A.8. THE HOMOLOGY GROUPS OF SPHERES 481

whenever n≥ 1. Letting n = 1, Im(g∗) = 0 which is the kernel of ∆ and so ∆ is one to one.Note that H0 (U ∩V )≈ Z⊕Z because U ∩V consists of two path components.

0H1 (U)⊕

0H1 (V )

g∗→ H1(S1) ∆,1−1→

≈Z⊕ZH0 (U ∩V )

f∗→≈Z

H0 (U)⊕≈Z

H0 (V )

Then since the sequence is exact, ∆(H1(S1))

= ker( f∗) . So what is ker( f∗)? It must beisomorphic to m(1,−1) ,m ∈ Z which is isomorphic to Z.

More precisely, if U ∩V is L∪R where L,R are the left and right sides of U ∩V in theabove picture, the path components of U ∩V . H0 (U ∩V ) would be of the form (m [c] ,n [d])because there are two path components. Here [c] , [d] are in H0 (L) ,H0 (R) respectively c,dbeing cycles. Say

c = ∑φ

mφ φ ,d = ∑ψ

nψ ψ

where the φ have values in L and the ψ have values in R. Now all of the φ are homologousto each other in L because φ − φ̂ is indeed a boundary, so [c] can be reduced to m [φ ]and similarly [d] is of the form n [ψ] for φ ,ψ simplices. Thus we can assume c,d are0 simplices. Both c and d are supported in U and both are supported in V . To be inker( f∗) we would need n [d] +m [c] = 0 in H0 (U) and also in H0 (V ). This would meanthat [nd +mc] = 0 in H0 (U) so nd+mc would be a boundary in U . Of course this happensexactly when n =−m so that you can pair the two to obtain their difference as a boundary,and so ker( f∗) is of the form m([c]− [d]) , with c,d being 0 simplices in U ∩V , c in L and din R. Note that c−d is indeed a boundary in U and also in V but this is not a boundaryin U ∩V . Thus [c]− [d] is nonzero in H0 (U ∩V ) and

ker( f∗) = {m([c]− [d]) : m ∈ Z} ≈ H1(S1)≈ Z.

If [c1] generates H1(S1), this means that ∆([c1]) = [c]− [d] where c,d are two 0 cycles.

This seems to be a pretty useful observation.

Lemma A.8.2 Let U,V be the open sets given above. Then for any 0 cycle c in L and d inR, it follows that H1

(S1)≈ Z and a generator for H1

(S1)

is ∆−1 ([c]− [d]), this by LemmaA.0.2.

So what about Hn(S1)

for n > 1? Consider n = 2.

0H2 (U)⊕

0H2 (V )

g∗→ H2(S1) ∆,1−1→ H1 (U ∩V )

f∗→0⊕0

H1 (U)⊕H1 (V )

As just noted, H2 (U)⊕H2 (V )= (0,0) so ∆ is one to one since ker∆= 0. However, ker f∗=Im(∆) =H1 (U ∩V ) because f∗ maps to 0⊕0= 0. Now H1 (U ∩V ) = 0. It is the direct sumof homology groups of the two path components of U ∩V each of which is 0. Therefore,∆(H2(S1))

and consequently H2(S1)

is 0. Similarly the other Hn(S1)= 0 for n≥ 2. Just

replace 2 with n and repeat.Now it is time to find the homology groups of spheres in any dimension. The case of

S1 was just done. In particular, H1(S1)

is of the form m([c]− [d]) for c,d 0 simplices.Therefore, it is assumed that n > 1 in what follows.

Rn

Snt

b

xy

U

V0y

xx→ y

x

A.8. THE HOMOLOGY GROUPS OF SPHERES 481whenever n > 1. Letting n = 1, Im(g..) =0 which is the kernel of A and so A is one to one.Note that Hp (UNV) + Z@®Z because UNV consists of two path components.0) 0 ee |! Al—1 SZeZ fe 2Z XZ,H\U)@A(V) > Ai (S') “> Ao(UNV) 4 Ho(U) PCV)Then since the sequence is exact, A (Hj (S')) = ker(f,). So what is ker(f,)? It must beisomorphic to m(1,—1),m € Z which is isomorphic to Z.More precisely, if UV is LUR where L,R are the left and right sides of UM V in theabove picture, the path components of UNV. Ho (UNV) would be of the form (m[c] ,n [d])because there are two path components. Here [c] , |d] are in Ho (L) , Ho (R) respectively c,dbeing cycles. Sayc= Ying 9,d = Ynyyo ywhere the @ have values in L and the y have values in R. Now all of the @ are homologousto each other in L because @ — @ is indeed a boundary, so [c] can be reduced to m[@]and similarly [d] is of the form n[w] for @, y simplices. Thus we can assume c,d are0 simplices. Both c and d are supported in U and both are supported in V. To be inker (f,) we would need n|d] + m[c] = 0 in Ho (U) and also in Hp (V). This would meanthat [nd + mc] = 0 in Ho (U) so nd +mc would be a boundary in U. Of course this happensexactly when n = —m so that you can pair the two to obtain their difference as a boundary,and so ker (f,.) is of the form m ([c] — [d]) , with c,d being 0 simplices in UNV, c in Landdin R. Note that c —d is indeed a boundary in U and also in V but this is not a boundaryin UNV. Thus |c] — [d] is nonzero in Hp (UNV) andker(f,) = {m([c] —[d]): me Z} ~My (S') &Z.If [ci] generates Hj (S') , this means that A({c1]) = [c] — [d] where c,d are two 0 cycles.This seems to be a pretty useful observation.Lemma A.8.2 Let U,V be the open sets given above. Then for any 0 cycle c in L and d inR, it follows that H, (S') © Z and a generator for H, (S') is A~' ([c] — [d]), this by LemmaA.O.2.So what about H,, (Ss!) for n > 1? Consider n = 2.0 0 e Al-1 fe 060H)(U)@M2(V) + Mm(S') “> A(UNV) 3S AU) (V)As just noted, Hy (U) ®H2 (V) = (0,0) so A is one to one since ker A = 0. However, ker f,, =Im (A) = Hi (UNV) because f, maps to 0@0 = 0. Now A; (UNV) =0. It is the direct sumof homology groups of the two path components of U MV each of which is 0. Therefore,A (Hz (S')) and consequently H2 (S') is 0. Similarly the other H, (S') =0 for n > 2. Justreplace 2 with n and repeat.Now it is time to find the homology groups of spheres in any dimension. The case ofS' was just done. In particular, H; (S!) is of the form m({c] —[d]) for c,d 0 simplices.Therefore, it is assumed that n > 1 in what follows.SAKLyIR”y xbL’U