane. The above computation shows that Y has second homotopy group π_2Y ≅ Z.
gap> K:=[ [1,2,3], [1,3,4], [1,2,6], [1,5,6], [1,4,5], > [2,3,5], [2,4,5], [2,4,6], [3,4,6], [3,5,6]];; gap> K:=MaximalSimplicesToSimplicialComplex(K); Simplicial complex of dimension 2. gap> Y:=RegularCWComplex(K); Regular CW-complex of dimension 2 gap> # Y is a regular CW-complex corresponding to the projective plane. gap> U:=UniversalCover(Y); Equivariant CW-complex of dimension 2 gap> G:=U!.group;; gap> # G is the fundamental group of Y, which by the next command gap> # is finite of order 2. gap> Order(G); 2 gap> U:=EquivariantCWComplexToRegularCWComplex(U,Group(One(G))); Regular CW-complex of dimension 2 gap> #U is the universal cover of Y gap> Homology(U,0); [ 0 ] gap> Homology(U,1); [ ] gap> Homology(U,2); [ 0 ]
For any path connected space Y with universal cover widetilde Y there is an exact sequence
→ π_4widetilde Y → H_4(widetilde Y, Z) → H_4( K(π_2widetilde Y,2), Z ) → π_3widetilde Y → H_3(widetilde Y, Z) → 0
due to J.H.C.Whitehead. Here K(π_2(widetilde Y),2) is an Eilenberg-MacLane space with second homotopy group equal to π_2widetilde Y.
Continuing with the above example where Y is the real projective plane, we see that H_4(widetilde Y, Z) = H_3(widetilde Y, Z) = 0 since widetilde Y is a 2-dimensional CW-space. The exact sequence implies π_3widetilde Y ≅ H_4(K(π_2widetilde Y,2), Z ). Furthermore, π_3widetilde Y = π_3 Y. The following commands establish that π_3Y ≅ Z.
gap> A:=AbelianPcpGroup([0]); Pcp-group with orders [ 0 ] gap> K:=EilenbergMacLaneSimplicialGroup(A,2,5);; gap> C:=ChainComplexOfSimplicialGroup(K); Chain complex of length 5 in characteristic 0 . gap> Homology(C,4); [ 0 ]
The following commands construct a 4-dimensional simplicial complex Y with 9 vertices and 36 4-dimensional simplices, and establish that
π_1Y=0 , π_2Y= Z , H_3(Y, Z)=0, H_4(Y, Z)= Z.
gap> smp:=[ [ 1, 2, 4, 5, 6 ], [ 1, 2, 4, 5, 9 ], [ 1, 2, 5, 6, 8 ], > [ 1, 2, 6, 4, 7 ], [ 2, 3, 4, 5, 8 ], [ 2, 3, 5, 6, 4 ], > [ 2, 3, 5, 6, 7 ], [ 2, 3, 6, 4, 9 ], [ 3, 1, 4, 5, 7 ], > [ 3, 1, 5, 6, 9 ], [ 3, 1, 6, 4, 5 ], [ 3, 1, 6, 4, 8 ], > [ 4, 5, 7, 8, 3 ], [ 4, 5, 7, 8, 9 ], [ 4, 5, 8, 9, 2 ], > [ 4, 5, 9, 7, 1 ], [ 5, 6, 7, 8, 2 ], [ 5, 6, 8, 9, 1 ], > [ 5, 6, 8, 9, 7 ], [ 5, 6, 9, 7, 3 ], [ 6, 4, 7, 8, 1 ], > [ 6, 4, 8, 9, 3 ], [ 6, 4, 9, 7, 2 ], [ 6, 4, 9, 7, 8 ], > [ 7, 8, 1, 2, 3 ], [ 7, 8, 1, 2, 6 ], [ 7, 8, 2, 3, 5 ], > [ 7, 8, 3, 1, 4 ], [ 8, 9, 1, 2, 5 ], [ 8, 9, 2, 3, 1 ], > [ 8, 9, 2, 3, 4 ], [ 8, 9, 3, 1, 6 ], [ 9, 7, 1, 2, 4 ], > [ 9, 7, 2, 3, 6 ], [ 9, 7, 3, 1, 2 ], [ 9, 7, 3, 1, 5 ] ];; gap> K:=MaximalSimplicesToSimplicialComplex(smp); Simplicial complex of dimension 4. gap> Y:=RegularCWComplex(Y); Regular CW-complex of dimension 4 gap> Order(FundamentalGroup(Y)); 1 gap> Homology(Y,2); [ 0 ] gap> Homology(Y,3); [ ] gap> Homology(Y,4); [ 0 ]
Previous commands have established H_4(K( Z,2), Z)= Z. So Whitehead's sequence reduces to an exact sequence
Z → Z → π_3Y → 0
in which the first map is H_4(Y, Z)= Z → H_4(K(π_2Y,2), Z )= Z. Hence π_3Y is cyclic.
HAP is currently unable to compute the order of π_3Y directly from Whitehead's sequence. Instead, we can use the Hopf invariant. For any map ϕ: S^3 → S^2 we consider the space C(ϕ) = S^2 ∪_ϕ e^4 obtained by attaching a 4-cell e^4 to S^2 via the attaching map ϕ. The cohomology groups H^2(C(ϕ), Z)= Z, H^4(C(ϕ), Z)= Z are generated by elements α, β say, and the cup product has the form
- ∪ -: H^2(C(ϕ), Z)× H^2(C(ϕ), Z) → H^4(C(ϕ), Z), (α,α) ↦ h_ϕ β
for some integer h_ϕ. The integer h_ϕ is the Hopf invariant. The function h: π_3(S^3)→ Z is a homomorphism and there is an isomorphism
π_3(S^2∪ e^4) ≅ Z/⟨ h_ϕ⟩.
The following commands begin by simplifying the cell structure on the above CW-complex Y≅ K to obtain a homeomorphic CW-complex W with fewer cells. They then create a space S by removing one 4-cell from W. The space S is seen to be homotopy equivalent to a CW-complex e^2∪ e^0 with a single 0-cell and single 2-cell. Hence S≃ S^2 is homotopy equivalent to the 2-sphere. Consequently Y ≃ C(ϕ ) = S^2∪_ϕ e^4 for some map ϕ: S^3 → S^2.
gap> W:=SimplifiedComplex(Y); Regular CW-complex of dimension 4 gap> S:=RegularCWComplexWithRemovedCell(W,4,6); Regular CW-complex of dimension 4 gap> CriticalCells(S); [ [ 2, 6 ], [ 0, 5 ] ]
The next commands show that the map ϕ in the construction Y ≃ C(ϕ) has Hopf invariant -1. Hence h: π_3(S^3)→ Z is an isomorphism. Therefore π_3Y=0.
gap> IntersectionForm(K); [ [ -1 ] ]
[The simplicial complex K in this second example is due to W. Kuehnel and T. F. Banchoff and is homeomorphic to the complex projective plane. ]
The following commands compute the second integral homology
H_2(π_1W, Z) = Z
of the fundamental group π_1W of the complement W of the Hopf-Satoh surface.
gap> L:=HopfSatohSurface(); Pure cubical complex of dimension 4. gap> W:=ContractedComplex(RegularCWComplex(PureComplexComplement(L))); Regular CW-complex of dimension 3 gap> GW:=FundamentalGroup(W);; gap> IsAspherical(GW); Presentation is aspherical. true gap> R:=ResolutionAsphericalPresentation(GW);; gap> Homology(TensorWithIntegers(R),2); [ 0 ]
From Hopf's exact sequence
π_2W stackrelh⟶ H_2(W, Z) ↠ H_2(π_1W, Z) → 0
and the computation H_2(W, Z)= Z^4 we see that the image of the Hurewicz homomorphism is im(h)= Z^3 . The image of h is referred to as the subgroup of spherical homology classes and often denoted by Σ^2W.
The following command computes the presentation of π_1W corresponding to the 2-skeleton W^2 and establishes that W^2 = S^2∨ S^2 ∨ S^2 ∨ (S^1× S^1) is a wedge of three spheres and a torus.
gap> F:=FundamentalGroupOfRegularCWComplex(W,"no simplification"); < fp group on the generators [ f1, f2 ]> gap> RelatorsOfFpGroup(F); [ < identity ...>, f1^-1*f2^-1*f1*f2, < identity ...>, <identity ...> ]
The next command shows that the 3-dimensional space W has two 3-cells each of which is attached to the base-point of W with trivial boundary (up to homotopy in W^2). Hence W = S^3∨ S^3∨ S^2 ∨ S^2 ∨ S^2 ∨ (S^1× S^1).
gap> CriticalCells(W); [ [ 3, 1 ], [ 3, 3148 ], [ 2, 6746 ], [ 2, 20510 ], [ 2, 33060 ], [ 2, 50919 ], [ 1, 29368 ], [ 1, 50822 ], [ 0, 21131 ] ] gap> CriticalBoundaryCells(W,3,1); [ ] gap> CriticalBoundaryCells(W,3,3148); [ -50919, 50919 ]
Therefore π_1W is the free abelian group on two generators, and π_2W is the free Zπ_1W-module on three free generators.
generated by GAPDoc2HTML