Algebraic Topology

DRAFT 2013 - U.T.

Contents
0 Introduction 2

1 Chain complexes 2

2 Simplicial homology 4

3 Singular homology 6

4 Chain homotopy and homotopy invariance 8

4.1 Chain homotopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.2 Homotopy invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

5 The Snake Lemma and relative homology 10

6 The Five Lemma and excision 12

6.1 A tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
6.2 Excision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
6.3 Locality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
6.4 Mayer-Vietoris sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

7 The degree of a map of spheres 15

8 Cellular homology 17

9 Cohomology and the Universal Coefficient Theorem 20

9.1 Cohomology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
9.2 Universal Coefficient Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

1
0 Introduction

We study topological spaces by associating algebraic objects which are invariant under homotopy
(deformation).
Sets: number of path components π0
Numbers: Euler characteristic χ
winding number w(γ)
Groups: fundamental group π1
Graded abelian groups: Homology H∗
Graded rings: Cohomology H∗
This course will introduce and study homology and cohomology.

Why?
• Nonabelian groups are difficult to work with
• π1 only tells us something about the 2-skeleton of a space
• higher dimensional analogues, πn = {homotopy classes of based maps from S n to the space
of interest}, are very hard to compute. Even πn (S k ) are not all know for n > k.
Homology, H∗ , is a refinement of the Euler characteristic and still relatively easy to compute.

References
Everything contained in this course is covered in
• chapters 2 and 3 of ‘Algebraic Topology’ by Allen Hatcher, available online at
http://www.math.cornell.edu/∼hatcher/

1 Chain complexes

Definition 1.1. Let C0 , C1 , . . . be abelian groups (or vector spaces) and let ∂n ∶ Cn → Cn−1 be
homomorphisms (or linear maps). Assume that ∂n ○ ∂n+1 = 0 for all n. Then
∂3 ∂2 ∂1 ∂0 =0
(C● , ∂● ) ..= ⋯ Ð
Ð→ C2 Ð
Ð→ C1 Ð
Ð→ C0 ÐÐÐ→ 0

is a chain complex. Furthermore define

• n-chains: Cn
• n-cycles: Zn ..= ker(∂n )
• n-boundaries: Bn ..= im(∂n+1 )
Sometimes we refer to a chain complex as C● for simplicity.
Note 1.2. As ∂n ○ ∂n+1 = 0, Bn ⊆ Zn .

2
Definition 1.3. The n-th homology of a chain complex (C● , ∂● ) is

Hn (C● , ∂● ) ..= Zn /Bn = ker(∂n )/ im(∂n+1 ).

C● is called exact at Cn if Hn = 0, i.e. if ker(∂n ) = im(∂n+1 ). C● is called exact if it is exact at

Cn for all n ≥ 0.
Example 1.4. Let
α β
0 Ð→ A Ð
Ð→ B Ð
Ð→ C Ð→ 0
be a short exact sequence of abelian groups, then

ker(α) = im(0) = {0} Ô⇒ α is injective

ker(β) = im(α) = A
ker(0) = C = im(β) Ô⇒ β is surjective

Thus C is isomorphic to the quotient B/A. Note that the group B is not determined by A and
C. For example, when A = Z and C = Z/2Z then B could be Z or Z ⊕ Z/2Z. On the otherhand,
if in a short exact sequence of vector spaces where A, B and C are finite dimensional and all
maps are linear, B is determined by A and C upto isomorphism: dim(B) = dim(A) + dim(C).
̃● , ∂̃● ) is a collection of homomorphisms φn ∶ Cn → C
Definition 1.5. A chain map φ ∶ (C● , ∂● ) → (C ̃n
such that the following diagram commutes.

∂2 ∂1
⋯ / C2 / C1 / C 0 ∂0 / 0

φ2 φ1 φ0
 ∂̃2
 ∂̃1
 ∂̃
⋯ /C
̃ /C
̃ /C
̃0 0 / 0
2 1

That is, such that φn−1 ○ ∂n = ∂̃n ○ φn for all n ≥ 0.

Proposition 1.6. φ ∶ C● → C ̃● induces a homomorphism φ∗ ∶ Hn (C● ) → Hn (C
̃● ) for all n ≥ 0
by φ∗ ([x]) = [φn (x)] where [x] = x + Bn ∈ Zn /Bn = Hn (C● ).

Proof. We show that the map is well-defined:

Let [x] = [x′ ], then x = x′ + b for some b ∈ Bn . So φn (x) = φn (x′ + b) = φn (x′ ) + φn (b). We need
̃n , then [φn (x)] = [φn (x′ )].
φn (b) ∈ B

∂n+1
/ Cn ∂n / Cn−1 Since b ∈ Bn , there is a b′ ∈ Cn+1 such that ∂n+1 (b′ ) = b.
Cn+1
 /b Then, by commutativity,
b′ _
φn (b) = φn (∂n+1 (b′ ))
φn+1 φn φn−1

 ∂̃n+1
 ∂̃n

̃n+1
C /C
̃n /C
̃n−1 = ∂̃n+1 ○ φn+1 (b′ )

φn (b) ̃m = im(∂̃n+1 ).
∈B

̃n : as x is a cycle, ∂̃n ○ φn (x) = φn−1 ○ ∂n (x) = 0.

Finally, one checks that φn (x) ∈ Z

We will study three different notions of homology groups Hn (X) for a topological space X,
which will agree when defined.

3
 Simplicial Homology easy to compute and good geometric intuition
Singular Homology good for theoretical work
Cellular Homology better for computations and applications
Through out the course we will use freely the following result from algebra.
Theorem 1.7. Any finitely generated abelian group is isomorphic to

Zr ⊕ (Z/pn1 1 Z) ⊕ (Z/pn2 2 Z) ⊕ ⋯ ⊕ (Z/pnk k Z)

for some r ∈ Z≥0 , k ∈ Z≥0 , each pi is prime (possibly with pi = pj for i ≠ j) and each ni ∈ Z>0 .
Moreover, r, k, pi and ni are unique up to reordering.
Example 1.8. (1) Z/6Z ≅ Z/2Z ⊕ Z/3Z
(2) Z/9Z ≇ Z/3Z ⊕ Z/3Z

2 Simplicial homology

Definition 2.1. The standard n-simplex ∆n is

{(t0 , . . . , tn ) ∈ Rn+1 ∶ ∑ ti = 1 and ti ≥ 0 ∀i}.

Example 2.2.

∆0 = e2
e1
e1
∆2 =
∆1 =

e0 e0

Definition 2.3. Let {v0 , . . . , vn } ⊆ Rn+k be such that v1 −v0 , . . . , vk −v0 are linearly independent.
The n-simplex spanned by v0 , . . . , vn is

{∑ ti vi ∶ ti ≥ 0 ∀i and ∑ ti = 1}.

Its vertices are {v0 , . . . , vn }. Let [v0 , . . . , vn ] denote the n-simplex spanned by v0 , . . . , vn together
with the given ordering of its vertices. The ordering induces an orientation of its edges [vi , vj ]
from vi to vj .
For any nonempty subset of the vertices, the simplex that they span is a face of [v0 , . . . , vn ].
The vertices of this face inherit an ordering.
There is a canonical homeomorphism

∆n Ð→ [v0 , . . . , vn ]
(t0 , . . . , tn ) z→ t0 v0 + ⋯ + tn vn .

Definition 2.4. A ∆-complex is obtained as follows:

(i) Start with an indexing set In for each n ∈ Z≥0 .

4
 (ii) For each α ∈ In take a copy σαn of the standard n-simplex.
(iii) Form the disjoint union of all these simplices over all n ∈ Z≥0 : ∐n∈Z≥0 ∐α∈In σαn .
(iv) We require that for each (n−1)-dimensional face of each n-simplex σαn there is an associated
(n − 1)-simplex σβn−1 for some β ∈ In−1 .
(v) Form the quotient space by identifying each (n − 1)-dimensional face of each σαn with the
associated simplex σβn−1 using the canonical homeomorphism. These homeomorphisms
preserve the ordering of the vertices.
Example 2.5. The torus admits the structure of a ∆-complex with
one 0-simplex
three 1-simplices
two 2-simplices

Note 2.6. The following is not a well-defined ∆-complex because

the vertices of each 2-simplex are not totally ordered.

Remark 2.7. Any simplicial complex is a ∆-complex, but in a ∆-complex an n-simplex may not
be uniquely determined by its vertices as in Example 2.5.
Definition 2.8. Let X● be a ∆-complex. The n-th chain group Cn∆ (X● ) of X● is the free abelian
group generated by the set of n-simplices Xn of X● . An element of Cn∆ (X● ) is ∑α∈In cα σαn , where
each cα ∈ Z and only finitely many of the cα ’s are non-zero. The boundary homomorphism is

∂n ∶ Cn∆ (X● ) Ð→ Cn−1

∆
(X● )
[v0 , . . . , vn ] z→ ∑(−1)i [v0 , . . . , v̂i , . . . , vn ]
i

where [v0 , . . . , v̂i , . . . , vn ] ..= [v0 , . . . , vi−1 , vi+1 , . . . , vn ].

Example 2.9.

⎛ v2 ⎞ ⎛ × −1
⎛ v1 ⎞ ⎛ +v1 ⎞ ⎜ ⎟ ⎜ ×1 ⎞
∂1 ⎜ ⎟=⎜ ⎟ ∂2 ⎜ ⎟=⎜ ⎟
⎜ ⎟ ⎟
⎝ v0 ⎠ ⎝ −v0 ⎠ ⎝ v0 ⎠ ⎝ ×1 ⎠
v1

Key Lemma 2.10. ∂n−1 ○ ∂n = 0 ∀n

Proof.

∂n−1 ○ ∂n ([v0 , . . . , vn ]) = ∂n−1 (∑(−1)i [v0 , . . . , v̂i , . . . , vn ])

i
= ∑(−1)i (−1)j [v0 . . . , v̂j , . . . , v̂i , . . . , vn ]
j<i

+ ∑(−1)i (−1)j−1 [v0 . . . , v̂i , . . . , v̂j , . . . , vn ]

j>i

= 0.

5
Definition 2.11. For n ≥ 0 the n-th simplicial homology group Hn∆ (X● ) is the n-th homology
group of the chain complex
∂n+1 ∂n
⋯ Ð→ Cn+1
∆
(X● ) ÐÐ→ Cn∆ (X● ) Ð
Ð→ Cn−1
∆
(X● ) Ð→ ⋯

Example 2.12. X● = S 1 ∨ S 1

∂1
⋯ / C ∆ (X ) / C ∆ (X ) / C ∆ (X ) /0
2 ● 1 ● 0 ●
a b
v 0 Z⊕Z Z
generated
generated by v
by a and b

∂1 (a) = v − v and ∂1 (b) = v − v so

H0 (X● ) = ker(∂0 )/ im(∂1 ) = Z/{0} = Z

H1 (X● ) = ker(∂1 )/ im(∂1 ) = (Z ⊕ Z)/{0} = Z ⊕ Z
Hn (X● ) = 0 for n ≥ 2.

Example 2.13. X● = torus T

v a v
1 2 ⋯ / C ∆ (X ) / C ∆ (X ) ∂2
/ C ∆ (X ) ∂1
/ C ∆ (X) /0
f 2 3 ● 2 ● 1 ● 0

b c b
g 0 Z⊕Z Z⊕Z⊕Z Z
0 generated generated
0 1 generated by v
by f and g by a, b, c
v a v

∂1 (a) = ∂1 (b) = ∂1 (c) = v − v = 0 and ∂2 (f ) = ∂2 (g) = a + b − c so

H0 (X● ) = ker(∂0 )/ im(∂1 ) = Z/{0} = Z

H1 (X● ) = ker(∂1 )/ im(∂2 ) = (Z ⊕ Z ⊕ Z)/ im(∂2 ) ≅ Z ⊕ Z generated by a and b
H2 (X● ) = ker(∂2 )/ im(∂3 ) = ker(∂n )/{0} ≅ Z generated by f − g
Hn (X● ) = 0 for n ≥ 3.

Example 2.14. A simplicial complex with only 0-simplices (vertices) and 1-simplices (edges) is
a graph G. If G is connected and finite then its first homology is free of rank equal the number
of edges minus the number of vertices plus one.

3 Singular homology

Singular homology is a theoretical tool. It is not practical for computations. Our immediate
goals are to show the following:
• continuous maps induce a map on H∗ (this section)
• H∗ is invariant under homotopy (section 4)

6
Definition 3.1. Let X be a topological space. A continuous map σ ∶ ∆n = [e0 , . . . , en ] → X is
called a singular n-simplex. The singular n-chains are the elements of the free abelian group
generated by the n-simplices:

Cn (X) ..= {∑ aα σα ∶ aα ∈ Z with only finitely many ≠ 0, σα ∶ ∆n → X}

α

The boundary operator is defined on generators by

∂n ∶ Cn (X) Ð→ Cn−1 (X)

n
σ z→ ∂n σ ..= ∑(−1)i σ∣[e0 ,...,êi ,...,en ]
i=0

Lemma 3.2. ∂n−1 ○ ∂n = 0 ∀n

Proof. Heuristically, ∂n restricts a singular simplex σ to its boundary ∂n [e0 , . . . , en ]. The bound-
ary has no boundary. So ∂n−1 (∂n σ) restricts σ restricted to the empty set and ∂n−1 (∂n [e0 , . . . , en ]) = 0.
For a formal proof imitate the proof of Key Lemma in section 2.
Definition 3.3. Singular homology Hn (X) ..= Hn (C● (X)).
Remark 3.4. For a simplicial complex X● , the map that assigns to an n-simplex α its canon-
ical homeomorphism σα ∶ ∆n → α induces a chain map C●∆ (X● ) → C● (X) which induces an
isomorphism of homology groups as we will see later.
Example 3.5. X = pt. For each n there is only one map cn ∶ ∆n → X.
n
∂n (cn ) = ∑(−1)i cn ∣[e0 ,...,êi ,...,en ]
i=0
⎧
⎪
⎪0 if n is odd
= ⎨ n−1
⎪
⎪
⎩c if n is even

1 0 1 0 0
Ô⇒ C● = ⋯ Ð→ZÐ →ZÐ →ZÐ→ZÐ
→0
⎧
⎪
⎪Z n = 0
Ô⇒ Hn (pt) = ⎨
⎪
⎩0 n ≠ 0
⎪

Naturality. Let f ∶ X → Y be a continuous maps. Define

f♯ ∶ Cn (X) Ð→ Cn (Y )
σ z→ f ○ σ ∶ ∆n → X → Y.

Note that ∂n ○ f♯ (σ) = f♯ ○ ∂n (σ). Hence f♯ induces a map of chain complexes and hence a
homomorphism of homology groups:

f∗ ∶ Hn (X) Ð→ Hn (Y ) for all n.

Furthermore, if g ∶ Y → Z is another continuous map then

(g ○ f )♯ = g♯ ○ f♯ and (g ○ f )∗ = g∗ ○ f∗
(idX )♯ = idC● (X) and (idX )∗ = idHn for all n

where idX ∶ X → X is the identity map.

7
Aside 3.6. ( Optional) This says that homology is a functor from the category of topological
spaces and continuous maps to the category of graded abelian groups and homomorphisms that
preserve the grading:
X z→ H∗ (X) = ⊕ Hn (X).
n≥0

4 Chain homotopy and homotopy invariance

We will first establish a criterion (chain homotopy) for when two chain maps induce the same
homomorphism in homology. Then we will show that a homotopy between two continuous maps
induces such a chain homotopy and hence the two maps induce the same homomorphism in
homology.

4.1 Chain homotopy

̃● , ∂̃● ) are chain homotopic

Definition 4.1. Two maps of chain complexes φ● , ψ● ∶ (C● , ∂● ) → (C
if there exist homomorphisms hn ∶ Cn → C̃n+1 such that
∂̃n+1 ○ hn + hn−1 ○ ∂n = φn − ψn .
h● is called a chain homotopy.
Proposition 4.2. If φ● and ψ● are chain homotopic then
̃● ).
φ∗ = ψ∗ ∶ Hn (C● ) Ð→ Hn (C

Proof. Let z ∈ Zn = ker ∂n . Then

φn (z) − φn (z) = (∂̃n+1 ○ hn + hn−1 ○ ∂n )(z) as z ∈ Zn
= (∂̃n+1 ○ hn )(z) ∈ B
̃n = im ∂̃n+1 .
Hence, φ∗ ([z]) = [φn (z)] = [ψn (z)] = ψ∗ ([z]).
/ Cn ∂n / Cn−1
Cn+1
ss ss
hn sss ss
s φ−ψ sss
sss ss
ys  yss hn−1
̃n+1
C /C
̃ /C̃
n n−1
∂̃n+1

Main Example 4.3. Let i0 , i1 ∶ X → X × [0, 1] be the inclusions x ↦ (x, 0) and x ↦ (x, 1).
Then there exists a chain homotopy h● between
(i0 )♯ and (i1 )♯ ∶ C● (X) Ð→ C● (X × [0, 1]).
Construction 4.4. Let ∆n × {0} = [v0 , . . . , vn ], ∆n × {1} = [w0 , . . . , wn ] and divide ∆n × [0, 1]
into n+1 (n+1)-simplices of the form si ..= [v0 , . . . , vi , wi , . . . , wn ] where i ranges from 0 to n. Let
Γn ..= ∑n+1
i=0 (−1) si ∈ Cn+1 (∆ × [0, 1]), then
i n

∂Γn = ”∂∆n × [0, 1] ∐ ∆n × ∂[0, 1]”

n+1 i
= ∑ ∑ (−1)i (−1)j [v0 , . . . , v̂j , . . . , vi , wi , . . . , wn ]
i=0 j=0
n+1 n+1
+ ∑ ∑ (−1)i (−1)j+1 [v0 , . . . , vi , wi , . . . , ŵj , . . . , wn ]
i=0 j=i

8
Terms i = j cancel except for [w0 , . . . , wn ] − [v0 , . . . , vn ]. Terms i ≠ j are ”Γ(∂∆n × [0, 1])”.
w0 w1
Γ1 = [v0 , w0 , w1 ] − [v0 , v1 , w1 ]
ÿ

⟲ ∂Γ1 = [w0 , w1 ] − 
[v
0 , w1 ] + [v0 , w0 ]

− [v1 , w1 ] + 
[v
0 , w1 ] − [v0 , v1 ]

v0 v1

Let (σ ∶ ∆n → X) ∈ Cn (X) and σ × 1 ∶ ∆n × [0, 1] → X × [0, 1], (x, t) ↦ (σ(x), t). Define

h ∶ Cn (X) Ð→ Cn+1 (X × [0, 1])

σ z→ (σ × 1)♯ (Γn )

Then
∂h(σ) = ∂(σ × 1)♯ (Γn )
= (σ × 1)♯ ∂(Γn )
= (i1 )♯ (σ) − h∂(σ) − (i0 )♯ (σ)

and so ∂● h● + h● ∂● = (i1 )♯ − (i0 )♯ .

4.2 Homotopy invariance

Definition 4.5. Two continuous maps f0 , f1 ∶ X → Y are homotopic if there exists a continuous
map
F (x, 0) = f0 (x)
F ∶ X × [0, 1] Ð→ Y such that
F (x, 1) = f1 (x).
F is called a homotopy and we write f0 ∼ f1 . If A is a subset of X and f0 ∣A = f1 ∣A and the
homotopy F satisfies F (x, t) = f0 (x) = f1 (x) for all x ∈ A and all t ∈ [0, 1], then f0 and f1 are
said to be homotopic relative to A.
Note 4.6. Homotopy is an equivalence relation.
Definition 4.7. Two spaces X and Y are homotopy equivalent if there exist maps

f ∶ X Ð→ Y g ○ f ∼ idX
such that
g ∶ Y Ð→ X f ○ g ∼ idY .

X is called contractible if it is homotopy equivalent to a point.

Example 4.8. Rn is contractible. f ∶ Rn → ∗, g ∶ ∗ → 0 ∈ Rn . Then f ○ g = id∗ and

F ∶ Rn × [0, 1] Ð→ Rn F (x, 0) = 0 = g ○ f (x)

with
(x, t) z→ tx F (x, 1) = x = idRn (x)

Theorem 4.9 (Homotopy invariance). If f0 ∼ f1 , then (f0 )∗ = (f1 )∗ ∶ Hn (X) → Hn (Y ).

9
Geometric intuition
Y

X f1
z cycle
F

f0

f1 (z) − f0 (z) is a boundary.

Proof. Let F be a homotopy between f0 and f1 . Then f0 = F ○ i0 , f1 = F ○ i1 and

(f1 )♯ − (f0 )♯ = F♯ (i1♯ − i0♯ ) by naturality

= F♯ (∂● h● + h● ∂● ) by Main Example 4.3
= ∂● F♯ (h● ) + F♯ (h● )∂.

So F♯ (h● ) is a chain homotopy between f0 ♯ and f1 ♯ , and so f0 ∗ = f1 ∗

Corollary 4.10. If X and Y are homotopy equivalent then Hn (X) ≅ Hn (Y ) for n ≥ 0.

Proof. By definition, there exists

f ∶ X Ð→ Y g ○ f ∼ idX
such that
g ∶ Y Ð→ X f ○ g ∼ idY .

So g∗ ○ f∗ = (g ○ f )∗ = (idX )∗ = idHn (X) and f∗ ○ g∗ = idHn (Y ) . Hence f∗ ∶ Hn (X) → Hn (Y ) is an

isomorphism.
⎧
⎪
⎪Z for n = 0
Corollary 4.11. If X is contractible then Hn (X) ≅ Hn (∗) ≅ ⎨
⎪
⎩0 for n > 0.
⎪

5 The Snake Lemma and relative homology

This is a computational tool: ‘Divide and conquer, Part I’

Definition 5.1. The sequence of chain complexes
i j
0 Ð→ A● Ð
Ð→ B● Ð
Ð→ C● Ð→ 0 (∗)

is short exact if for all n ≥ 0

i j
0 Ð→ An Ð
Ð→ Bn Ð
Ð→ Cn Ð→ 0

is short exact.

10
Theorem 5.2 (The Snake Lemma). Given a short exact sequence of chain complexes (∗) there
exist connecting homomorphisms

δ ∶ Hn (C● ) Ð→ Hn−1 (A● )

such that the following sequence is long exact:

δ i∗ j∗ δ
⋯Ð
Ð→ Hn (A● ) Ð
Ð→ Hn (B● ) Ð
Ð→ Hn (C● ) Ð
Ð→ Hn−1 (A● ) Ð→ ⋯ Ð→ H0 (B● ) Ð→ H0 (C● ) Ð→ 0

Proof. Definition of δ:
Let c ∈ Zn (C● ). Since j is a surjection there exists b ∈ Bn such that j(b) = c. Then ∂b ∈ Bn−1
and j(∂b) = ∂j(b) = ∂c = 0. By exactness there exists a ∈ An−1 such that i(a) = ∂b. Define
δ[c] ..= [a].
Well-defined:
(1) a is a cycle: i(∂a) = ∂(i(a)) = ∂(∂b) = 0, and so ∂a = 0 as i is injective.
(2) Independence of choice of c: assume [c] = [c′ ]

/ b − b′ + i(e) Ô⇒ c′ − c = ∂c′′ and c′′ = j(b′′ )

b′′
Ô⇒ ∂b′′ = b − b′ + i(e) and as ∂∂b′′ = 0
  Ô⇒ ∂i(e)) = i(∂e) = ∂b − ∂b′ = ia − ia′
c′′ / c − c′
Ô⇒ [a] = [a′ ] ∈ Hn−1 (A● ).

Exactness at Hn (C● ): im j∗ ⊆ ker δ: let [c] ∈ im j∗ , then there exists b such that j∗ [b] = [c]. In
particular ∂b = 0 so if i(a) = ∂b then a = 0 so δ[c] = 0.
im j∗ ⊇ ker δ: let δ[c] = [a] = 0

/a Ô⇒ ∃a′ ∶ ∂a′ = a
a′
Ô⇒ ∂ia′ = i∂a′ = ia = ∂b where jb = c
 put b′ = b − ia′
b / ∂b
note jb′ = jb − jia′ = jb = c
  Ô⇒ ∂b′ = ∂b − ∂ia′ = ∂b − ∂b = 0
/0
c Ô⇒ [c] = [jb′ ] = j∗ [b′ ].
Exactness at Hn (A● ) and Hn (B● ) follow in a similar way.
Note 5.3. If A● ⊂ B● and C● = B● /A● with ∂(b + An ) = ∂b + An−1 then δ[b + An ] = [∂b].
Let (X, A) be a pair of spaces, i.e. a space X and a subspace A. Then the inclusion A ↪ X
induces an inclusion of chain complexes C● (A) ↪ C● (X). Define the relative homology of (X, A)
as
Hn (X, A) ..= Hn (C● (X)/C● (A)).
Note 5.4. A cycle in C● (X, A) ..= C● (X)/C● (A) is a chain x ∈ Cn (X) with ∂x ∈ Cn−1 (A) and
δ[x] = [∂x].

11
It will be convenient to introduce the reduced homology of a space, which is defined to be
H̃n (X) ..= ker(Hn (X) → Hn (pt)). Note that H̃n (X) = Hn (X) for n > 0. Furthermore, a
continuous map f ∶ X → Y defines a map of reduced homology groups.
Corollary 5.5. There is a long exact sequence in reduced homology

̃1 (A) Ð→ H
⋯ Ð→ H ̃1 (X) Ð→ H
̃1 (X, A) Ðδ
̃0 (A) Ð→ H
Ð→ H ̃0 (X) Ð→ H0 (X, A) Ð→ 0.

Proof. By the Snake Lemma

δ
⋯ Ð→ H1 (X, A) Ð
Ð→ H0 (A) Ð→ H0 (X) Ð→ H0 (X, A) Ð→ 0
≅
is exact. It is an exercise to show that H̃0 (A) → H
̃0 (X), and that A ↪ X maps Z Ð → Z
in H0 (A) = H̃0 (A) ⊕ Z and H0 (X) = H̃0 (X) ⊕ Z. So the sequence in reduced homology is
exact.
̃n (X) = 0 for all n ≥ 0 so
Example 5.6. (X, A) = (Dk , S k−1 ). Note Dk ≃ ∗ and so H
̃n (Dk , S k−1 ) ≅ H
H ̃n−1 (S k−1 )

for all n ≥ 0.
Let (X, A) and (Y, B) be two pairs of spaces. A map of pairs of spaces is a continuous map
f ∶ X → Y such that f (A) ⊂ B.
Proposition 5.7. A map of pairs of spaces induces a map of long exact sequences.

Proof. This is left as an exercise.

6 The Five Lemma and excision

Excision distinguishes homology from homotopy and makes the former more accessible to com-
putations.

6.1 A tool

Five Lemma 6.1. In the diagram below, if the rows are exact then γ is an isomorphism.

A /B /C /D /E

α ≅ β ≅ γ δ ≅ ε ≅
    
A′ / B′ / C′ / D′ / E′

Proof. Diagram chasing.

(1) β, δ surjective, ε injective Ô⇒ γ surjective: Let c′ ∈ C ′ , ∂c′ = d′ = δd.

∂d′ = ∂∂c′ = 0 Ô⇒ ∂(δd) = ε(∂d) = 0 ∂(c′ − γc) = 0 Ô⇒ ∃b′ ∶ ∂b′ = c′ − γc

Ô⇒ ∂d = 0 Ô⇒ ∃b ∶ ∂βb = c′ − γc
Ô⇒ ∃c ∶ ∂c = d. Ô⇒ γ(∂b) = c′ − γc
Ô⇒ c′ = γ(∂b + c).

12
(2) β, δ injective, α surjective Ô⇒ γ injective: similarly.

Example 6.2. Assume 0 /G α / H Ui β / K / 0 is split exact with β ○ γ = id. Ap-

_ i
γ
plying the Five Lemma to

0 /G /H /K /0
O
α⋅γ

0 /G / G⊕K /K /0

gives H ≃ G ⊕ K.

6.2 Excision

Definition 6.3. Let A ⊂ X be a subspace. A map r ∶ X → X is a retract onto A if r(X) = A

and r∣A = idA (so r2 = r). If r is a homotopy equivalence (rel A) then r is a deformation retract.

A pair (X, A) is called good if A is a non-empty, closed subset of X such X

that it is a deformation retract of some neighbourhood V of A in X. A
V

Define the quotient X/A as the quotient under the equivalence relation x ∼ y iff (x ∈ A and
y ∈ A) or (x = y).
Example 6.4. X = S 1 ∨ S 1 ⊇ A = S 1 V = X/A ≃ with [A] =

Example 6.5. X = Dk ⊇ A = S k−1 V = X/A ≃ S k with [S k−1 ] = a point.

Theorem 6.6 (Excision and relative homology as the homology of a space). Let A ⊂ V ⊂ X be
subspaces of X such that Ā ⊂ V̊ . Then the map of pairs (X ∖ A, V ∖ A) → (X, V ) induces an
≅
isomorphism Hn (X ∖ A, V ∖ A) Ð → Hn (X, V ) for all n ≥ 0.
Corollary 6.7. For a good pair (X, A), the quotient map (X, A) → (X/A, ∗) induces an iso-
≅
morphism Hn (X, A) Ð
→ Hn (X/A, ∗) = H̃n (X/A) for all n ≥ 0.

Proof of corollary. As (X, A) is good there exists V ⊃ A such that A ↪ V is a homotopy

equivalence. Consider the following commutative diagram.
≅ / Hn (X, V ) o ≅
Hn (X, A) Hn (X ∖ A, V ∖ A)
q q q
 
≅ / Hn (X/A, V /A) o ≅
Hn (X/A, A/A) Hn (X/A ∖ A/A, V /A ∖ A/A)

Using Proposition 5.7 and the Five Lemma one shows that the left horizontal arrows are isomor-
≃ ≃
phisms since A Ð
→ V and A/A Ð → V /A are homotopy equivalences. The right horizontal arrows
are isomorphisms by excision and the right vertical arrow is the identity as X/A ∖ A/A = X ∖ A.
The diagram commutes and therefore also the other two vertical arrows are isomorphisms.
̃n (Dk /S k−1 ) ≅ H
Example 6.8. (Dk , S k−1 ) is good Ô⇒ Hn (Dk , S k−1 ) ≅ H ̃n (S k ).

13
 ⎧
⎪Z
̃n (S k ) = ⎪ n=k
Claim 6.9. H ⎨
⎪
⎪ n≠k
⎩0
̃n (Dk , S k−1 ) ≅ H
Proof. Recall from section 5, H ̃n−1 (S k−1 ). So

⎧
⎪Z n = k
̃n (S k ) ≅ H
H ̃n (Dk , S k−1 ) ≅ H
̃n−1 (S k−1 ) ≅ ⋯ ≅ H ̃n−k (S 0 ) = ⎪
̃n−k+1 (S 1 ) ≅ H ⎨
⎪
⎩0 n ≠ k.
⎪

Remark 6.10. The generator of Hn (S n ) ≅ Hn (Dk , S k−1 ) ≅ Z can be represented by the n-cycle
∆n1 − ∆n2 where S n = (∆1 ∐ ∆2 )/(∂∆1 = ∂∆2 ) or by the relative cycle ∆n1 ≃ Dn . By excision
Hn (∆n1 , ∂∆n1 ) ≅ Hn (S n , ∆n2 ).

6.3 Locality

Let U = {Ui }i∈I be a collection of open subsets of X with X = ⋃i∈I Ůi . Define C●U (X) ⊆ C● (X)
to be the subcomplex generated by the n-simplices σ with σ(∆n ) ⊂ Ui for some i ∈ I.
≅
Theorem 6.11 (Locality). Hn (C●U (X)) Ð
Ð→ Hn (C● (X)) = Hn (X) for all n ≥ 0.

Sketch.
(1) Given an n-simplex σ ∶ ∆n → X, we can subdivide ∆n into smaller n-simplices such that
the image is contained in some Ui by applying repeatedly the barycentric subdivision:

Barycentric subdivision: S(∆1 ) = S(∆2 ) =

(2) S defines a chain map C● (X) → C● (X), σ ↦ σ ○ S with ∂S = S∂.

(3) S is chain homotopic to the identity:

T ∶ Cn (X) Ð→ Cn+1 (X)

proj σ
σ z→ T σ ∶ ∆n × I ÐÐ→ ∆n Ð
Ð→ X ∆1 × I = ∆2 × I =
and ∂T + T ∂ = S − id

Proof of Theorem 6.6 (Excision).

Let A ⊆ V ⊆ X, put U = X ∖ A and U = {U, V }. Then U

Cn (X ∖ A)/Cn (V ∖ A) = Cn (U )/Cn (V ∩ U ) ≅ CnU (X)/Cn (V ). A V

There is a map of long exact sequences

Hn (V ) / Hn (C U ) / Hn (C U /C● (V )) / Hn−1 (V ) / Hn−1 (C U (X))

● ● ●

≅ ≅
  
Hn (V ) / Hn (X) / Hn (C● (X)/C● (V )) / Hn−1 (V ) / Hn−1 (X)

Hn (X, V )

14
So by the Five Lemma:

Hn (X ∖ A, V ∖ A) ≅ Hn (C●U /C● (V )) ≅ Hn (X, V ).

6.4 Mayer-Vietoris sequence

B
‘Divide and conquer: Part II’
Let X be a topological space and A, B ⊂ X with Å ∪ B̊ ⊇ X.
A
Theorem 6.12. With A, B and X as above there is a long exact sequence

⋯ Ð→ Hn (A ∩ B) Ð→ Hn (A) ⊕ Hn (B) Ð→ Hn (X) Ð→ Hn−1 (A ∩ B) Ð→ ⋯

⋯ Ð→ H0 (A ∩ B) Ð→ H0 (A) ⊕ H0 (B) Ð→ H0 (X) Ð→ 0.

Similarly, there is a long exact sequence in reduced homology.

Proof. Applying locality (Theorem 6.11) to U = {A, B},

C●U (X) Ð→ C● (X)

induces an isomorphism in homology. Consider the short exact sequence

0 / C● (A ∩ B) / C● (A) ⊕ C● (B) / C U (X) /0

●
σ / (σ, −σ)
 / σ+µ
(σ, µ)

and apply the Snake Lemma.

Note 6.13. δ[σ + µ] = [∂σ] = [−∂µ].
Definition 6.14. The cone on X is the quotient space CX ..=
(X × [0, 1]/ ∼) ≃ ∗ where (x, s) ∼ (y, t) if and only if (x, s) = (y, t)
or s = t = 1.
The suspension on X is the quotient space ΣX ..= X × [0, 1]/ ∼ where
(x, s) ∼ (y, t) if and only if (x, s) = (y, t) or s = t = 0 or s = t = 1.
Example 6.15. Let A = (X × [0, 2/3]/∼) ⊂ ΣX and let B = (X × [1/3, 1]/ ∼) ⊂ ΣX, then
A ≃ B ≃ CX ≃ ∗ and A ∩ B = X × [1/2, 2/3] ≃ X. Apply Mayer-Vietoris to get
̃n−1 (X).
Hn (ΣX) ≅ H

7 The degree of a map of spheres

̃n (S n ) ≃ Z for n ≥ 0.
Let f ∶ S n → S n be a continuous map. Recall H
Definition 7.1. f∗ in dimension n is a homomorphism Z → Z, i.e. f∗ in dimension n is multi-
plication by an integer deg(f ), the degree of f .

15
Note 7.2. (i) deg(idS n ) = 1 as (idS n )∗ = idHn
(ii) deg(f ○ g) = deg(f ) ⋅ deg(g) as (f ○ g)∗ = f∗ ○ g∗
(iii) f ≃ g Ô⇒ deg(f ) = deg(g) as f∗ = g∗
(iv) f ≃ ∗ Ô⇒ deg(f ) = 0 as f∗ = 0.
Example 7.3.
(1) f is induced by the reflection in Rn i × {0} ⊂ Rn+1 , then 1 0
deg(f ) = −1:
Hn (S n ) = ⟨∆1 − ∆2 ⟩, f interchanges ∆1 and ∆2 . So 2
f∗ (∆1 − ∆2 ) = (∆2 − ∆1 ) = −(∆1 − ∆2 ).
(2) f is the antipodal map S n → S n , x ↦ −x, then deg(f ) = (−1)n+1 :
−1 0 ⋯ 0 1 0 ⋯0 1⋯0 0
0 −1 ⋮ ⋱ ⋮
f is induced by −I ∶ Rn+1 → Rn+1 , −I = ( 0 1
⋮ ⋱ ⋮ )( ⋮
0
⋱ ⋮ )⋯( 0
0
1 0 ).
0 0⋯1 0 0 ⋯1 0 ⋯ 0 −1
Each of the n + 1 matrices on the right hand side is a reflection homotopic to the one in (1)
hence, using (ii) and (iii), deg(f ) = (−1)n+1 .

Application 7.4. A continuous function v ∶ S n → Rn+1 defines

a tangent vector field on S n if v(x) ⊥ x for all x ∈ S n . S n has a
continuous nowhere zero tangent vector field if and only if n is odd.

Proof. Suppose v ∶ S n → Rn+1 is continuous with v(x) ⊥ x for all x. Suppose further v(x) ≠ 0
v(x)
for all x. Consider F ∶ S n × [0, 1] → S n given by F (x, t) = (cos πt)x + (sin πt) ∣v(x)∣ .

Note F (x, 0) = x and F (x, 1) = −x. Hence F defines a homotopy v(x)

from the identity map to the antipodal map. Hence they must have ∣v(x)∣

the same degree (by (iii)) and 1 = (−1)n+1 . So n must be odd.

Conversely, if n is odd, define −x x

v ∶ S n Ð→ Rn+1
(x1 , . . . , x2k ) z→ (−x2 , x1 , . . . , −x2k , x2k−1 ).
Corollary 7.5 (Hairy ball theorem). You cannot comb a hairy ball (S 2 ).
Definition 7.6. Let f (x) = y and U ∋ x, V ∋ y be open neighbourhoods of x and y respectively
such that
f ∶ (U, U ∖ x) Ð→ (V, V ∖ y) (∗)
i.e. the only point in U mapping to y is x.
From the relative long exact sequence and excision we have
(f ∣x )∗
̃n (S n ) ≅ Hn (S n , S n ∖ x) ≅ Hn (U, U ∖ x) ÐÐÐ→
Z≅H ̃n (S n ) ≅ Z.
Hn (V, V ∖ y) ≅ H

Then (f ∣x )∗ is multiplication by an integer deg(f ∣x ), the local degree of f at x.

16
Proposition 7.7. Assume f −1 (y) = {x1 , . . . , xk } and for each
i = 1, . . . , k there exist disjoint open sets Ui ∋ xi each satisfying
condition (∗). Then
k x2
deg(f ) = ∑ deg(f ∣xi ). x1 y
i=1

If f −1 (y) = ∅ then deg(f ) = 0.

Proof. The proposition follows from the following commutative diagram:

f∗
̃n (S n )
H /H
̃n (S n )
q ≅
 f∗ 
Hn (S n , S n ∖ {x1 , . . . , xk }) / Hn (V, V ∖ y)
4
jj j jj jj
≅ jjjj
 jjjj ∑i (f ∣xi )∗
⊕ki=1 Hn (Ui , Ui ∖ xi )

Example 7.8. Let f ∶ S 2 = C ∪ {∞} → S 2 be define by f (z) = z n . Let y = 1, then

f −1 (y) = {e2πki/n ∶ k = 1, . . . n}. Since f is orientation preserving and is a homeomorphism near
each root of unity, deg(f ∣e2πki/n ) = 1. Hence deg(f ) = n.
• For p an arbitrary polynomial of degree n, p ≃ f (via a homotopy that moves all zeros to
0). Hence deg(p) = n.
• deg(f ∣0 ) = n.

8 Cellular homology

Definition 8.1 (CW-complex). Inductive construction:

(1) Start with X 0 , a disjoint set (X −1 = ∅).
(2) Let Dαn be an n-disk and φα ∶ ∂Dαn ≃ S n−1 → X n−1 be a continuous map. For a collection
{Dαn , φα } construct X n from X n−1 by X n ..= X n−1 ∐α Dαn /x ∼ φα (x) for all x ∈ ∂Dαn .
(3) X = ⋃n X n is given the weak topology (A ⊂ X is open if and only if A ∩ X n is open for all
n).
D̊αn = enα is called an n-cell. Each cell has a characteristic map enα ↪ X.
Example 8.2.

X 0 = pt
X1 =
X2 = ⋃D
2
φ ∶ S 1 → S 1 of degree 3
φ

17
Example 8.3. Real projective space - the space of lines in Rn+1 .
X 0 = pt
X1 = = RP 1
X 2 = X 1 ⋃ D2 = RP 2 φ ∶ S 1 → S 1 of degree 2
RP = S /v ∼ −v
n n
φ

⋮
X = X n−1 ⋃ Dn
n
φ ∶ S n−1 → RP n−1 , x ↦ {±x}
φ

Example 8.4. Complex projective space - the space of complex lines in Cn+1 .
X 0 = pt = CP 0
X1 = X0
X 2 = CP 0 ⋃ D2 = CP 1 φ ∶ S 1 → pt
φ

CP = S
n 2n+1
/v ∼ λv, λ ∈ S 1
X =X
3 2

X 4 = CP 1 ⋃ D4 = CP 2 φ ∶ S 3 → S 3 /S 1 ≃ S 2
φ

⋮
X 2k+1 = X 2k

Let X be a cell complex with m skeleton X m , m ≥ 0. Then (X m , X m−1 ) is good and the
quotient space X m /X m−1 ≃ ⋁ S m is a wedge of m-spheres, one for each m-cell.
Definition 8.5. Define Cm CW
(X) ..= Hm (X m , X m−1 ) = free abelian group generated by the
m-cells. Define the boundary map
δ q
d ∶ Cm
CW
(X) = Hm (X m , X m−1 ) Ð→ Hm−1 (Xm−1 ) Ð→ Hm−1 (Xm−1 , Xm−2 ) = Cm−1
CW
(X).

The boundary homomorphism δ is induced by the attaching maps, δ[Dαm ] = [∂Dαm ≡ φα (S m−1 )].
α and an m − 1 cell eβ
Therefore d also has the following description. For an m-cell em m−1
define

dαβ ..= deg ( S m−1 / X m−1 / X m−1 /X m−2 ≃ ⋁ S m−1 / S m−1 )

φα q

∂Dαm Dβm−1 /∂

α ) .= ∑β dαβ eβ
and d(em . m−1
. This is a finite sum as the image of a compact set is compact.
Lemma 8.6. d ○ d = 0

Proof. Consider the following three long exact sequences in relative homology.

Hm+1 (X m+1 , X m ) / Hm (X m )
δ / Hm (X m+1 ) / (∗∗)
jjj
jj
jjjj
jjjj
/ Hm (X m ) / Hm (X m , X m−1 ) / Hm−1 (X m−1 )
δ
q∗ j
j
jjjjjjj
jjjj
/ Hm−1 (X m−1 ) / Hm−1 (X m−1 , X m−2 )
q∗

18
d ○ d = (q∗ δ)(q∗ δ) = q∗ (δq∗ )δ = 0 as δq∗ = 0 by the exactness of the middle row.
Definition 8.7 (Cellular homology).

HnCW (X) ..= Hn (C●CW (X), d)

Remark 8.8. A ∆-complex X● is naturally a cell-complex with X n the union of all k simplices,
k ≤ n. Furthermore, with this identification there is an isomorphism of chain complexes

(C●CW (X● ), d) ≅ (C●∆ (X● ), ∂).

Example 8.9. S m = ∗ ∪φ Dm , φ ∶ ∂Dm = S m−1 → ∗

m≥2 0 Ð→ Zm Ð→ ⋯ Ð→ 0 Ð→ Z0 Ð→ 0
d=0
m=1 0 Ð→ 0 Ð→ ⋯ Ð→ Z ÐÐ→ Z Ð→ 0 φ ∶ S 0 → ∗, deg φ = 0
⎧
⎪
⎪Z n = 0, m
so Hn (S ) =
m
CnCW (S n )
=⎨
⎪
⎩0 n ≠ 0, m.
⎪
In particular we deduce that the cell complex above for S m has the minimal number of cells as
we must have at least as many cells as generators for the homology.
CW
Theorem 8.10. For any cell complex X, Hm (X) = Hm (X), and hence also for any ∆-
complex, Hm (X● ) = Hm (X● ).
∆

Proof. We note that Hk (X m+1 , X m ) is zero for k ≠ m + 1 as X m+1 /X m is a wedge of m + 1-

spheres. Using induction and the l.e.s. for triples of space (X m+l , X m+1 , X m ), one shows that
Hk (X) = Hk (X m+1 ) for all k ≤ m, and Hk (X m ) = 0 for all k > m.

Hm (X) ≅ Hm (X m+1 ) (from relative l.e.s.)

δ
≅ Hm (X m )/ im (Hm+1 (X m+1 , X m ) Ð
→ Hm (X m )) (from first row in (∗∗))

q ∶ Hm (X m ) Ð→ Hm (X m , X m−1 ) is an injection as Hm (X m−1 ) = 0. So im δm+1 ≅ im dm+1

(as d = q ○ δ). Now ker dm = ker δm ≅ im q∗ ≅ Hm (X m ) so Hm CW
(X) = ker dm / im dm+1 ≅
Hm (X )/ im δm+1 ≅ Hm (X).
m

Example 8.11. Real projective plane (cf. Example 8.3).

RP n = S n /∼ x ∼ y iff x = ±y
= space of lines in Rn+1 .
RP 0 = S 0 /∼ = ∗
RP 1 = S 1 /∼ ≅ S 1
RP 2 =
RP 3 ≅ SO(3) = {A ∈ M3×3 (R) ∶ det A = 1, AAt = I}.

RP n has a cell structure with one cell in each dimension m = 0, . . . , n and m-skeleton
RP m = RP m−1 ⋃φ Dm where φ ∶ ∂Dm = S m−1 → RP m−1 = S m−1 / ∼ is the natural quotient
map.
φ q
Ð→ S m−1 /∼ = RP m−1 Ð
d = dm = deg(S m−1 Ð Ð→ RP m−1 /RP m−2 ≅ S m−1 )

19
q ○ φ∣∆
˚ + , q ○ φ∣∆
˚ − are homeomorphisms onto their images which are related to each other via the
antipodal map.
Ô⇒ d = dm = deg(q ○ φ) = deg(id) + deg(antipodal map)
= 1 + (−1)m
⎧
⎪ 0 2 2 0 0
⎪0 Ð→ Z Ð
Ð→ Z ÐÐ→ ⋯ Ð
Ð→ Z Ð
Ð→ Z Ð
Ð→ 0 m odd
Ô⇒ C●CW (RP n )
=⎨
⎪
⎪
0 2 2 0 0
⎩0 Ð→ 0 Ð
Ð→ Z Ð
Ð→ ⋯ ÐÐ→ Z Ð
Ð→ Z Ð
Ð→ 0 m even
⎧
⎪ m=0
⎪
⎪
⎪
Z
⎪
⎪
⎪Z/2 m odd and m < n
Ô⇒ Hm (RP n ) = ⎨
⎪
⎪
⎪Z m = n odd
⎪
⎪
⎪
⎪
⎩0 otherwise.

9 Cohomology and the Universal Coefficient Theorem

9.1 Cohomology

Definition 9.1. Let (C● , ∂● ) be a chain complex of free Z-modules (or F-vector spaces). Define
• n-cochains: C n ..= Hom(Cn , Z), group of group homomorphisms (or the dual vector space)
• coboundary map: ∂ n ∶ C n → C n+1 , φ ↦ φ ○ ∂
• n-cocycle: Z n = ker(∂ n )
• n-coboundary: B n = im(∂ n−1 )
• n-th cohomology group: H n (C, ∂) ..= Hn (C ● , ∂ ● ) = Z n /B n .
We note that Hom(Zn , Z) ≃ Zn , and if a matrix A represents an element in Hom(Zn , Zm then
the dual map is represented by its transpose.
Example 9.2. The cellular chain complex for RP 3 is
0 2 0
0 Ð→ Z Ð
Ð→ Z Ð
Ð→ Z Ð
Ð→ Z Ð→ 0.
As Hom(Z, Z) ≅ Z the dual complex is
0 2 0
0 ←Ð Z ←Ð
Ð Z ←Ð
Ð Z ←Ð
Ð Z ←Ð 0.
Hence
⎧
⎪ n=0
⎪
⎪
⎪
Z
⎪
⎪
⎪0 n=1
⎪
⎪
⎪
H (RP ) = ⎨Z/2
n 3
n=2
⎪
⎪
⎪
⎪
⎪
⎪Z n=3
⎪
⎪
⎪
⎪
⎩0 n > 3.

Induced Homomorphism. f ∶ X Ð→ Y induces

f♯ ∶ Cn (X) Ð→ Cn (Y ) and hence
f ♯ ∶ C n (Y ) Ð→ C n (X), φ z→ φ ○ f♯
f ∗ ∶ H n (Y ) Ð→ H n (X)

20
Lemma 9.3. If f● , g● ∶ (C● , ∂● ) → (C ̃● , ∂̃● ) are chain homotopic then so are f ● , g ● ∶ (C
̃● , ∂̃● ) →
̃● , ∂̃● ) → H n (C● , ∂● ).
(C ● , ∂ ● ) and f ● = g ● ∶ H n (C

Proof. If f● − g● = ∂̃● h● + h● ∂● then f ● − g ● = h● ∂̃● + ∂ ● h● .

Homotopy Invariance. If f ≃ g ∶ X → Y then f ∗ = g ∗ ∶ H n (Y ) → H n (X).

i● j●
→ C● Ð→ B● → 0 is an exact sequence of free chain complexes then so
Lemma 9.4. If 0 → A● Ð
i● j●
Ð C ● ←Ð B ● ← 0.
is 0 ← A● ←

Proof. Consider 0 → An → Cn → Bn → 0. As Bn is free there is a splitting s ∶ Bn → Cn and

i⊕s
Cn ÐÐ→ An ⊕ Bn . Hence
≅

C n = Hom(Cn , Z) ≅ Hom(An , Z) ⊕ Hom(Bn , Z) ≅ An ⊕ B n

φ ↦ ((φ ○ i), (φ ○ s))

and ker i∗ = im j ∗ .

i∗
Excision. If Z̄ ⊂ Å ⊂ X then H n (X, A) Ð→ H n (X ∖ Z, A ∖ Z).
≅

Long Exact Relative Cohomology Sequence. For any (X, A) there is a long exact se-
quence
δ1
⋯ ←Ð H 1 (X, A) ←Ð
Ð H 0 (A) ←Ð H 0 (X) ←Ð H 0 (X, A) ←Ð 0.

Mayer-Vietoris. For Å ∪ B̊ = X there is a long exact sequence

δ1
⋯ ←Ð H 1 (X) ←Ð
Ð H 0 (A ∩ B) ←Ð H 0 (A) ⊕ H 0 (B) ←Ð H 0 (X) ←Ð 0.

9.2 Universal Coefficient Theorem

Relating H ∗ to H∗ .
Let (C● , ∂● ) be a chain complex of free abelian groups (or vector spaces). Then Cn , Bn and Zn
are all free. Hence the short exact sequence
∂n
0 Ð→ Zn Ð→ Cn Ð
Ð→ Bn−1 Ð→ 0 (9.1)

gives rise to a short exact sequence of dual chain complexes

∂n
0o ZO n o CO n o B n−1
O
o 0
∂ n =0 ∂n ∂ n−1 =0
∂ n−1
0o Z n−1 o C n−1 o B n−2 o 0

21
and a long exact sequence in homology

⋯ ←Ð B n ←Ð Z n ←Ð H n (C● ) ←Ð B n−1 ←Ð Z n−1 ←Ð ⋯

so there are short exact sequences

0o ker in So H n (C● ) o coker in o 0 (∗)

_ k6
∗
and ker in = {φ ∈ Z n ∶ φ∣Bn ≅ 0} = {φ ∈ (Zn /Bn ) } = (Hn (C● ))∗ .

If we are working over a field and assuming that H n (C● ) is finite dimensional, coker in = 0 as
every f ∈ B n−1 can be extended to f̃ ∈ Z n−1 .
Theorem 9.5. If (C● , ∂● ) is a chain complex of vector spaces with finite dimensional cohomol-
ogy, then H n (C● ) = (Hn (C● ))∗ .
Corollary 9.6. For any space X of finite type and any field F, H n (X; F) ≅ (Hn (X; F))∗ .
Let (C● , ∂● ) now be a chain complex of free, finitely generated abelian groups. Then
d1
⎛ ⋱ ⎞
⎜ dk ⎟
⎜ 1 ⎟
in−1 = ⎜ ⋱ ⎟ ∶ Bn−1 Ð→ Zn−1
⎜ 1⎟
⎜ 0 ⋯ ⋯ ⋯ ⋯ 0⎟
⎝ ⋮ ⋮⎠
0 ⋯ ⋯ ⋯⋯0

and hence Hn−1 = Zl ⊕ (Z/d1 Z) ⊕ ⋯ ⊕ (Z/dk Z). Also

d1 0⋯0
d1 ⎛ t ⋱ ⋮ ⋮⎞
n−1
=( ⋱ ) =⎜ dk ⋮ ⋮⎟
i 1 ⎜ 1 ⋮ ⋮⎟
0 ⋯0 ⎝ ⋱ ⋮ ⋮⎠
10⋯0

and
coker in−1 = B n−1 / im in−1 ≅ (Z/d1 Z) ⊕ ⋯ ⊕ (Z/dk Z) ≅ Tor(Hn−1 )
Theorem 9.7. If (C● , ∂● ) is a chain complex of finitely generated free abelian groups then

H n (C● ) ≅ (Hn (C● ))∗ ⊕ Tor(Hn−1 (C● ))

Corollary 9.8. For a cell complex of finite type, i.e. a complex with finitely many cells in each
dimension,
H n (X) ≅ (Hn (X)/ Tor(Hn (X))) ⊕ Tor(Hn−1 (X)).
Furthermore, (Hn (C● ))∗ is free and so the short exact sequence (∗) splits.
Remark 9.9. A map of chain complexes C● → C ̃● induces a map of exact sequences in (∗).
However, the splitting in Theorem 9.7 does not correspond, i.e. the splitting may not be natural.
Example 9.10.

H0 (RP 3 ) = Z H1 (RP 3 ) = Z/2Z H2 (RP 3 ) = 0 H3 (RP 3 ) = Z

H 0 (RP 3 ) = Z H 1 (RP 3 ) = 0 H 2 (RP 3 ) = Z/2Z H 3 (RP 3 ) = Z

22