LECTURES ON HOMOLOGY THEORY

DRAFT VERSION

SERGEY MOZGOVOY

Contents
1. Triangulated spaces and ∆-sets 2
1.1. Motivation 2
1.2. Triangulated spaces 3
1.3. Geometric realization 8
1.4. Simplicial complexes 10
1.5. Product triangulation 13
1.6. Barycentric subdivision 14
1.7. Simplicial homology 15
2. Homological algebra 19
3. Singular homology 23
3.1. Definition and basic properties 23
3.2. Homotopy invariance 25
3.3. Relative homology groups 27
3.4. Excision and Mayer-Vietoris 28
3.5. Long exact sequence for good pairs 30
3.6. Equivalence of simplicial and singular homologies 32
3.7. CW-complexes and cellular homology 33
4. Applications of Homology theory 35
4.1. Degree 35
4.2. Hedgehog theorem 36
4.3. Jordan curve theorem 36
4.4. Invariance of domain 38
4.5. Algebraic applications 38
4.6. Borsuk-Ulam theorem 39
4.7. Lefschetz fixed point theorem 41
Appendix A. Preliminaries 42
A.1. Relations 42
A.2. Quotient topology 43
Appendix B. Categories and functors 45
Appendix C. Simplicial approximation 47
Index 48

Date: March 13, 2019.

2 HOMOLOGY THEORY

1. Triangulated spaces and ∆-sets

1.1. Motivation. Algebraic topology relates problems of topology and algebra. At the first
level which is the content of this course one reduces topological problems to algebra and in
particular to linear algebra. At the next level one reduces algebraic problems to topology.
The main approach is the following. Given a topological space X, one constructs an algebraic
object H(X). This can be a group (like a fundamental group), a vector space, an algebra, etc.
This construction is usually functorial, meaning that given a continuous map f : X → Y , there
is a homomorphism f∗ : H(X) → H(Y ). In this way we obtain a functor (see Appendix B) from
the category of topological spaces to the category of groups, vector spaces, algebras, etc. The
objective is to relate topological properties of X to algebraic properties of H(X). We will study
just one construction of this type, called the homology theory.
As applications of the theory that we will develop, we will later prove the following statements:
(1) Rn are not homeomorphic to each other for different n 3.35.
(2) A continuous map f : Dn → Dn has a fixed point (Brouwer theorem) 3.37.
(3) One can not comb a hedgehog smoothly, meaning that there is no continuous non-
vanishing tangent vector field on S 2 4.4.
(4) If f : S 1 → R2 is injective and continuous, then R2 \f (S 1 ) consists of exactly two connected
components (Jordan curve theorem) 4.7.
(5) One can not embed S n in Rn 4.10. Generally, if f : M → N is an embedding of topological
n-manifolds, with compact M and connected N , then f is a homeomorphism 4.13.
(6) The field C is algebraically closed (Fundamental theorem of algebra) 4.17.
(7) If f : S n → Rn is continuous, then there exists x ∈ S n with f (x) = f (−x) (Borsuk-Ulam
theorem) 4.18.
(8) If F1 , F2 , F3 is a closed covering of S 2 , then at least one Fi contains antipodal points
(x, −x ∈ Fi ). Generally, if F1 , . . . , Fn+1 is a closed covering of S n , then at least one Fi
contains antipodal points (Borsuk-Ulam theorem) 4.18.
Example 1.1. Let us consider a finite connected graph on a plane. It consists of several points
(called vertices) connected by line segments (called edges), without intersections. Connected
components of the complement are called faces (we consider also the unbounded component).
We assume that all bounded faces are homeomorphic to an open disc. Let v, e, f be the numbers
of vertices, edges and faces respectively. Then Euler’s formula states that
v − e + f = 2.
For example, consider a graph consisting of one point. Then v = 1, e = 0, f = 1 and the formula
is satisfied. For a triangle on a plane, we have v = e = 3, f = 2 and the formula is again
satisfied.
The above decomposition of R2 can be interpreted as a decomposition (also called a triangu-
lation if all faces are triangles) of the two-dimensional sphere S 2 – the sphere is obtained from
R2 by adding one point at infinity and we consider the unbounded component as containing this
additional point. The left hand side of Euler’s formula can be associated with any triangulation
of S 2 . According to the formula, this number is independent of the triangulation, hence it is
an invariant of S 2 , called the Euler characteristic of S 2 . Homology theory in a nutshell is a
generalization of Euler’s formula to other topological spaces.
 HOMOLOGY THEORY 3

1.2. Triangulated spaces. By a space we will always mean a topological space. By a map
between spaces we will always mean a continuous map, unless otherwise stated. We will study
spaces that can be obtained by gluing together points, segments, triangles and higher-dimensional
building blocks, called simplices. The structure that one obtains is called a triangulated space.
Its combinatorial counterpart is called a ∆-set or a semi-simplicial set. Having this combinatorial
structure, one can apply linearization to it and get an algebraic structure (an abelian group or
a vector space), called the homology of the original space.
Definition 1.2. Define the standard n-dimensional simplex (or n-simplex) to be the
topological space n X o
∆n = (t0 , . . . , tn ) ∈ Rn+1 ti = 1, ti ≥ 0 ∀i .

Example 1.3. ∆0 is a point, ∆1 is a line segment (edge), ∆2 is a triangle, ∆3 is a tetrahedron.

e1 ∆1 e2 ∆2

e0 e1
e0

Remark 1.4. Let us define an n-dimensional disc

Dn = {x ∈ Rn | kxk ≤ 1} ,
an n-dimensional sphere
S n = x ∈ Rn+1 kxk = 1 ,


and an n-dimensional cube

I n = |I × .{z
. . × I}, I = [0, 1].
n times
Then
∆n ' Dn ' I n
and ∂∆n ' ∂I n ' ∂Dn = S n−1 . We can obtain S n by gluing two hemispheres D±
n
(both
n n−1
homeomorphic to D ) along their boundary S .
For every n ≥ 0, define [n] = {0, . . . , n}. We will consider it as an ordered set.
Definition 1.5.
(1) For every i ∈ [n], define the i-th vertex of ∆n to be the point ei ∈ ∆n with ti = 1.
Note that every point t ∈ ∆n can be uniquely written in the form t = ni=0 ti ei , where
P
P
ti ≥ 0 and ti = 1.
(2) More generally, for every subset I ⊂ [n], define the I-th face ∆I ⊂ ∆n to be
∆I = {t ∈ ∆n | ti = 0 for i ∈
/ I} .
This face is a simplex of dimension #I − 1.
(3) For any map f : [m] → [n], consider a map
X
f∗ = ∆f : ∆m → ∆n , (s0 , . . . , sm ) 7→ (t0 , . . . , tn ), tj = si
f (i)=j
P P
or equivalently, i si ei 7→ i si ef (i) . Every subset I ⊂ [n] can be identified with a
(strictly) increasing map f : [m] → [n] having the image I. Then f∗ : ∆m → ∆n is injective
and f∗ (∆m ) = ∆I . We call it the f -face of ∆n .
4 HOMOLOGY THEORY

(4) A dimension n−1 face of ∆n is called a facet. There are n+1 facets in ∆n , corresponding
to coface maps δi : [n − 1] → [n] which are increasing maps that miss i ∈ [n].
(5) Define the boundary ∂∆n to be the union of all facets. It consists of t ∈ ∆n with at
least one ti = 0. ◦
(6) Define the open simplex ∆n to be the interior of ∆n
◦
∆n = {t ∈ ∆n | ti > 0 ∀i} = ∆n \∂∆n .
◦
Note that ∂∆0 = ∅ and ∆0 = ∆0 .
Example 1.6. Consider all faces of ∆2 . Observe how they correspond to subsets of [2] or
to increasing maps f : [m] → [2] for m ≤ 2. Consider f : [0] → [2] for 0 ≤ i ≤ 2 and the
corresponding maps f∗ : ∆0 → ∆2 .
Definition 1.7. A triangulation K of a space X is a collection of maps
(φσ = σ: ∆n → X)σ∈K ,
where n depends on σ which is called a simplex of dimension n = dim σ and
◦
(1) The restriction σ|∆◦ n is injective and X is the disjoint union of cells eσ = σ(∆n ).
(2) The restriction of σ to a face of ∆n is one of the maps τ : ∆m → X with τ ∈ K.
(3) A subset A ⊂ X is open ⇐⇒ σ −1 (A) is open in ∆n for each σ.
The pair (X, K), consisting of a space X with a triangulation K, is called a triangulated
space (or a ∆-complex).
Remark 1.8. The map σ: ∆n → X is not necessarily injective even though its restriction to the
◦
open simplex ∆n is injective. Nevertheless, we will often identify σ with its image σ(∆n ) ⊂ X
(especially on the drawings of triangulations).
Example 1.9.
(1) Circle S 1 as a space homeomorphic to ∂∆2 . There are 3 0-simplices and 3 1-simplices.
v1 a

b v0

v2 c
(2) Circle obtained by gluing one point and one interval.

a v0

We have K = {v0 , a}, where v0 is a 0-simplex and a is a 1-simplex. Consider S 1 =

{z ∈ C | |z| = 1} and φv0 : ∆0 → S 1 , pt 7→ 1,
φa : ∆1 ' [0, 1] → S 1 , t 7→ e2πit .
(3) Circle obtained by gluing two points and two intervals.
a

v1 v0

b
 HOMOLOGY THEORY 5

(4) Sphere S 2 as a space homeomorphic to ∂∆3 (boundary of a tetrahedron). There are 4

0-simplices, 6 1-simplices and 4 2-simplices. Note that 4 − 6 + 4 = 2 in accordance with
Euler’s formula.
(5) Sphere S 2 obtained by gluing 3 points, 3 intervals between them and 2 triangles at the
top and at the bottom. Note that 3 − 3 + 2 = 2 in accordance with Euler’s formula.

Let Kn be the set of all n-simplices of a triangulation. For any increasing map f : [m] → [n]
the composition σf∗ of f∗ : ∆m ,→ ∆n and σ: ∆n → X is equal to τ : ∆m → X for a unique
τ ∈ Km which we denote by f ∗ (σ) ∈ Km :
τ
m f∗ σ
∆ ∆n X
In this way we obtain a map f ∗ : Kn → Km for every increasing map f : [m] → [n]. Given two
g f
increasing maps [`] →
− [m] →
− [n], consider
g∗ f∗
∆l ∆m ∆n
τ σ
γ
X
Then
γ = (f g)∗ (σ) = σ(f g)∗ = (σf∗ )g∗ = g ∗ (f ∗ σ),
hence hence (f g)∗ = g ∗ f ∗ . This data comprises a ∆-set.

Definition 1.10. A ∆-set (or a semi-simplicial set) K is a collection of sets (Kn )n≥0 together
with maps f ∗ = Kf : Kn → Km , for every increasing map f : [m] → [n], such that

1∗ = 1,
g f
(f g)∗ = g ∗ f ∗ , [`] →
− [m] →
− [n].
F
Elements of Kn are called n-dimensional simplices of K. We denote n≥0 Kn by K.

Example 1.11. The faces of ∆r are parametrized by subsets I ⊂ [r] or by increasing maps
σ: [n] → [r], for all 0 ≤ n ≤ r. This ∆-set is denoted by ∆[r]. The set ∆[r]n of n-simplices
consists of increasing maps σ: [n] → [r]. If f : [m] → [n] is an increasing map, then f ∗ is given by
f ∗ : ∆[r]n → ∆[r]m , ∆[r]n 3 σ 7→ σ ◦ f ∈ ∆[r]m .

Remark 1.12. Given a ∆-set K, let ∂K be a ∆-set obtained from K by removing all maximal
simplices (simplices σ ∈ Km that are not of the form f ∗ (τ ) for f : [m] → [n], m < n). For
example ∂∆[n] is obtained from ∆[n] by removing the unique n-simplex [n] → [n]. The simplices
of ∂∆[n] correspond to all proper non-empty subsets I ⊂ [n]. Note that ∂∆[n] is a triangulation
of the boundary ∂∆n ' S n−1 .

Remark 1.13. Let K be a ∆-set. For every 0 ≤ i ≤ n, consider the map

v̄i : [0] → [n], 0 7→ i,
and the corresponding map vi : Kn → K0 = V , called the i-th vertex map. Given a 1-simplex
a ∈ K1 with x0 = v0 (a) and x1 = v1 (a), we will depict it as an arrow:
x0 a x1
Assume that the map
v: Kn → V n+1 , σ 7→ (v0 (σ), . . . , vn (σ)),
6 HOMOLOGY THEORY

is injective. This means that every n-simplex is uniquely determined by the sequence of its
vertices. Then it is customary to represent a simplex σ ∈ Kn in the form

σ = [x0 , . . . , xn ], xi = vi (σ).

In many examples we will have the above injectivity and use the described representation of
simplices. We will usually denote a 0-simplex [v] simply by v.

Example 1.14.
(1) Triangulation of an interval. We have K0 = {v0 , v1 } and K1 = {[v0 , v1 ]}.

v0 a v1
1
(2) Triangulations of S :
(1) K0 = {v0 }, K1 = {[v0 , v0 ]} (one vertex, one loop).
(2) K0 = {v0 , v1 }, K1 = {[v0 , v1 ], [v1 , v0 ]}.
(3) K = ∂∆[v2 ].
(3) Consider the canonical triangulation K of the standard simplex ∆2 :
2

c b

0 a 1

Every face of ∆2 should be a simplex of the triangulation. Therefore we have

K0 : 0, 1, 2, K1 : a = [0, 1], b = [1, 2], c = [0, 2], K2 : σ = [0, 1, 2].

Note that simplices of K correspond to all subsets of [2] = {0, 1, 2}.

(4) Standard triangulation of S n−1 , represented as ∂∆n , and the corresponding ∆-set ∂∆[n].

Example 1.15. Triangulations of S 2 , Torus, Klein bottle, Möbius strip.

b 2 b b b
1 2S 0 0 T 0 0 K 1 0M
σ σ σ σ
a c b a c a a c a a c a
τ τ τ τ
0 a 1 0 0 0 0 0 1
b b d

Example 1.16. Consider the real projective plane RP 2 = (R3 − {0})/R∗ ' S 2 /Z2 (see
Example A.4). We can construct RP 2 by taking the upper hemisphere of S 2 (homeomorphic to
the disk D2 ) and identifying the opposite (antipodal) points on its boundary.
b
w v
σ
a c a
τ
v w
b
 HOMOLOGY THEORY 7

Remark 1.17. For every 0 ≤ i ≤ n, we defined the coface map δi : [n − 1] → [n] to be an

increasing map that misses i ∈ [n] in the image:
(
k k < i,
δi (k) =
k + 1 k ≥ i.
The corresponding map ∂i = δi∗ : Kn → Kn−1 is called the face map (or the face operator).
Every increasing map f : [m] → [n] can be written as a composition of maps δi . Therefore the
map f ∗ : Kn → Km can be written as a composition of face maps ∂i . These maps can not be
just arbitrary as they should satisfy equation (2). This equation is necessary and sufficient for
the existence of a ∆-set structure.
Lemma 1.18. We have
δj δi = δi δj−1 , i<j (1)
Therefore
∂i ∂j = ∂j−1 ∂i , i < j. (2)
Proof. To prove that first equality, we note that both compositions are increasing and miss i, j.
Therefore they are equal. The second equality follows from the first:
∂i ∂j = δi∗ δj∗ = (δj δi )∗ = (δi δj−1 )∗ = δj−1
∗
δi∗ = ∂j−1 ∂i .

Example 1.19. Given a poset S (see Definition A.1), let ∆(S) consist of all finite non-empty
chains (ordered subsets) I ⊂ S, or equivalently, of increasing maps σ: [m] → S. This is a
∆-set, called the ordered complex or the nerve of S (cf. Example 1.29). The set ∆(S)m of
m-simplices consists of increasing maps [m] → S. The set of vertices is ∆(S)0 = S. Note that
for S = [n] we obtain ∆(S) = ∆[n], the ∆-set corresponding to the triangulation of ∆n , see
Example 1.11.
The following triangulation of a square
1 3
σ
τ
0 2
produces a ∆-set ∆(S) for the poset
S = {0, 1, 2, 3 | 0 < 1 < 3, 0 < 2 < 3} .
Its 2-simplices are σ = [0 < 1 < 3] and τ = [0 < 2 < 3].
Remark 1.20.* Let ∆+ be the category with objects [n] for n ≥ 0 and (strictly) increasing
maps between them. A ∆-set K can be interpreted as a functor K: ∆op
+ → Set, where [n] 7→ Kn
∗
and f : [m] → [n] is mapped to f : Kn → Km .
Remark 1.21.* Let P+ be the category of posets with (strictly) increasing maps between them.
Then ∆+ ⊂ P+ is a subcategory and every poset S induces a functor P+ (−, S): ∆op + → Set,
which is exactly the ∆-set ∆(S) described above.
8 HOMOLOGY THEORY

1.3. Geometric realization. So far we have seen that a triangulation of a space produces a
∆-set structure. But we can also start with a ∆-set K and canonically construct a triangulated
space |K|, called a geometric realization of K. Consider the space
G G
K̄ = Kn × ∆n = ∆σ , ∆σ = {σ} × ∆n , σ ∈ Kn ,
n≥0 σ∈K

with the equivalence relation generated by (see Definition A.5)

Km × ∆m 3 (σ, s) ∼ (τ, t) ∈ Kn × ∆n ⇐⇒ σ = f ∗ (τ ), t = f∗ (s), (3)
f
for some increasing f : [m] → [n]. We will write (σ, s) →
− (τ, t) for the above situation and call it
a simple equivalence. Define the geometric realization
|K| = K̄/∼
and equip it with the quotient topology (see Section A.2) with respect to the projection map
π: K̄ → |K|. This means that U ⊂ |K| is open ⇐⇒ π −1 (U ) is open.

Remark 1.22. For every σ ∈ Kn , we have a map πσ : ∆n ' ∆σ → K̄ → |K| (which is not
◦ ◦
◦
necessarily injective) and its restriction πσ : ∆n ' ∆σ → |K| which is injective as we will see.

Theorem 1.23. The following map is a bijection:

G ◦
π: K̄ ◦ =
◦
(Kn × ∆n ) → |K| .
n≥0

Proof. For any point (τ, t) ∈ Kn × ∆n , let m ≤ n be the minimal dimension of a face of ∆n
containing t. Then there exists an increasing map f : [m] → [n] such that t = f∗ (s) for some
◦
s ∈ ∆m . By the assumption on m, the point s ∈ ∆m is not contained in any facet, hence s ∈ ∆m .
◦
◦
Taking σ = f ∗ (τ ) ∈ Km , we obtain (τ, t) ∼ (σ, s) ∈ Km × ∆m , hence π is surjective.
◦
Let us show that π is injective. Consider two points (σ, s), (τ, t) in K̄ ◦ and assume that they
are equivalent. Then there is a chain of simple equivalences
f =f0 g=f1 f2
(σ, s) = (σ0 , s0 ) −−−→ (σ1 , s1 ) ←−−− (σ2 , s2 ) −
→ . . . ← (τ, t),
with σi ∈ Kmi . We will show that one can make this chain shorter. We have increasing maps
◦
f : [m0 ] → [m1 ], g: [m2 ] → [m1 ] with s1 = f∗ (s0 ) = g∗ (s2 ). As s = s0 ∈ ∆m0 , we have
f∗ (s0 ) ∈ g∗ (∆m2 ) ⇐⇒ f∗ (∆m0 ) ⊂ g∗ (∆m2 ).
We conclude from f∗ (s0 ) = g∗ (s2 ) that f∗ (∆m0 ) ⊂ g∗ (∆m2 ), hence Im f ⊂ Im g, hence there
exists an increasing map h: [m0 ] → [m2 ] with f = gh. Then g∗ (s2 ) = f∗ (s0 ) = g∗ h∗ (s0 ), hence
h
s2 = h∗ (s0 ) and we obtain (σ0 , s0 ) →
− (σ2 , s2 ). Taking its composition with f2 we get a chain
f2 g
(σ, s) −−→ (σ3 , s3 ) ← . . . ← (τ, t).
f
Finally we will get an equivalence (σ, s) → − (τ, t) with t = f∗ (s). As t is internal, we conclude
that f is a bijection, hence (σ, s) = (τ, t). This proves injectivity. 

Corollary 1.24. If K is a ∆-set, then |K| is canonically triangulated.

π
Proof. For every σ ∈ Kn , there is a canonical map πσ : {σ} × ∆n → K̄ →
− |K|. By the previous
theorem these maps satisfy the first axiom of a triangulation. Other axioms follow from the
construction of |K|. 
 HOMOLOGY THEORY 9

Remark 1.25. A map f : K → L between ∆-sets is called a simplicial morphism if f (Kn ) ⊂

Ln and if for every increasing φ: [m] → [n] we have φ∗ f (σ) = f φ∗ (σ), for all σ ∈ Kn . Such
map induces a continuous map between geometric realizations |f | : |K| → |L| which sends
{σ} × ∆n ⊂ |K| to {f (σ)} × ∆n ⊂ |L|, for all σ ∈ Kn .
Theorem 1.26. Let X be a space with a triangulation K. Then the map
φ: |K| → X, Kn × ∆n 3 (σ, t) 7→ σ(t) ∈ X,
is a homeomorphism.
Proof. Let us show first that φ respects equivalence classes, that is, well-defined.
f∗ iτ π
∆m ∆n K̄ |K|
τ
φ
σ

X
f
Consider a simple equivalence (σ, s) → − (τ, t) with f : [m] → [n] (3). Then σ = f ∗ (τ ) = τ f∗ . This
implies that σ(s) = τ f∗ (s) = τ (t), hence the images of equivalent elements coincide.
◦
We proved that the map π: K̄ ◦ → |K| is bijective. On the other hand the composition
◦
π φ
K̄ ◦ →
− |K| →
− X
is bijective by the first axiom of a triangulation, hence φ: |K| → X is also bijective.
The map K̄ → X, (σ, t) 7→ σ(t) is continuous, hence φ: |K| → X is also continuous, by the
definition of the quotient topology. On the other hand, if U ⊂ |K| is open, then A = φ(U ) ⊂ X
satisfies
σ −1 (A) = π −1 (U ) ∩ ({σ} × ∆n ) ∀σ ∈ Kn .
This set is open as π is continuous. By the third axiom of a triangulation, this implies that A is
open, hence φ is a homeomorphism. 
10 HOMOLOGY THEORY

1.4. Simplicial complexes. In this section we will introduce an important particular class of
∆-sets, called simplicial complexes.
1.4.1. Simplicial complexes.
Definition 1.27. An (abstract) simplicial complex (V, K) consists of a vertex set V and a
collection K of non-empty finite subsets of V (called simplices) such that
(1) {v} ∈ K for every v ∈ V .
(2) If σ ∈ K and ∅ 6= τ ⊂ σ, then τ ∈ K. The set τ is called a face of σ.
We say that a simplex σ ∈ K has dimension n if σ consists of n + 1 elements. Let Kn denote
the set of all n-dimensional simplices of K. We will identify K0 and V .
Example 1.28. Let ∆[n] be the set of non-empty subsets of [n] = {0, . . . , n}. It is a simplicial
complex with the set of vertices [n].
Example 1.29 (Order complex). More generally, for any poset S, let ∆(S) be the set of all
non-empty finite chains (totally ordered sets) in S (cf. Example 1.19). Then ∆(S) is a simplicial
complex, called the order complex (or the nerve) of S. The set of vertices of ∆(S) is S.
In particular, for the poset [n] = {0, . . . , n} with the usual order, the order complex ∆([n])
coincides with ∆[n].
Definition 1.30. We say that a simplicial complex K is ordered if V = K0 is equipped with
a partial order such that every simplex σ ⊂ V is a chain (totally ordered set).
Lemma 1.31. An ordered simplicial complex has a canonical structure of a ∆-set.
Proof. Every σ ∈ Kn can be written in the form σ = (v0 , . . . , vn ), where vi ∈ V and v0 < · · · < vn .
Therefore σ can be identified with an increasing map σ: [n] → V , i 7→ vi . For any increasing
map f : [m] → [n], the composition σf : [m] → V corresponds to a simplex in Km , hence we
obtain a map f ∗ : Kn → Km , σ 7→ σf . This means that K has a structure of a ∆-set. 
Remark 1.32. Let K be a ∆-set (or a simplicial complex) and let V = K0 be the set of its
vertices. Denote by 2V be the set of all subsets of V , called the power set of V . For every
0 ≤ i ≤ n, consider the map v̄i : [0] → [n], 0 7→ i, and the corresponding map vi : Kn → K0 = V
(cf. Remark 1.13). These maps induce
v: K → 2V , Kn 3 σ 7→ {vi (σ) | 0 ≤ i ≤ n} ⊂ V,
which associates with every σ ∈ K its set of vertices v(σ). If K is a simplicial complex, then
every simplex of K is uniquely determined by its set of vertices, hence v: K → 2V is injective.
Example 1.33. Consider a triangulation of S 1 consisting of two vertices and two segments
between them. Then v: K → 2V is not injective.
Lemma 1.34. Let K be a ∆-set such that v: K → 2V is injective (every simplex is uniquely
determined by its set of vertices). Then K has a structure of a simplicial complex.
Proof. As v: K → 2V is injective, we can consider K as a subset of 2V . For every v ∈ V = K0 ,
the one-point set {v} is in K, hence the first axiom of a simplicial complex is satisfied.
For every σ ∈ Kn , the subset v(σ) ⊂ V has exactly n + 1 elements. Indeed, we know that
v(σ) has ≤ n + 1 elements. If it has < n + 1 elements, then vi (σ) = vj (σ) for some i 6= j and
then τ = δi∗ (σ) (where δi : [n − 1] → [n] is the i-th facet) would have the same set of vertices
as σ, contradicting to our assumption. Given  σ = {v0 , . . . , vn } ∈ Kn∗ and τ ⊂ σ, there exists an
increasing map f : [m] → [n] such that τ = vf (0) , . . . , vf (m) . But f (σ) ∈ Km has the same set
of vertices, hence τ ∈ Km and the second axiom of a simplicial complex is satisfied. 
 HOMOLOGY THEORY 11

Example 1.35. Consider a triangulation of S 1 which corresponds to a simplicial complex.

Such triangulation should have at least 3 vertices and 3 edges. On the other hand, if we allow
arbitrary triangulations of S 1 , then we can consider a triangulation with just one vertex and
one edge.
Remark 1.36. Let K be a ∆-set with K0 = {v0 , v1 , v2 } and K1 = {[v0 , v1 ], [v1 , v2 ], [v2 , v0 ]}.
This ∆-set structure does not come from any order on the corresponding simplicial complex
as this would require v0 < v1 < v2 < v0 . If we consider instead K1 = {[v0 , v1 ], [v1 , v2 ], [v0 , v2 ]},
then it corresponds to a simplicial complex with an order v0 < v1 < v2 .
1.4.2. Geometric realization. We assume that we have an ordered simplicial complex K, hence
also a ∆-set structure on it. We will write a simplex in the form (v0 , . . . , vn ), where v0 < · · · < vn .
Definition 1.37. Given a simplicial complex K, define the simplicial polyhedron of K
 X 
V
|K|p = t ∈ [0, 1] tv = 1, supp(t) ∈ K ⊂ [0, 1]V ,
v∈V

where supp(t) = {v ∈ V | tv 6= 0}. We equip |K|p ⊂ [0, 1]V with the subspace topology. The
numbers tv are called barycentric coordinates of t ∈ |K|p .
Example 1.38. Let K = ∆[n] be the set of all non-empty subsets of [n]. Then
n X o
|K|p = t ∈ [0, 1]n+1 ti = 1 = ∆n .
Example 1.39. Let K = ∂∆[n] be the set of all non-empty subsets of [n] except the set [n]
itself. Then n X o
|K|p = t ∈ [0, 1]n+1 ti = 1, ∃ti = 0 = ∂∆n ' S n−1 .
Theorem 1.40. Let K be an ordered simplicial complex equipped with the canonical ∆-set
structure. Then there is a continuous bijection φ: |K| → |K|p , where for any simplex σ =
(v0 , . . . , vn ) ∈ Kn we consider
n o
φ
{σ} × ∆n → − ∆(σ) = t ∈ |K|p supp(t) ⊂ σ ⊂ |K|p ,
X
(t0 , . . . , tn ) 7→ ti δvi ∈ [0, 1]V ,
and δv ∈ RV is the delta-function corresponding to v ∈ V (a standard basis vector). If K is
finite, then φ: |K| → |K|p is a homeomorphism.
Proof. The map G
φ̄: K̄ = Kn × ∆n → |K|p
n≥0
described above is continuous and respects equivalence classes. nTherefore it induces a continuous
◦ ◦
o
map φ: |K| → |K|p . The interior ∆n is mapped to ∆(σ) = t ∈ |K|p supp(t) = σ and we
F ◦
have |K|p = σ∈K ∆(σ).
◦ ◦
π
Kn × ∆n
F
n≥0 |K|

φ̄ φ
F ◦
|K|p = σ∈K ∆(σ)
◦
◦ F
But we proved earlier that π: n≥0 Kn × ∆n → |K| is a bijection, hence φ is a bijection.
If K is finite, then |K| is compact. As |K|p is Hausdorff, we conclude that the bijective
continuous map φ: |K| → |K|p is a homeomorphism. 
12 HOMOLOGY THEORY

1.4.3. Geometric simplicial complexes.

Definition 1.41.
(1) The points v0 , . . . , vn ∈ RN are called affinely independent if the vectors
v1 − v0 , . . . , vn − v0
are linearly independent. Equivalently, the map
X
σ = [v0 , . . . , vn ]: ∆n → RN , t 7→ ti vi
is injective.
(2) The map σ: ∆n → RN as above is called an n-simplex and the points vi = σ(ei ) are
called its vertices. We identify σ with its image (the convex hull)
nX X o
∆(σ) = Conv(v0 , . . . , vn ) = ti vi ti = 1, ti ≥ 0 ∀i ⊂ RN
equipped with an ordering of vertices.
(3) Define a face of a simplex σ = [v0 , . . . , vn ] to be a simplex [vi0 , . . . , vim ] corresponding
to a non-empty subset of vertices of σ.
Remark 1.42. The standard n-simplex is given by ∆n = [e0 , . . . , en ] ⊂ Rn+1 , where (e0 , . . . , en )
is the standard basis of Rn+1 .
Definition 1.43. A geometric simplicial complex K in RN is a collection of simplices
in RN such that
(1) If σ ∈ K, then every face of σ is in K.
(2) The intersection of two simplices in K is a face of both of them.
Define the underlying space |K|u = σ∈K ∆(σ) ⊂ RN .
S

Remark 1.44. Let V ⊂ RN be the set of 0-simplices of K and let Kn be the set of n-simplices
of K for n ≥ 0. Every simplex σ = [v0 , . . . , vn ] in K can be identified with an (ordered) subset of
V (the set of vertices of σ). By the first axiom above K has a structure of an abstract simplicial
◦
complex with the set of vertices V . Let ∆(σ) denote the interior of ∆(σ) ⊂ RN . The second
F ◦
axiom above implies that |K|u = σ∈K ∆(σ), a disjoint union of open simplices.
Theorem 1.45. Let K be a finite geometric simplicial complex in RN . Then the map
X
f : |K|p → |K|u , t 7→ tv v
v∈V

is a homeomorphism, where V ⊂ RN is the set of vertices of K.

Proof. Under this map, for any σ = [v0 , . . . , vn ] ∈ Kn , the simplex
n o
∆(σ) = t ∈ |K|p supp t ⊂ σ ⊂ |K|p
F ◦ F ◦
is mapped bijectively to ∆(σ) ⊂ |K|u . We have seen that |K|p = σ ∆(σ) and |K|u = σ ∆(σ),
hence f is a bijection. It is easy to see that f is continuous. The fact that |K|p is compact and
|K|u is Hausdorff, implies that f is a homeomorphism. 
 HOMOLOGY THEORY 13

1.5. Product triangulation.

Theorem 1.46 (Triangulation of the product). The product ∆m × ∆n has a triangulation
parametrized by the order complex ∆(S) (see Example 1.29), where S = [m] × [n] is equipped
with the partial order (i, j) ≤ (i0 , j 0 ) if i ≤ i0 and j ≤ j 0 . Namely, there is a homeomorphism
X X
|∆ (S)| −∼→ ∆m × ∆n , tij eij 7→ tij (ei , ej ) ∈ Rm+n+2 .
Proof. We will consider only the case S = [n] × [1] and the corresponding map
f : |∆(S)| → ∆n × [0, 1].
We have S = {v0 , . . . , vn , w0 , . . . , wn }, where vi = (i, 0), wi = (i, 1) and vi ≤ vj , vi ≤ wj ,
wi ≤ wj for i ≤ j. The maximal simplices have dimension n + 1 and are of the form
σi = (v0 , . . . , vi , wi , . . . , wn ) for 0 ≤ i ≤ n. A point
(t0 , . . . , ti−1 , t0i , t00i , ti+1 , . . . , tn ) ∈ ∆(σi ) ' ∆n+1
is mapped to
(t0 , . . . , ti−1 , ti = t0i + t00i , ti+1 , . . . , tn ; s) ∈ ∆n × [0, 1], s = t00i + ti+1 + · · · + tn .
Let us introduce new coordinates on ∆n by sending t ∈ ∆n to
x = (x1 , . . . , xn ) ∈ Rn , xj = t0 + · · · + tj−1 , 1 ≤ j ≤ n.
In these coordinates ∆n is given by 0 ≤ x1 ≤ · · · ≤ xn ≤ 1. Note that 1 − s = t0 + · · · + ti−1 + t0i ,
hence the above point in ∆n × [0, 1] satisfies
0 ≤ x1 ≤ · · · ≤ xi ≤ 1 − s ≤ xi+1 ≤ · · · ≤ xn .
For any point in ∆n × [0, 1], we can find an appropriate i such that this point is contained in the
image of ∆(σi ). This implies that f is surjective. Injectivity follows from the same analysis. 
Example 1.47. For n = 1, consider the triangulation of ∆1 × ∆1 .
w0 w1

v0 v1
The 1-simplex [v0 , w1 ] divides the cylinder into 2-triangles, which are maximal simplices given
by [v0 , w0 , w1 ], [v0 , v1 , w1 ].
Example 1.48. For n = 2, consider the triangulation of the cylinder ∆2 × ∆1 .
w2

w0 w1

v2

v0 v1
The 2-simplices [v0 , w1 , w2 ], [v0 , v1 , w2 ] divide the cylinder into 3-tetrahedrons, which are maximal
simplices given by [v0 , w0 , w1 , w2 ], [v0 , v1 , w1 , w2 ], [v0 , v1 , v2 , w2 ].
14 HOMOLOGY THEORY

1.6. Barycentric subdivision. Let K be a simplicial complex with the set of vertices V . Then
K ⊂ 2V is a poset (ordered by inclusion), and we can construct its order complex ∆(K) (see
Example 1.29), consisting of all chains in K. It is called the barycentric subdivision of K.
Example 1.49. For K = ∆[2] (subsets of [2]), we have
K0 : 0, 1, 2, K1 : a = [0, 1], b = [1, 2], c = [0, 2], K2 : σ = [0, 1, 2].
The simplices of ∆(K) (which are chains in K) can be identified with the simplices of the
triangulation
2

c5 b4
σ6

0 1
a3
if we identify simplices of K with their barycenters (centers of gravity), namely σ with 6, a with
3, b with 4 and c with 5. For example
(1) [0, a, σ] ∈ ∆(K) is identified with [0, 3, 6].
(2) [1, a, σ] ∈ ∆(K) is identified with [1, 3, 6].
(3) [0, σ] ∈ ∆(K) is identified with [0, 6].
One can see on this example that the geometric realization of K can be identified with the
geometric realization of ∆(K) (both are given by the large triangle).
Theorem 1.50. There is a canonical bijection between geometric realizations of ∆(K) and K.
Proof. If L = ∆(K), then L0 = K. For any simplex σ ∈ Kn , define its barycenter (or center
of gravity)
1 X
bσ = δv ∈ ∆(σ) ⊂ |K|p ,
n + 1 v∈σ
where δv : V → [0, 1] is the delta-function. Define the map
X
f : |L|p → |K|p , t = (tσ )σ∈K 7→ tσ bσ .
σ∈K
◦
Note that if supp(t) = (σ0 < · · · < σn ) ∈ L, then supp(f (t)) = σn , hence f (t) ∈ ∆(σn ) ⊂ |K|p .
◦
One can show that there is a bijection between t ∈ |L|p with max supp(t) = σn and ∆(σn ) ⊂ |K|p .
Therefore f : |L|p → |K|p is a bijection (and a homeomorphism if K is finite). 
 HOMOLOGY THEORY 15

1.7. Simplicial homology. Consider a ∆-set K (or a triangulated space (X, K)). For every
0 ≤ i ≤ n, consider the coface map δi : [n − 1] → [n] which is an increasing map that misses
i ∈ [n] in the image. Consider the corresponding face map (see Remark 1.17)
∂i = δi∗ : Kn → Kn−1 , 0 ≤ i ≤ n.
By Lemma 1.18 we have
∂i ∂j = ∂j−1 ∂i , i < j.
Definition 1.51. Let K be a ∆-set.
(1) Define the group of n-chains Cn = Cn (K) to be the free abelian group with a basis
consisting of n-simplices of K. An element of Cn (K), called an n-chain, is a finite sum
P
σ∈Kn nσ σ, where nσ ∈ Z.
(2) Define the boundary operator (also called the differential)
n
X
d = dn : Cn → Cn−1 , dn (σ) = (−1)i ∂i (σ).
i=0

(3) The sequence of homomorphisms between abelian groups

dn+1 n d 1 0 d d
· · · → Cn+1 −−−→ Cn −→ Cn−1 → · · · → C1 −
→ C0 −
→ 0 → ...
satisfies the equation dn dn+1 = 0 for all n (see the next Lemma) and is called the chain
complex of K, denoted as C• (K), C• or (C• , d• ).
(4) Given a triangulated space (X, K), define C•K (X) = C• (K).
Example 1.52. For a simplex σ = [v0 , . . . , vn ], we have
n
X
d(σ) = (−1)i [v0 , . . . , v̂i , . . . , vn ].
i=0

For example d[v0 , v1 ] = v1 − v0 and d[v0 , v1 , v2 ] = [v1 , v2 ] − [v0 , v2 ] + [v0 , v1 ].

n dn+1 d
Lemma 1.53. The composition Cn+1 −−−→ Cn −→ Cn−1 is zero (also written as d2 = 0).
Proof. We have
X X
dn dn+1 (σ) = (−1)i+j ∂i ∂j (σ) + (−1)i+j ∂i ∂j (σ)
i<j≤n+1 n≥i≥j
X X
= (−1)i+j+1 ∂j ∂i (σ) + (−1)i+j ∂i ∂j (σ) = 0.
i≤j≤n n≥i≥j


Note that the equation dn dn+1 = 0 implies
Im dn+1 ⊂ Ker dn .
Definition 1.54. Let K be a ∆-set and C• = C• (K) be the corresponding chain complex.
(1) Define the n-th homology group of K (or of the complex C• ) to be
Hn (K) = Hn (C) = Ker dn / Im dn+1 .
(2) The elements of Zn (C) = Ker dn ⊂ Cn are called n-cycles and the elements of Bn (C) =
Im dn+1 ⊂ Cn are called n-boundaries.
(3) If (X, K) is a triangulated space, define the n-th simplicial homology group of X to
be HnK (X) = Hn (K) which is the n-homology group of C•K (X).
16 HOMOLOGY THEORY

Remark 1.55. More generally, for any ring R, we define the chain complex C• (K, R) with
Cn (K, R) = R ⊗Z Cn (K) (the free R-module with the basis Kn ) and simplicial homology groups
(or R-modules) Hn (K, R) = Hn (C• (K, R)). Note that Hn (K, R) 6' R ⊗Z Hn (K) in general,
although this is true if R is a field of characteristic zero (or does not have torsion over Z).
Remark 1.56. We will show later that the simplicial homology groups HnK (X) are independent
of the triangulation K. But this statement will require a lot of work.
Example 1.57. Consider a sequence of 1-simplices (called a path)
a1 a2 an
v0 v1 v2 vn−1 vn

and a 1-chain c = a1 + · · · + an ∈ C1 corresponding to this path. Then

Xn
d(c) = (vi − vi−1 ) = vn − v0 .
i=1
This means that c is a 1-cycle if and only if vn = v0 , that is, the above path is an actual cycle.
Example 1.58. The signs are used to take orientation into account. Consider a triangulation
of a square
1 3
σ
τ
0 2
with σ = [0, 1, 3] (clockwise) and τ = [0, 2, 3] (anti-clockwise). Then
d(σ) = [1, 3] − [0, 3] + [0, 1] = [0, 1] + [1, 3] − [0, 3]
which corresponds to the boundary of σ going clockwise along the cyclic path 0 → 1 → 3 → 0.
This is why d(σ) is called a 1-boundary. On the other hand
d(σ − τ ) = [0, 1] + [1, 3] − [2, 3] − [0, 2]
corresponds to the boundary of the square going clockwise along the cyclic path 0 → 1 → 3 →
2 → 0.
Example 1.59. Consider the ∆-set K = ∂∆[2] which a triangulation of S 1
2

c b

0 a 1

Then
d(a) = [1] − [0], d(b) = [2] − [1], d(c) = [2] − [0]
Therefore d(a + b − c) = 0, hence a + b − c is a 1-cycle. It corresponds to the cyclic path
0 → 1 → 2 → 0, hence the name. Let us compute homologies. We have C0 = Z3 , C1 = Z3 with
d = d1 : C1 → C0 described above. Then Ker d1 = (a+b−c)Z, hence H1 (C) = Ker d1 / Im d2 ' Z.
We have n0 [0] + n1 [1] + n2 [2] ∈ Im d1 ⇐⇒ n0 + n1 + n2 = 0. Therefore H0 (C) = Z3 / Im d1 ' Z,
where n0 [0] + n1 [1] + n2 [2] 7→ n0 + n1 + n2 . We conclude that
H0K (S 1 ) ' Z, H1K (S 1 ) ' Z.
 HOMOLOGY THEORY 17

Example 1.60. Consider a triangulation of S 1 consisting of a vertex v and a segment e = [v, v].
Then C0 = Zv, C1 = Ze and
d: C1 → C0 , de = v − v = 0
with all other differentials equal zero. This implies that
H1 (C) = Ker d/0 = Z, H0 (C) = Z/ Im d = Z
and all other homology groups are zero. We conclude that H0K (S 1 ) = H1K (S 1 ) = Z.
Let us consider a different triangulation of S 1 , with two vertices v0 , v1 and two segments
e = [v0 , v1 ], e0 = [v1 , v0 ] between them. Then C0 = Z2 , C2 = Z2 and d(e) = v0 −v1 , d(e0 ) = v1 −v0
can be written as a matrix
 
1 −1
d: C1 → C0 , d= .
−1 1
Then Ker d = Z(1, 1), Im d = Z(1, −1) and
H1 (C) = Ker d/0 ' Z, H0 (C) = Z2 / Im d ' Z.
We see that the homology groups are the same as before, although our chain complex is different.
We will see later that this is a general phenomenon. Therefore it is enough to find the simplest
possible triangulation of a space in order to compute its homologies.

Example 1.61. Hn (∆1 ) = Z for n = 0 and zero otherwise. Hn (∆2 ) = Z for n = 0 and zero
otherwise. More generally, one can show that Hn (∆k ) = Z for n = 0 and zero otherwise. The
reason is that ∆n is homotopic to a point.

Example 1.62. We can triangulate S 2 by taking two copies of ∆2 and identifying them along
the boundary. Similarly, S n is obtained by taking two copies σ, τ of ∆n and identifying them
along the boundary. We can show that Hn (S n ) ' Z. Let us compute homology groups for S 2 .

Example 1.63. Consider the following triangulation of the torus T

b
v v
σ
a c a
τ
v v
b
with vertices of σ going along a, b and vertices of τ going along b, a. Then d2 (σ) = b − c + a,
d2 (τ ) = a − c + b, d1 (a) = d1 (b) = d1 (c) = 0
d d
Z2 −
→2
Z3 −
→1
Z.
Therefore H0 (T ) ' Z, H1 (T ) = Z3 / Im d2 ' Z2 , H2 (T ) = Ker d2 ' Z.

Example 1.64. Define an n-dimensional projective space

RP n = (Rn+1 \ {0})/R∗ ' S n /Z2 ,
where R∗ acts on Rn+1 by multiplication and Z2 acts on S n by the antipodal map x 7→ −x.
For example, we can describe RP 2 as the quotient of the disc D2 with antipodal points on the
boundary identified. We construct a triangulation of RP 2 by identifying arrows in the following
picture, where antipodal points on the boundary are identified.
18 HOMOLOGY THEORY

b
2 1
σ
a c a
τ
0 3
b
Let v = 0 ∼ 1, w = 2 ∼ 3. We consider σ = [012] and τ = [013] so that
d2 (σ) = b − a + c, d2 (τ ) = a − b + c, d1 (a) = d1 (b) = w − v, d1 (c) = 0.
The corresponding chain complex is
d2 d1
C2 C1 C0
 
−1 1  
−1 −1 0
d2 =  1 −1 , d1 =

1 1 0
d2 d1 1 1
Z2 Z3 Z2
We have        
1 0 1 0
Im(d2 ) = −1 Z + 0 Z,
    Ker d1 = −1 Z + 0 Z.
  
1 2 1 1
Therefore H1 (K) = Ker d1 / Im d2 ' Z/2Z = Z2 . The map d2 is injective, hence H2 (K) =
Ker d2 = 0. Finally, H0 (K) = Z2 / Im d1 ' Z. Note that over rational numbers we have
H1 (K, Q) ' Q/2Q = 0.
Example 1.65. Let us assume that we have a triangulated surface (X, K). Consider its chain
complex Cn = CnK (X, Q) with rational coefficients
d
2 1 d
· · · → 0 → C2 −
→ C1 −
→ C0 → 0 → . . .
and the corresponding homology groups H0 , H1 , H2 . It’s a simple exercise in linear algebra to
show that
e(X) = #K0 − #K1 + #K2 = dim C0 − dim C1 + dim C2 = dim H0 − dim H1 + dim H2 .
As we will see later, the vector spaces Hn are independent of a triangulation. In particular, the
above number e(X), called the Euler characteristic of X, is independent of a triangulation.
Compare this with Example 1.1, where we considered a triangulation of S 2 . Of course all the
arguments work also for higher dimensional spaces.
 HOMOLOGY THEORY 19

2. Homological algebra
Let R be a ring. Recall that an R-module M is an abelian group equipped with a map

R × M → M, (a, m) 7→ am,

such that, for all a, b ∈ R and x, y ∈ M ,

(1) a(bx) = (ab)x.
(2) 1R x = x.
(3) a(x + y) = ax + ay.
(4) (a + b)x = ax + ay.
If R = Z, then an R-module is just an abelian group. If R is a field (for example R or C), then
an R-module is a vector space over R. A map f : M → N between two R-modules is called a
homomorphism (or an R-linear map) if for all a ∈ R and x, y ∈ M
(1) f (ax) = af (x).
(2) f (x + y) = f (x) + f (y).

Definition 2.1.
(1) A (chain) complex C• (or (C• , d• )) is a sequence of R-modules and homomorphisms
dn+1 dn
· · · → Cn+1 −−−→ Cn −→ Cn−1 → . . .

such that dn dn+1 = 0 for all n ∈ Z (also written as d2 = 0). The map dn is called a
differential (or a boundary operator).
(2) The elements of Cn are called n-chains. The elements of

Zn (C) = Ker dn , Bn (C) = Im dn+1

are called n-cycles and n-boundaries respectively. Note that d2 = 0 implies Bn (C) ⊂
Zn (C). The module

Hn (C) = Zn (C)/Bn (C) = Ker dn / Im dn+1

is called the n-th homology group (or homology module). Its elements are called the
n-th homology classes.

Definition 2.2.
f g
(1) A sequence of homomorphisms M1 →− M2 → − M3 is called exact at M2 if Im f = Ker g.
(2) A complex C• is called exact (or acyclic) if it is exact at every Cn . Equivalently,
Hn (C) = 0 for all n.

Lemma 2.3.
f
(1) A sequence 0 → M1 →
− M2 is exact at M1 ⇐⇒ f is a monomorphism.
f
(2) A sequence M1 →
− M2 → 0 is exact at M2 ⇐⇒ f is an epimorphism.
f g
(3) A sequence 0 → M1 →− M2 → − M3 → 0 with gf = 0 is exact (at all components) ⇐⇒ f
is a monomorphism, g is an epimorphism and Im f = Ker g. Such sequence is called a
short exact sequence.
(4) If M 0 ⊂ M is a submodule, then 0 → M 0 → M → M/M 0 → 0 is a short exact sequence.

Definition 2.4. A chain map f : A• → B• between complexes A• , B• is a collection of homo-

morphisms fn : An → Bn for n ∈ Z such that dn fn = fn−1 dn ∀n. This means that we have a
20 HOMOLOGY THEORY

commutative diagram
dn+1 dn
... An+1 An An−1 ...
fn+1 fn fn−1
dn+1 dn
... Bn+1 Bn Bn−1 ...
We will always assume that a map between complexes is a chain map unless otherwise stated.

Lemma 2.5. A chain map f : A• → B• induces homomorphisms f∗ = Hn f : Hn (A) → Hn (B)

between homology modules.

Proof. Given x ∈ Zn (A) = Ker(d: An → An−1 ), we have df (x) = f d(x) = 0, hence f (x) ∈
Zn (B) = Ker(d: Bn → Bn−1 ). We define then f∗ (x) = [f (x)] ∈ Hn (B) = Zn (B)/Bn (B). We
need to show that this definition is independent of the choice of x in its homology class, that is,
f∗ (x + dy) = f∗ (x) for y ∈ An+1 . But f∗ (dy) = [f d(y)] = [df (y)] = 0 as df (y) ∈ Bn (B). 

Definition 2.6.
(1) Let f, g: A• → B• be two chain maps. A chain homotopy s between f and g (or from
f to g, written as s: f ∼ g) is a sequence of homomorphisms sn : An → Bn+1 such that
dn+1 sn + sn−1 dn = gn − fn , ∀n ∈ Z
dn+1 dn
... An+1 An An−1 ...
sn fn sn−1
fn+1 fn−1
dn+1 dn
... Bn+1 Bn Bn−1 ...
Chain maps f, g are called homotopic (written as f ∼ g) if such s exists.
(2) A chain map f : A• → B• is called null-homotopic if f ∼ 0.
(3) A chain map f : A• → B• is called a homotopy equivalence if there exists a chain
map g: B• → A• such that f g ∼ 1B and gf ∼ 1A . In this case g is called a homotopy
inverse of f and A• , B• are called homotopy equivalent complexes.
(4) A chain complex A• is called contractible (or null-homotopic) if 1A ∼ 0.

Remark 2.7. A chain complex A• is null-homotopic ⇐⇒ A• is homotopy equivalent to the

zero complex. Indeed, consider the unique chain maps f = 0: A• → 0 and g = 0: 0 → A• . If
1A ∼ 0, then gf = 0 ∼ 1A and f g = 0 = 10 . Therefore f is a homotopy equivalence. Conversely,
if f : A• → 0 has a homotopy inverse g: 0 → A• , then 0 = gf ∼ 1A .

Lemma 2.8.
(1) The homotopy relation is an equivalence relation.
(2) The homotopy relation is compatible with composition: if f, g: A• → B• are homotopic
and f 0 , g 0 : B• → C• are homotopic, then f 0 f ∼ g 0 g.

Proof. (1) It is clear that f ∼ f and if s: f ∼ g, then −s is a homotopy between g and f :

f − g = d(−s) + (−s)d. We need to show transitivity: if s: f ∼ g and t: g ∼ h, then f ∼ h. We
have h − f = (g − f ) + (h − g) = d(s + t) + (s + t)d, hence s + t is a homotopy between f and h.
(2) Let us show that f 0 f ∼ f 0 g. Similarly one can show that f 0 g ∼ g 0 g and then by transitivity
f 0 f ∼ g 0 g. If s: f ∼ g, then
f 0 g − f 0 f = f 0 (ds + sd) = d(f 0 s) + (f 0 s)d,
where we used the property f 0 d = df 0 . This implies that f 0 s is a homotopy between f 0 f, f 0 g. 
 HOMOLOGY THEORY 21

Lemma 2.9. If f, g: A• → B• are homotopic, then they induce the same morphisms of homolo-
gies: f∗ = g∗ : Hn (A) → Hn (B).
Proof. Let s: f ∼ g. Given x ∈ Zn (A) = Ker(dn : An → An−1 ), we need to show that f (x), g(x) ∈
Zn (B) coincide modulo Bn (B) = Im(dn+1 : Bn+1 → Bn ). But gx−f x = (g−f )(x) = dsx+sdx =
dsx ∈ Im dn+1 , where we used the fact that dx = 0. 
Corollary 2.10. A homotopy equivalence induces an isomorphism between homology groups.
Proof. Let f : A• → B• have a homotopy inverse g: B• → A• . Then f g ∼ 1B implies f∗ g∗ = 1
and gf ∼ 1A implies g∗ f∗ = 1. This means that f∗ : Hn (A) → Hn (B) is an isomorphism. 
Theorem 2.11 (Snake lemma). Consider a commutative diagram
Ker f Ker g Ker h
∂

i p
X Y Z 0
f g h
j q
0 X0 Y0 Z0

Coker f Coker g Coker h

with exact middle rows. Then there is an exact sequence
∂
Ker f → Ker g → Ker h →
− Coker f → Coker g → Coker h,
where ∂ is defined by j −1 gp−1 .
Proof. We will show exactness only at Ker g and Ker h. First let us show that ∂ is well defined.
For z ∈ Ker h choose y ∈ Y with py = z. Then qgy = hpy = hz = 0 =⇒ gy = jx0 for a unique
x0 ∈ X 0 . Then we define ∂(z) = [x0 ] = x0 + Im f . If ȳ ∈ Y is another element with pȳ = z =⇒
p(ȳ − y) = 0 =⇒ ȳ − y = ix for some x ∈ X =⇒ g ȳ − gy = gix = jf x =⇒ g ȳ = j(x0 + f x)
and we note that x0 and x0 + f x are equivalent modulo Im f .
To prove exactness at Ker g, consider y ∈ Ker g with py = 0. Then y = ix for some x ∈ X
=⇒ jf x = gix = gy = 0 =⇒ f x = 0 =⇒ x ∈ Ker f .
To prove exactness at Ker h, we first note that ∂p = 0. Indeed, if z = py for some y ∈ Ker g,
then by the above construction ∂(z) = [x0 ], where jx0 = gy = 0, hence x0 = 0. This implies
that ∂p(y) = ∂(z) = 0. On the other hand, assume that z ∈ Ker ∂. Using the above notation
we conclude that x0 ∈ Im f =⇒ x0 = f x for some x ∈ X. Then gy = jx0 = jf x = gix =⇒
ȳ = y − ix ∈ Ker g and pȳ = py − pix = py = z. This implies that z ∈ Im(p: Ker g → Ker h). 
Definition 2.12. A sequence of chain maps
f g
0 → A• →
− B• →
− C• → 0
is called a short exact sequence of complexes if
fn gn
0 → An −→ Bn −→ Cn → 0
are short exact sequences of modules for all n.
Theorem 2.13 (Long exact sequence). Consider a short exact sequence of complexes
f g
0 → A• →
− D• →
− C• → 0.
22 HOMOLOGY THEORY

Then there is a long exact sequence of homology groups:

f∗ g∗ ∂ f∗ g∗
· · · → Hn (A) −
→ Hn (D) −
→ Hn (C) →
− Hn−1 (A) −
→ Hn−1 (D) −
→ Hn−1 (C) → . . .
Proof. We will apply the Snake lemma to the following diagram
f g
An /Bn (A) Dn /Bn (D) Cn /Bn (C) 0
d d d
f g
0 Zn−1 (A) Zn−1 (D) Zn−1 (C)
We need to check that the rows are exact. Let us show that the bottom row is exact. It is clear
that Zn−1 (A) → Zn−1 (D) is injective as An−1 → Dn−1 is injective. Let us show exactness at
Zn−1 (D). Assume that x ∈ Zn−1 (D) is such that g(x) = 0. Then there exists y ∈ An−1 such
that f (y) = 0. Then f d(y) = df (y) = 0, hence d(y) = 0 as f is injective. This implies that
y ∈ Zn−1 (A), hence the bottom sequence is exact at Zn−1 (D).
Note that
Ker(d: An /Bn (A) → Zn−1 (A)) ∼ Zn (A)/Bn (A) = Hn (A),
Coker(d: An /Bn (A) → Zn−1 (A)) ∼ Zn−1 (A)/Bn−1 (A) = Hn−1 (A)
and the same applies to other vertical maps. From the Snake lemma we obtain an exact sequence
f∗ g∗ ∂ f∗ g∗
Hn (A) −
→ Hn (D) −
→ Hn (C) →
− Hn−1 (A) −
→ Hn−1 (D) −
→ Hn−1 (C)

Theorem 2.14 (Five-lemma). Consider a commutative diagram with exact rows:
d d d d
A1 A2 A3 A4 A5
f1 f2 f3 f4 f5
d d d d
B1 B2 B3 B4 B5
If f1 , f2 , f4 , f5 are isomorphisms, then f3 is also an isomorphism.
Proof. We will show that f3 is injective using the so-called diagram chasing. Let x ∈ A3
and f3 x = 0 =⇒ 0 = df3 x = f4 dx =⇒ dx = 0 =⇒ x = dy for some y ∈ A2 =⇒
0 = f3 x = f3 dy = df2 y =⇒ f2 y = dz for some z ∈ B1 . Let z = f1 z 0 for some z 0 ∈ A1 . Then
f2 y = df1 z 0 = f2 dz 0 =⇒ y = dz 0 =⇒ x = dy = d2 z 0 = 0. This means that Ker f3 = 0. 
 HOMOLOGY THEORY 23

3. Singular homology
3.1. Definition and basic properties.
Definition 3.1. Let X be a space.
(1) A continuous map σ: ∆n → X is called a singular n-simplex of X.
(2) Let Sn (X) be the set of singular n-simplices of X. For any increasing f : [m] → [n], define
σ ∆f σ
f ∗ : Sn (X) → Sm (X), [∆n →
− X] 7→ [∆m −→ ∆n →
− X].
In particular, define ∂i = δi∗ : Sn (X) → Sn−1 (X), where δi : [n − 1] → [n] is the i-th facet.
We obtain a ∆-set S(X), called a singular ∆-set.
(3) Define the singular chain complex C(X) with the group of n-chains Cn (X) equal
the free abelian group generated by singular n-simplices and differential (also called a
boundary operator),
n
X
d: Cn (X) → Cn−1 (X), d(σ) = (−1)i ∂i (σ).
i=0

(4) For any abelian group R, define the chain complex C(X, R) with the group of n-chains
Cn (X, R) = Cn (X) ⊗Z R.
If R is a ring, then Cn (X, R) is equal to the free R-module generated by singular
n-simplices. We have Cn (X, Z) = Cn (X).
Remark 3.2. Note that C(X) is the chain complex associated with the ∆-set S(X), hence we
automatically have d2 = 0 by Lemma 1.53.
Remark 3.3.* Let ssSet be the category of ∆-sets and Top be the category of topological
spaces. Then the above construction produces a functor S: Top → ssSet, X 7→ S(X). On the
other hand we have the geometric realization functor ssSet → Top, K 7→ |K|. One can show
that Top(|K| , X) ' ssSet(K, S(X)), meaning that these functors are adjoint to each other.
Definition 3.4. Define the singular homology groups of X:
Hn (X) = Hn (C(X)), n ≥ 0.
More generally, define Hn (X, R) = Hn (C(X, R)) for any abelian group R.
F
Proposition 3.5. Let X = α Xα be a decomposition of X into its connected components.
L
Then Hn (X) ' α Hn (Xα ).
Proof. Every singular n-simplex σ: ∆n → X has its image in one connected component. This
L
implies that Cn (X) = α Cn (Xα ). The differential d takes Cn (Xα ) to Cn−1 (Xα ), hence there is
a similar splitting of Ker d and Im d, for example,
  M  
d d
Ker Cn (Xα ) →− Cn−1 (Xα ) = Ker Cn (Xα ) →
− Cn−1 (Xα ) .
α
L
This implies that the same is true for homology groups, hence Hn (X) ' Hn (Xα ).  α
F
Proposition 3.6. If X = 6 ∅ is path-connected, then H0 (X) ' Z. Generally, if X = α Xα is
L
a decomposition into connected components, then H0 (X) ' α Z.
Proof. The singular chain complex has the form
d1
· · · → C1 (X) −
→ C0 (X) → 0 → . . . ,
24 HOMOLOGY THEORY

hence H0 (X) = C0 (X)/ Im d1 . We can identify ∆0 with a point and ∆1 with an interval [0, 1].
Then a 0-simplex can be identified with a point x ∈ X and a 1-simplex can be identified with
a path γ: [0, 1] → X. Under this identification d1 (γ) = γ(1) − γ(0) ∈ C0 (X). Consider a
homomorphism X X
ε: C0 (X) → Z, ni xi 7→ ni .
This map is surjective as X 6= ∅. We will show that Ker ε = Im d1 and this would imply that
Z ' C0 (X)/ Ker ε ' C0 (X)/ Im d1 as required.
P
We have εd1 (γ) = ε(γ(1) − γ(0)) = 0, hence Im d1 ⊂ Ker ε. Assume that i ni xi ∈ Ker ε,
P
which means that i ni = 0. Choose a basepoint x0 ∈ X and a path γi : [0, 1] → X from x0 to
P P
xi (meaning that γ(0) = x0 and γ(1) = xi ) for every i. Then d1 ( ni γi ) = ni (xi − x0 ) =
P P P
ni xi − ( ni ) x0 = ni xi . This implies that Ker ε ⊂ Im d1 . 
Proposition 3.7. If X is a point, then
(
Z n = 0,
Hn (X) =
0 n > 0.
Proof. If X is a point, there is a unique singular n-simplex σn : ∆n → X. Moreover, we have
∂i (σn ) = σn−1 , hence dn (σn ) = ni=0 (−1)i σn−1 = 0 for n odd and equals σn−1 for n even. This
P

means that the singular chain complex C(X) has the form
0 0
· · · → Z −∼ − Z −∼
→Z→ →Z→
− Z→0
We see that for n > 0 we have Hn (X) = Ker(0)/ Im(Id) = Z/Z = 0 if n is odd and Hn (X) =
Ker(Id)/ Im(0) = 0 if n is even. We also have H0 (X) = Z. 
Definition 3.8. Given a nonempty space X, define the augmented chain complex
d 2 1 d ε
· · · → C2 (X) −
→ C1 (X) −
→ C0 (X) →
− Z → 0,
P P
where ε ( i ni xi ) = i ni . Its homology groups H̃n (X) are called reduced homology groups.
Remark 3.9. We have H̃n (X) = Hn (X) for n > 0. On the other hand ε∗ : H0 (X) → Z is
surjective and H̃0 (X) = Ker ε∗ . If X is path-connected, then H̃0 (X) = 0. In general there is a
(non-canonical) isomorphism H0 (X) ' H̃0 (X) ⊕ Z.
Example 3.10. For a 1-point set {x}, we have H̃n ({x}) = 0 for all n.
 HOMOLOGY THEORY 25

3.2. Homotopy invariance. Let f : X → Y be a continuous map between two spaces. It

induces a map f∗ : S(X) → S(Y ) between ∆-sets, where a simplex σ: ∆n → X is sent to
f] (σ) = f ◦ σ: ∆n → Y . It also induces a map f] : C(X) → C(Y ) between chain complexes.
Proposition 3.11. The map f] : S(X) → S(Y ) is a simplicial morphism 1.25.
Proof. Given an increasing map φ: [m] → [n], we need to show that the following diagram
commutes
φ∗
Sn (X) Sm (X)
f] f]
φ∗
Sn (Y ) Sm (Y )
For any σ: ∆n → X, we have f] φ∗ (σ) = f] (σ∆φ ) = f ◦ (σ∆φ ) and φ∗ f] (σ) = φ∗ (f σ) = (f σ) ◦ ∆φ ,
where ∆φ : ∆m → ∆n . 
Proposition 3.12. The map f] : C(X) → C(Y ) is a chain map 2.4. Therefore it induces a
homomorphism f∗ = Hn (f] ): Hn (X) → Hn (Y ) for all n ≥ 0.
Proof. By the previous result we have f] ∂i = ∂i f] , where ∂i = δi∗ and δi : [n − 1] → [n] is the i-th
face. This implies that f] dn = dn f] , where dn = ni=0 (−1)i ∂i .
P

dn
Cn (X) Cn−1 (X)
f] f]
dn
Cn (Y ) Cn−1 (Y )

Definition 3.13. Let X, Y be two spaces.
(1) Two maps f, g: X → Y are called homotopic (written as f ∼ g) if there exists a
(continuous) map F : X × I → Y such that F0 = f and F1 = g, where Ft (x) = F (x, t).
The map F as above is called a homotopy between f and g.
(2) A map f : X → Y is called null-homotopic if it is homotopic to a constant map.
(3) A map f : X → Y is called a homotopy equivalence if there exists a map g: Y → X
such that gf ∼ 1X and f g ∼ 1Y . The spaces X, Y are called homotopy equivalent
in this case.
(4) A space X is called contractible (or null-homotopic) if 1X is null-homotopic, meaning
that there exists a (continuous) map F : X × I → X such that F0 = 1X and F1 is a
constant map. Equivalently, X is homotopy equivalent to a point.
Theorem 3.14. If two continuous maps f, g: X → Y are homotopic, then they induce the same
homomorphisms f∗ = g∗ : Hn (X) → Hn (Y ).
Proof. For every t ∈ I = ∆1 , consider the map η t : X → X × I, x 7→ (x, t). In particular,
consider chain maps η]0 , η]1 : C(X) → C(X × I). We will construct a chain homotopy P : η]0 ∼ η]1 ,
that is, maps P : Cn (X) → Cn+1 (X × I), called prism operators, satisfying
dP + P d = η]1 − η]0 .
If F : X × I → Y is a homotopy between f, g, then f = F η 0 and g = F η 1 . Therefore we obtain
a chain homotopy F] P between f] = F] η]0 and g] = h] η]1 . This implies that f∗ = g∗ .
A singular simplex σ: ∆n → X induces a map
(σ × 1)] : Cn+1 (∆n × I) → Cn+1 (X × I).
26 HOMOLOGY THEORY

We will construct
P (σ) = (σ × 1)] (an ) ∈ Cn+1 (X × I),
where an ∈ Cn+1 (∆ × I) is a certain chain corresponding to a triangulation of ∆n × I (the
n

prism). Then the equation

dP = η]1 − η]0 − P d
can be interpreted as a statement that the boundary of ∆n × I can be represented as a
combination of ∆n × {1}, ∆n × {0} and the sides ∂∆n × I.
We will use the triangulation of ∆n × I constructed in Theorem 1.46. In particular, the
parametrization of maximal simplices of ∆n × I that appeared there. Consider the vectors
vi = (ei , 0) and wi = (ei , 1) in ∆n × I and the singular n + 1-simplices [v0 , . . . , vi , wi , . . . , wn ] ∈
Cn+1 (∆n × I) (cf. Theorem 1.46). Define
Xn
an = (−1)i [v0 , . . . , vi , wi , . . . , wn ] ∈ Cn+1 (∆n × I)
i=0

and P (σ) = (σ × 1)] (an ). Then

(−1)i+j (σ × 1)|[v0 , . . . , v̂j , . . . , vi , wi , . . . , wn ]

X
dP (σ) =
j≤i

(−1)i+(j+1) (σ × 1)|[v0 , . . . , vi , wi , . . . , ŵj , . . . , wn ].

X
+
j≥i

On the other hand

(−1)j (∂j σ × 1)] (an−1 )

X X
P d(σ) = (−1)j P (∂j σ) =
j j

(−1)j+(i−1) (σ × 1)|[v0 , . . . , v̂j , . . . , vi , wi , . . . , wn ]

X
=
j<i

(−1)j+i (σ × 1)|[v0 , . . . , vi , wi , . . . , ŵj , . . . , wn ].

X
+
j>i

The sum dP (σ) + P d(σ) contains only summands with i = j and all of them cancel except the
following two:
(σ × 1)|[v̂0 , w0 , . . . , wn ] − (σ × 1)|[v0 , . . . , vn , ŵn ] = η 1 σ − η 0 σ.
We conclude that dP + P d = η]1 − η]0 . 
Corollary 3.15. We have
(1) If f : X → Y is a homotopy equivalence, then f∗ : Hn (X) → Hn (Y ) is an isomorphism
(the same is true for the reduced homologies).
(2) If X is contractible, then H̃n (X) = 0 for all n.
Proof. (1) Let g: Y → X be homotopy inverse to f , meaning that gf ∼ 1X and f g ∼ 1Y . This
implies that (gf )∗ = 1: Hn (X) → Hn (X) and (f g)∗ = 1: Hn (Y ) → Hn (Y ) by the previous
theorem. Therefore f∗ : Hn (X) → Hn (Y ) has the inverse g∗ : Hn (Y ) → Hn (X), hence f∗ is an
isomorphism.
(2) If X is contractible, then X is homotopy equivalent to a point, hence H̃n (X) ' H̃n (pt) by
the previous statement. But we know that H̃n (pt) = 0 for all n, hence H̃n (X) = 0 for all n. 
Example 3.16. The space Dn is contractible, hence H̃i (Dn ) ' H̃i (pt) = 0 for all i. The same
is true for Rn .
 HOMOLOGY THEORY 27

3.3. Relative homology groups.

Definition 3.17.
(1) A pair (X, A) of topological spaces consists of a space X and a subspace A ⊂ X.
(2) A map of pairs f : (X, A) → (Y, B) is a continuous map f : X → Y such that f (A) ⊂ B.
Given pair (X, A), define
Cn (X, A) = Cn (X)/Cn (A), n ≥ 0.
The differential d: Cn (X) → Cn−1 (X) maps Cn (A) to Cn−1 (A), hence it induces a map
d: Cn (X, A) → Cn−1 (X, A). We obtain a chain complex
d d
· · · → Cn+1 (X, A) →
− Cn (X, A) →
− Cn−1 (X, A) → . . .
Its homology groups Hn (X, A) are called relative homology groups.
Theorem 3.18. For any pair (X, A) there is a long exact sequence
∂ ∂
· · · → Hn (A) → Hn (X) → Hn (X, A) →
− Hn−1 (A) → Hn−1 (X) → Hn−1 (X, A) →
− ...
· · · → H0 (X) → H0 (X, A) → 0
Proof. We have a short exact sequence of chain complexes
0 → C(A) → C(X) → C(X, A) → 0
by the construction of the complex C(X, A). Applying Theorem 2.13 we obtain the required
long exact sequence. 
Remark 3.19. If A 6= ∅, then we can use reduced homology groups:

· · · → H̃n (A) → H̃n (X) → Hn (X, A) → H̃n−1 (A) → H̃n−1 (X) → Hn−1 (X, A) → . . .
· · · → H̃0 (X) → H0 (X, A) → 0

Example 3.20. Given a point x ∈ X, we have H̃n (x) = 0 for all n, hence we obtain from the
long exact sequence
Hn (X, x) ' H̃n (X) ∀n.
Example 3.21. Consider a disc Dn and its boundary ∂Dn = S n−1 . We have an exact sequence
· · · → H̃i (Dn ) → Hi (Dn , S n−1 ) → H̃i−1 (S n−1 ) → H̃i−1 (Dn ) → . . .
As Dn is contractible, we obtain that H̃i (Dn ) = 0 for all i. Therefore there is an isomorphism
Hi (Dn , S n−1 ) ' H̃i−1 (S n−1 )
Remark 3.22. A subspace A ⊂ X is called a retract of X if there exists a map r: X → A such
that ri = 1A , where i: A → X is the embedding. In this case the complex
i]
0 → C(A) −
→ C(X) → C(X, A) → 0
splits by the map r] (we get a direct sum decomposition C(X) = C(A) ⊕ C(X, A)). Therefore
there is a short exact sequence
0 → Hn (A) → Hn (X) → Hn (X, A) → 0
which splits by r∗ . This means that the boundary map ∂: Hn (X, A) → Hn−1 (A) in the above
long exact sequence is zero.
28 HOMOLOGY THEORY

3.4. Excision and Mayer-Vietoris. Let X be a space and U = (Ui )i∈I be a collection of
subspaces, such that X = i Ui◦ . Define SnU (X) ⊂ Sn (X) to be the subset of simplices σ: ∆n → X
S

with Im σ ⊂ Ui for some i. In the same way we define the chain complex C U (X) ⊂ C(X), and
its homology groups HnU (X).

Theorem 3.23. The inclusion j: C U (X) → C(X) is a chain homotopy equivalence. In particular,
it induces isomorphisms HnU (X) ' Hn (X).

Theorem 3.24 (Excision theorem). Let Z ⊂ A ⊂ X be subspaces such that Z̄ ⊂ A◦ . Then the
inclusion (X − Z, A − Z) ,→ (X, A) induces isomorphisms
Hn (X − Z, A − Z) ' Hn (X, A) ∀n.
Equivalently, for any subspaces A, B ⊂ X with X = A◦ ∪ B ◦ (we take B = X − Z), the inclusion
(B, A ∩ B) ,→ (X, A) induces isomorphisms
Hn (B, A ∩ B) ' Hn (X, A) ∀n.

Proof. Let A, B ⊂ X and X = A◦ ∪ B ◦ . Consider the cover U = (A, B). We have embeddings
Cn (A) ⊂ Cn (X), Cn (B) ⊂ Cn (X) and
CnU (X) = Cn (A) + Cn (B), C n (A ∩ B) = Cn (A) ∩ Cn (B).
Therefore
Cn (B) Cn (B) Cn (A) + Cn (B) C U (X)
= ' = n
Cn (A ∩ B) Cn (A) ∩ Cn (B) Cn (A) Cn (A)
and
Hn (B, A ∩ B) ' Hn (C U (X)/C(A)) ' Hn (C(X)/C(A)) = Hn (X, A).
To see the second isomorphism in more detail, consider a commutative diagram

0 C(A) C U (X) C U (X)/C(A) 0

0 C(A) C(X) C(X)/C(A) 0

Taking the long exact sequences of both rows we get

Hn (A) HnU (X) Hn (C U (X)/C(A)) Hn+1 (A) U

Hn+1 (X)

Hn (A) Hn (X) Hn (C(X)/C(A)) Hn+1 (A) Hn+1 (X)

The outer arrows are isomorphisms by the previous theorem. Therefore the middle arrow is also
an isomorphism by the 5-lemma. 

Example 3.25. We know that H̃i (S n ) ' Hi (S n , pt). Let D− n

⊂ S n be the lower hemisphere.
n
Then H̃i (D− ) = H̃i (pt) = 0 and we obtain Hi (S n , pt) ' Hi (S n , D−
n
) using the 5-lemma. We
n n n
apply the excision theorem to X = S , A = D− and Z ⊂ D− a slightly smaller disc. Note that
n
X − Z is homotopic to D+ (the upper hemisphere) and A − Z is homotopic to S n−1 . We obtain
Hi (S n , D−
n n
) ' Hi (X − Z, A − Z) ' Hi (D+ , S n−1 ) ' H̃i−1 (S n−1 ),
where the last isomorphism follows from the long exact sequence for the pair (Dn , S n−1 ). We
obtain
H̃i (S n ) ' Hi (S n , pt) ' Hi (S n , D−
n
) ' H̃i−1 (S n−1 ).
 HOMOLOGY THEORY 29

Note that S 0 = {x ∈ R | |x| = 1} = {±1}, hence H̃0 (S 0 ) = Z and H̃i (S 0 ) = 0 for i 6= 0. We

obtain inductively (
Z i=n
H̃i (S n ) =
0 i 6= n.
Theorem 3.26 (Mayer-Vietoris sequence). Let A, B ⊂ X be subspaces such that A◦ ∪ B ◦ = X.
Then there is a long exact sequence
∂
· · · → Hn (A ∩ B) → Hn (A) ⊕ Hn (B) → Hn (X) →
− Hn−1 (A ∩ B) → · · · → H0 (X) → 0
If A ∩ B 6= ∅, then we have a similar exact sequence of reduced homology groups.
Proof. Let Cn (A + B) = Cn (A) + Cn (B) ⊂ Cn (X). There is a short exact sequence of complexes
0 → C(A ∩ B) → Cn (A) ⊕ Cn (B) → Cn (A + B) → 0
which induces the corresponding long exact sequence. The required long exact sequence follows
from the fact that the inclusion C(A + B) → C(X) induces an isomorphism of homology groups
by excision. 
Example 3.27. Let X = S n and let A ⊃ D+ n
and B ⊃ D− n
be open neighborhoods. Then
n−1
A ∩ B is homotopic to S and A, B are contractible. We obtain an exact sequence
H̃i (A) ⊕ H̃i (B) → H̃i (S n ) → H̃i−1 (S n−1 ) → H̃i−1 (A) ⊕ H̃i−1 (B)
The outer groups are trivial, hence H̃i (S n ) ' H̃i−1 (S n−1 ). This allows us to compute H̃i (S n ) in
the same way as before.
Example 3.28. Define the suspension
X ×I
ΣX = .
(x, 0) ∼ (y, 0), (x, 1) ∼ (y, 1)
One can think about it as a union of two cones CX = X × I/X × {0} glued along the boundary
X. For example C(S n−1 ) ' Dn and Σ(S n−1 ) is a union of two discs Dn glued along the
boundary, hence Σ(S n−1 ) ' S n . Let A, B ⊂ ΣX be the above cones (more precisely, their
open neighborhoods). Then A, B ' CX are contractible (prove this) and A ∩ B is homotopy
equivalent to X. We obtain an exact sequence
H̃i (A) ⊕ H̃i (B) → H̃i (ΣX) → H̃i−1 (A ∩ B) → H̃i−1 (A) ⊕ H̃i−1 (B)
The outer groups are zero, hence
H̃i (ΣX) ' H̃i−1 (A ∩ B) ' Hi−1 (X).
This generalizes the formula H̃i (S n ) ' H̃i−1 (S n−1 ).
30 HOMOLOGY THEORY

3.5. Long exact sequence for good pairs.

Definition 3.29. Let A ⊂ X be a subspace and i: A ,→ X be an inclusion. Then A is called
(1) A retract if there exists a map r: X → A such that ri = 1A . The map r as above is
called a retraction.
(2) A deformation retract if there exists a map r: X → A such that ri = 1A and ir ∼ 1X .
Equivalently, there exists a homotopy h: X × I → X such that h0 = 1X , h1 (X) ⊂ A and
h1 |A = 1A (we define ht (x) = h(x, t) and r = h1 : X → A). The map h as above is called
a deformation retraction. In this case i∗ : H• (A) → H• (X) is an isomorphism.
(3) A strong deformation retract if there exists a homotopy h: X × I → X such that
h0 = 1X , h1 (X) ⊂ A and ht |A = 1A for all t ∈ I. The map h as above is called a strong
deformation retraction.
Example 3.30. Consider A = S n−1 ⊂ X = Rn \ {0} and define
h: X × I → X, (x, t) 7→ (1 − t + t/ kxk)x.
This is a strong deformation retract, as h0 (x) = x for all x ∈ Rn \ {0}, h1 (x) = x/ kxk ∈ A for
all x ∈ Rn \ {0}, ht (x) = (1 − t + t)x = x for all x ∈ S n−1 . Therefore H• (S n−1 ) → H• (Rn \ {0})
is an isomorphism.
Definition 3.31. A pair (X, A) is called good if A ⊂ X is nonempty, closed and is a strong
deformation retract of its neighborhood V ⊂ X (this means that A ⊂ V ◦ ).
Theorem 3.32. If (X, A) is a good pair, then Hn (X, A) ' H̃n (X/A).
Proof. Consider the quotient map q: (X, A) → (X/A, A/A). The space A/A has just one
point, hence Hn (X/A, A/A) ' H̃n (X/A) by Example 3.20. Therefore we need to show that
q∗ : Hn (X, A) → Hn (X/A, A/A) is an isomorphism.
Let A ⊂ V ⊂ X be a neighborhood such that A is a deformation retract of V . Consider the
diagram

Hn (X, A) Hn (X, V ) Hn (X − A, V − A)
q∗ q∗ q∗

Hn (X/A, A/A) Hn (X/A, V /A) Hn (X/A − A/A, V /A − A/A)

We need to show that the left vertical arrow is an isomorphism. For this we will show that all
horizontal arrows and the right vertical arrow are isomorphisms.
As A ⊂ V is a deformation retract, we obtain that Hn (A) → Hn (V ) are isomorphisms, hence
Hn (X, A) → Hn (X, V ) are isomorphisms (by the long exact sequence and 5-lemma). The same
argument applies to the left bottom arrow: if h: V × I → V is a strong deformation retraction of
A ⊂ V , then it induces a homotopy h: V /A × I → V /A which is a strong deformation retraction
of A/A ⊂ V /A. Then we conclude that Hn (X/A, A/A) ' Hn (X/A, V /A) as before.
The right horizontal arrows are isomorphisms by excision. The right vertical arrow is an
isomorphism because it is induced by a homeomorphism between pairs. 
Theorem 3.33. Let (X, A) be a good pair. Then there is a long exact sequence
i ∗ π ∗ ∂ i
∗ ∗ π
· · · → H̃n (A) −
→ H̃n (X) −→ H̃n (X/A) →
− H̃n−1 (A) −
→ H̃n−1 (X) −→ ...
π ∗
. . . −→ H̃0 (X/A) → 0
where i: A → X is the inclusion and π: X → X/A is the quotient map.
 HOMOLOGY THEORY 31

Proof. By Theorem 3.18 we have a similar long exact sequence for the relative homology groups
Hn (X, A) (note that we can consider reduced homology groups H̃n (X) and H̃n (A) there)
i ∗ ∂ ∗ i
· · · → H̃n (A) −
→ H̃n (X) →
− Hn (X, A) →
− H̃n−1 (A) −
→ H̃n−1 (X) →
− ...
π ∗
. . . −→ H0 (X, A) → 0
By the previous theorem we have Hn (X, A) ' H̃n (X/A). 
Lemma 3.34. We have (
Z i=n
H̃i (S n ) =
0 i 6= n.
Proof. Choose a good pair (X, A) = (Dn , S n−1 ) for n ≥ 1. Then X/A ' S n . We have
H̃i (Dn ) = 0 for all i ≥ 0 as Dn is contractible. From the long exact sequence for the pair (X, A)
we obtain
∂
0 = H̃i (Dn ) → H̃i (S n ) →
− H̃i−1 (S n−1 ) → H̃i−1 (Dn ) = 0,
Hence H̃i (S n ) ' H̃i−1 (S n−1 ) for i > 0 and H̃0 (S n ) = 0. Now the result follows by induction. In
the case of S 0 , we have S 0 = {±1} ⊂ R, hence H0 (S 0 ) ' Z2 and H̃0 (S 0 ) ' Z. We also have
H̃i (S 0 ) = 0 for i > 0. 
Corollary 3.35. Rn are not homeomorphic to each other for different n.
Proof. We have seen in Example 3.30 that S n−1 ⊂ Rn \ {0} is a deformation retract. This implies
H̃i (Rn \ {0}) ' H̃i (S n−1 ) and we have seen that spheres of different dimension have different
homology groups. 
Remark 3.36. Similarly, if U ⊂ Rm and V ⊂ Rn are nonempty open and homeomorphic,
then m = n. Indeed, we have Hi (U, U − {x}) ' Hi (Rm , Rm − {x}) by excision: consider
Z = Rm − U ⊂ A = Rm − {x}. From the long exact sequence for the pair (Rm , Rm − {x}) and
the fact that H̃i (Rm ) = 0, we obtain
Hi (Rm , Rm − {x}) ' H̃i (Rm − {x}) ' H̃i (S m−1 ).
We conclude that H̃i (S m−1 ) ' H̃i (S n−1 ), hence m = n.
Corollary 3.37 (Brouwer fixed point theorem ). A continuous map f : Dn → Dn has a fixed
point.
Proof. Suppose that f (x) 6= x for all x ∈ Dn . Define a map r: Dn → ∂Dn = S n−1 by letting
r(x) to be the point where the ray going from f (x) to x intersects ∂Dn . Then r(x) = x for
x ∈ ∂Dn , hence ri = 1, where i: ∂Dn ,→ Dn (r is a retraction). But this implies that
H̃n−1 (∂Dn ) → H̃n−1 (Dn ) → H̃n−1 (∂Dn )
is an identity. However, H̃n−1 (∂Dn ) ' H̃n−1 (S n−1 ) ' Z while H̃n−1 (Dn ) = 0 as Dn is con-
tractible. This is a contradiction. 
32 HOMOLOGY THEORY

3.6. Equivalence of simplicial and singular homologies. Let (X, K) be a triangulated

space and C K (X) = C(K) be its simplicial chain complex. Every simplex σ: ∆n → X from
the triangulation can be considered as a singular n-simplex. Therefore we have an embedding
i: C K (X) ,→ C(X) to the singular chain complex. It induces maps between homology groups

i∗ : HnK (X) → Hn (X)

and we are going to prove that these are isomorphisms. This means that simplicial homologies
are independent of the triangulation and coincide with singular homologies.
More generally, assume that A ⊂ X is a union of simplices. Then we can construct the
simplicial chain complex of the pair (X, A)

C K (X, A) = C K (X)/C K (A)

and the corresponding homology groups HnK (X, A). As before, there is a map i: C K (X, A) →
C(X, A) which induces a map between homology groups

i∗ : HnK (X, A) → Hn (X, A)

and we want to show that these are isomorphisms.

Theorem 3.38. The homomorphisms HnK (X, A) → Hn (X, A) are isomorphisms for all n.

Proof. Let us assume that A = ∅. Let X i ⊂ X be the i-skeleton of X consisting of simplices of

dimension ≤ i. Consider the following commutative diagram of exact sequences:

∆
Hn+1 (X i , X i−1 ) Hn∆ (X i−1 ) Hn∆ (X i ) Hn∆ (X i , X i−1 ) ∆
Hn−1 (X i−1 )

Hn+1 (X i , X i−1 ) Hn (X i−1 ) Hn (X i ) Hn (X i , X i−1 ) Hn−1 (X i−1 )

We can assume that the second and the fifth vertical maps are isomorphisms by induction
on k. Let us show that the forth (hence also the first) vertical map is an isomorphism. The
group of n-chains CnK (X i , X i−1 ) is zero for n 6= i, and is free abelian with the basis consisting of
n-simplices of X for n = i. Therefore the same is true for the homology groups HnK (X i , X i−1 ).
Consider the characteristic maps Φα : ∆iα = ∆i → X for i-simplices of X, and the corresponding
map Φ: α (∆iα , ∂∆iα ) → (X i , X i−1 ). It induces a homeomorphism
F

G G
∆iα / ∂∆iα → X i /X i−1 ,
α α

hence isomorphisms of singular homology groups. But ∆i is homeomorphic to Di and ∂∆i is

homeomorphic to ∂Di = S i−1 and we know by Lemma 3.34
(
Z n = i,
H̃n (Di /S i−1 ) = H̃n (S i ) =
0 n 6= i.

By Theorem 3.32 we have Hn (X i , X i−1 ) ' H̃n (X i /X i−1 ). Hence we conclude that it is zero for
n 6= i and is a free abelian group with a basis given by n-simplices of X for n = i. We conclude
that the forth vertical map is indeed an isomorphism. By the 5-lemma 2.14 the third vertical
map is also an isomorphism. 
 HOMOLOGY THEORY 33

3.7. CW-complexes and cellular homology. Define an n-cell to be a space homeomorphic

◦
to Dn . A CW-complex (or cell complex) is a space X equipped with a decomposition
into cells (eα )α∈K such that, if X n ⊂ X is the union of cells up to dimension n (called the
n-skeleton), then
(1) X 0 is a discrete space.
(2) For every n-cell eα , there is a continuous map Φα : Dn → X n which induces a homeomor-
◦
phism Dn → eα and sends S n−1 = ∂Dn to X n−1 .

Definition 3.39. A CW-decomposition of a space X is a chain

∅ = X −1 ⊂ X 0 ⊂ X 1 ⊂ . . . ⊂ X

of spaces such that

(1) The space X n is obtained from X n−1 by attaching n-cells via attaching maps

φα : Sαn−1 → X n−1 ,

meaning that X n is the quotient space (X n−1 α Dαn )/ ∼ under the identifications
`

φα (x) ∼ x for x ∈ Sαn−1 . The map Φα : Dαn → X n is called the characteristic map of
◦
the n-cell enα = Φα (Dαn ). There is a commutative (pushout) diagram
φ
Sαn−1 X n−1
F
α

Φ
Dαn Xn
F
α

X n has the weak topology: U ⊂ X is open ⇐⇒ U ∩ X n ⊂ X n is open ∀n.

S
(2) X = n≥0

The space X equipped with a CW-decomposition is called a CW-complex (or cell complex).
The space X n is called the n-skeleton of X. A space is called a CW-space if it has some
CW-decomposition.

Theorem 3.40. We have

(1) Hk (X n , X n−1 ) = 0 for k 6= n and is free with a basis given by the n-cells if k = n.
(2) H k (X n ) = 0 for k > n.
(3) The inclusion i: X n ,→ X induces isomorphisms i∗ : Hk (X n ) → Hk (X) for k < n.

Proof. (1) We can identify Dn /S n−1 with a pointed sphere S n . The quotient X n /X n−1 can be
identified with ( α Dαn )/( α Sαn−1 ) ' α Sαn (wedge sum). We have H̃n ( α Sαn ) ' α H̃n (Sαn ).
` ` W W L

Therefore
M M
Hn (X n , X n−1 ) ' H̃n (X n /X n−1 ) ' H̃n (Sαn ) ' Z.
α α

(2) We have an exact sequence

Hk+1 (X n , X n−1 ) → Hk (X n−1 ) → Hk (X n ) → Hk (X n , X n−1 )

By the previous statement the first and the fourth components are zero for k > n. By induction
we can assume that Hk (X n−1 ) = 0, hence we conclude that Hk (X n ) = 0.
(3) From the previous argument we obtain Hk (X n−1 ) ' Hk (X n ) for k 6= n, n − 1. Therefore
Hk (X n ) ' H k (X n+1 ) for k 6= n, n+1 and in particular, for k < n. This implies Hk (X n ) ' H k (X)
(if X is finite-dimensional). 
34 HOMOLOGY THEORY

The long exact sequence for the pair (X n , X n−1 ) gives in particular
jn ∂ pn
0 → Hn (X n ) −
→ Hn (X n , X n−1 ) −→
n
Hn−1 (X n−1 ) −→ Hn−1 (X n ) → 0
Let us organize these maps into a commutative diagram, with dn = jn−1 ∂n ,
Hn (X n ) Hn−1 (X n−1 )
∂n+1 jn ∂n jn−1

dn+1 dn
... Hn+1 (X n+1 , X n ) Hn (X n , X n−1 ) Hn−1 (X n−1 , X n−2 ) ...
Note that dn dn+1 = 0 as ∂n jn = 0. The above complex is called a cellular chain complex
and its homology groups are called cellular homology groups, denoted by HnCW (X).
Theorem 3.41. We have HnCW (X) ' Hn (X).
Proof. We have Ker dn = Ker ∂n ' Hn (X n ) as jn−1 is injective. Therefore
Ker dn / Im dn+1 ' Hn (X n )/ Im dn+1 = Hn (X n )/ Im ∂n+1 ' Hn (X n+1 ) ' Hn (X).

Recall that Hn (X n , X n−1 ) is a free abelian group with a basis given by n-cells enα . For any
φα
such cell we have the attaching map Sαn−1 −→ X n−1 as well as the quotient map
q qα
_
Xn →− X n /X n−1 ' (Dαn /Sαn−1 ) −→ Dαn /Sαn−1 ' Sαn
α

Theorem 3.42. The differential dn : ZKn → ZKn−1 is given by dn (enα ) = dαβ en−1
P
β∈Kn−1 β , where
dαβ is the degree of the composition Sαn−1 → X n−1 → Sβn−1 .
 HOMOLOGY THEORY 35

4. Applications of Homology theory

4.1. Degree. Given a map f : S n → S n , consider the induced map
f∗ : Hn (S n ) → Hn (S n ).
We know that Hn (S n ) ' Z, hence f∗ is of the form σ 7→ dσ for some d ∈ Z. This integer is
called the degree of f , denoted by deg f .
Proposition 4.1. We have
(1) deg 1 = 1.
(2) f ' g =⇒ deg f = deg g.
(3) deg f g = deg f · deg g.
(4) If f is a reflection with respect to some hyperplane in Rn+1 , then deg f = −1.
(5) If f = −1, the antipodal map, then deg f = (−1)n+1 .
(6) If f : S n → S n is not surjective, then deg f = 0.
Proof. (1) We have 1∗ = 1, the identity map Hn (S n ).
(2) If f ' g, then f∗ = g∗ .
(3) (f g)∗ = f∗ g∗ .
(4) Let S n−1 ⊂ S n be fixed by the reflection. Consider a triangulation of S n with two
hemispheres σ, τ as its n-simplices. Then Hn (S n ) has a generator σ − τ . The reflection
f interchanges σ and τ , hence f∗ (σ − τ ) = τ − σ. This means that deg f = −1.
(5) The antipodal map can be represented as a composition of n+1 reflections, each changing
the sign of one coordinate.
(6) If f is not surjective then it is a composition of S n → S n − {x} → S n for some
x ∈ S n − Im f . But S n − {x} is contractible, hence f∗ : Hn (S n ) → Hn (S n ) is zero.

Proposition 4.2. If f : S n → S n does not have fixed points, then deg f = (−1)n+1 .
Proof. If f (x) 6= x for all x ∈ S n , then gt (x) = tf (x) − (1 − t)x =
6 0 for t ∈ [0, 1]. Define
the homotopy ht : S → S , ht (x) = gt (x)/ kgt (x)k between h0 = −1 and h1 = f . Then
n n

deg f = deg(−1) = (−1)n+1 . 

In order to compute the degree of a map one introduces local degrees. Recall that if U ⊂ Rn
is open and x ∈ U , then
H n (U, U − x) ' H n (Rn , Rn − x) ' H̃n−1 (Rn − x) ' H̃n−1 (S n−1 ) ' Z,
where the first isomorphism follows from excision, the second from the long exact sequence
H̃n (Rn ) → H n (Rn , Rn − x) → H̃n−1 (Rn − x) → H̃n−1 (Rn )
and the third follows from the fact that Rn − x is homotopic to S n−1 .
Assume now that we have f : S n → S n and a point y ∈ S n such that the preimage f −1 (y)
consists of finitely many points x1 , . . . , xm . We can choose disjoint neighborhoods U1 , . . . , Um of
these points which are mapped to a neighborhood V of y. Then we obtain a homomorphism
f∗ : Z ' Hn (U, U − xi ) → Hn (V, V − y) ' Z
which is a multiplication by an integer, called the local degree of f at xi , denoted by deg f |xi .
Note that if f maps Ui homeomorphically to V , then the above homomorphism is an isomorphism,
hence deg f |xi = ±1.
P
Theorem 4.3. We have deg f = i deg f |xi .
36 HOMOLOGY THEORY

Proof. Consider X = S n , B = i Ui and A = S n − f −1 (y). Then by excision Hn (X, A) '

S
S
Hn (B, A ∩ B) and we have A ∩ B = i (Ui − xi ). Therefore
M
Hn (S n , S n − f −1 (y)) ' Hn (Ui , Ui − xi ) ' Zm .
i
To prove the theorem we consider
(i)
f∗
Hn (Ui , Ui − xi ) Hn (V, V − y) = Z
∼ ji

∼
f∗0
Hn (Sn , Sn − xi ) Hn (S n , S n − f −1 (y)) Hn (S n , S n − y) = Z
j

∼
∼
f∗
Hn (S n ) Hn (S n ) = Z
P
and observe that j(1) = i ji (1). But this means that f∗ (1) = deg f can be identified with
P (i)
f∗0 j(1) = i f∗0 ji (1) or with i f∗ (1) = i deg f |xi .
P P

4.2. Hedgehog theorem. For any point x ∈ S n ⊂ Rn+1 , its tangent space Tx S n consists
of vectors in Rn+1 perpendicular to x. A (continuous) vector field is a continuous map
f : S n → Rn+1 such that f (x) ∈ Tx S n , a tangent vector at x ∈ S n .
The following statement says that one can not comb a hairy ball without a cowlick:
Theorem 4.4 (Hedgehog theorem). There is no continuous tangent vector field on S 2 that is
nowhere zero.
We can generalize this statement to any even dimension:
Theorem 4.5 (2.28). If n is even, there is no continuous tangent vector field on S n that is
nowhere zero.
Proof. Let f : S n → Rn+1 be a vector field. If f (x) 6= 0 for all x, we can normalize it by taking
f (x)/ kf (x)k, hence we can assume that f (x) ∈ S n . The vectors x and f (x) are orthogonal to
each other in Rn+1 (the tangent space of S n at x is orthogonal to x). The vectors
ht (x) = cos(t)x + sin(t)f (x), t ∈ [0, π]
are contained in S n . Therefore ht : S n → S n is a homotopy between h0 (x) = 1 and h1 = −1
(the antipodal map). Therefore deg(−1) = deg 1, hence (−1)n+1 = 1. A contradiction as n is
even. 
Remark 4.6. Note that there is a non-vanishing vector field on S 1 . Indeed, represent S 1 =
{z ∈ C | kzk = 1} and define f (z) = iz, which is orthogonal to z. In real coordinates this means
z = x1 + ix2 7→ −x2 + ix1 , hence (x1 , x2 ) 7→ (−x2 , x1 ). More generally, for any odd n = 2k − 1,
consider the map
R2k 3 x 7→ f (x) = (−x2 , x1 , . . . , −x2k , x2k−1 ).
Then f (x) is orthogonal to x, hence f defines a non-vanishing vector field on S n .
4.3. Jordan curve theorem.
Theorem 4.7 (Jordan curve theorem). A subspace of R2 homeomorphic to S 1 separates R2
into two connected components.
We can consider S 2 instead of R2 as one obtains R2 from S 2 be removing a point and this
does not affect connectedness. The result then follows from a more general statement.
 HOMOLOGY THEORY 37

Theorem 4.8. Given an embedding f : Dk → S n , we have H̃i (S n − f (Dk )) = 0 for all i.

Proof. For k = 0, we have S n − f (D0 ) ' Rn and the statement is obvious. Generally, we can
identify Dk with a cube I k . Consider a decomposition I k = D1 ∪ D2 , where I 0 = I k−1 × [0, 21 ]
and I 00 = I k−1 × [ 12 , 1]. Let A = S n − f (I 0 ) and B = S n − f (I 00 ). Then

A ∪ B = S n − f (D1 ∩ D2 ) = S n − f (I k−1 ), A ∩ B = S n − f (D1 ∪ D2 ) = S n − f (I k ).

By induction H̃i (S n − f (I k−1 )) = 0, hence by the Meyer-Vietoris we have

H̃i (S n − f (I k )) ' H̃i (A) ⊕ H̃i (B),

induced by the embeddings S n − f (Ik ) → A and S n − f (Ik ) → B. Assume there exists

σ ∈ Zi (S n − f (I k )) which is not in the boundary. Then its image in Zi (A) or in Zi (B) is not in
the boundary. Continuing this process, we find a chain of shrinking intervals I1 ⊃ I2 ⊃ . . . such
that σ is not in the boundary in Um = S n − f (I k−1 × Im ) for all m.
We have U1 ⊂ U2 ⊂ . . . and U = m Um = S n − f (I k−1 × {p}), where {p} = m Im . The
S T

image of σ is in the boundary in U = S n − f (I k−1 × {p}) by induction. Let τ ∈ Ci+1 (U ) be

such that σ = dτ and let Z ⊂ U be its support. Then (S n − Z) ∪ m Um is an open cover of
S

S n , hence it contains a finite subcover and we conclude that (S n − Z) ∪ Um = S n for some m.

Then Z ⊂ Um , hence τ ∈ Ci+1 (Um ) and we conclude that σ ∈ Bi (Um ), a contradiction. 

Theorem 4.9. Given an embedding f : S k → S n with k < n, we have

(
Z i = n − k − 1,
H̃i (S n − f (S k )) =
0 otherwise.

Proof of Theorem 4.7. Taking k = 1 and n = 2, we obtain H̃0 (S 2 − f (S 1 )) ' Z. Therefore

H0 (S 2 − f (S 1 )) ' Z2 , hence there are two connected components in the complement. 

Proof. If k = 0, then S n − f (S 0 ) ' Rn − {0} which is homotopic to S n−1 . But we know how to
compute H̃i (S n ). For the induction on k, write S k as a union of hemispheres D+
k k
, D− intersecting
k−1 n k n k
at S . Let A = S − f (D+ ), B = S − f (D− ). Then

A ∩ B = S n − (f (D+
k k
) ∪ f (D− )) = S n − f (S k ), A ∪ B = S n − f (S k−1 ).

The spaces A, B have trivial reduced homology 4.8, hence by the Meyer-Vietoris 3.26 we have
H̃i (A ∪ B) ' H̃i−1 (A ∩ B). Therefore

H̃i−1 (S n − f (S k )) ' H̃i (S n − f (S k−1 ))

and we apply induction. 

Corollary 4.10. One cannot embed S n in Rn .

Proof. Given such an embedding, we would get a non-surjective embedding f : S n → S n . Using

notation from the previous theorem, consider A = S n − f (D+
n
) and B = S n − f (D−
n
). Assuming
n n
if A ∩ B = S − f (S ) 6= ∅, then MV gives an exact sequence

H̃0 (A) ⊕ H̃0 (B) → H̃0 (A ∪ B) = H̃0 (S n − f (S n−1 )) → 0

But the left group is zero, while the right group is nonzero by the previous theorem. We conclude
that A ∩ B = ∅, hence f (S n ) = S n and f is sujrective, a contradiction. 
38 HOMOLOGY THEORY

4.4. Invariance of domain.

Theorem 4.11. Let U ⊂ Rn be open and f : U → Rn be injective continuous. Then f (U ) is
open in Rn and f maps U homeomorphically onto f (U ).
Proof. We can embed Rn ⊂ S n using a one-point compactification and consider f : U → S n .
It is enough to show that f (U ) ⊂ S n is open. Given a ∈ U , choose ε > 0 such that D =
{x ∈ Rn | kx − ak ≤ ε} ⊂ U . It suffices to show that f (D◦ ) is open. We have ∂D ' S n−1
and by the previous results S n − f (∂D) has two path-connected components. Note that
f (D − ∂D) is path-connected as D − ∂D is path connected, and S n − f (D) is path-connected as
H̃0 (S n − f (D)) = 0 by the previous results. This means that these two sets are path-connected
components of S n − f (∂D). Such components are closed, hence open as there are finitely many
of them. We conclude that f (D◦ ) = f (D − ∂D) is open. 
Definition 4.12. Define a topological n-manifold to be a Hausdorff space locally homeo-
morphic to Rn (this means that every point has an open neighborhood homeomorphic to an
open subset U ⊂ Rn ).
Corollary 4.13. If M, N are topological n-manifolds, M is compact and N is connected, then
any embedding f : M → N is a homeomorphism.
Proof. By the previous result f (M ) is open. But it is also closed as M is compact. As N is
connected, we conclude that f (M ) = N , hence f : M → N is a bijection. It is an open/closed
map, hence a homeomorphism. 
4.5. Algebraic applications.
Lemma 4.14. We have H1 (RP n ) ' Z2 for n ≥ 2. In particular, RP n 6' S n for n ≥ 2.
Proof. The space RP n ' (Rn+1 \ {0})/R∗ ' S n /Z2 can be constructed as a quotient of Dn ,
where we identify the antipodal points of ∂Dn = S n−1 . Under this identification we get
S n−1 /Z2 ' RP n−1 . In this way we obtain RP n−1 ⊂ RP n such that
RP n /RP n−1 ' Dn /S n−1 ' S n .
Therefore we have a long exact sequence
· · · → H̃i+1 (S n ) → H̃i (RP n−1 ) → H̃i (RP n ) → H̃i (S n ) → . . .
This implies H̃1 (RP n ) ' H̃1 (RP n−1 ) for n ≥ 3. Now we use the fact H1 (RP 2 ) ' Z2 . 
Theorem 4.15. The only finite field extensions of R are R and C.
Proof. Let K = Rn be equipped with a field structure. Let kxk denote the Euclidean norm of
x ∈ Rn . Consider a continuous map
f : S n−1 → S n−1 , x 7→ x2 / x2 .
If f (x) = f (y), then x2 = a2 y 2 , where a2 = kx2 k / ky 2 k ∈ R. Therefore (x − ay)(x + ay) = 0,
hence x = ±ay. As x, y ∈ S n−1 , we conclude that x = ±y. Therefore we get an injective
continuous map
f¯: RP n−1 = S n−1 / {±1} → S n−1 .
By 4.13, f¯ is a homeomorphism. But this is not possible if n ≥ 3.
Let us show that for n = 2, we have K ' C. Let x ∈ K\R and x2 = a + 2bx for some a, b ∈ R.
Then (x − b)2 = a + b2 . If a + b2 ≥ 0, then x ∈ R, a contradiction. If a + b2 = −c2 < 0, then
y = (x − b)/c satisfies y 2 = −1, hence K ' C. 
 HOMOLOGY THEORY 39

Remark 4.16. It was proved by Frobenius that the only finite-dimensional division algebras
over R (not necessarily commutative) are R, C and H (the algebra of quaternions).

Theorem 4.17 (Fundamental theorem of algebra). The field C is algebraically closed (every
non-constant polynomial has a root).

Proof. By the previous result every finite field extension of C coincides with C. But this implies
that C is algebraically closed: if f (x) is an irreducible polynomial, then K = C[x]/(f ) is a field,
hence K = C and deg f = 1. Therefore f has a root. 

4.6. Borsuk-Ulam theorem. Let A ⊂ Rm and B ⊂ Rn be such that −A = A and −B = B.

We call a map f : A → B odd if f (−x) = −f (x) for all x ∈ A.

Theorem 4.18. The following statements are equivalent (all maps are continuous)
(1) For any map f : S n → Rn , there exists x ∈ S n such that f (x) = f (−x).
(2) For any odd map f : S n → Rn , there exists x ∈ S n such that f (x) = 0.
(3) There does not exist an odd map f : S n → S n−1 .
(4) There does not exist a map f : Dn → S n−1 which is odd on the boundary S n−1 .
(5) An odd map f : S n−1 → S n−1 is not null homotopic.
(6) Let S n = F1 ∪ · · · ∪ Fn+1 with closed Fi . Then at least one Fi has antipodal points.

Proof. (1) =⇒ (2). There exists x ∈ S n with f (x) = f (−x). If f is antipodal, then f (−x) =
−f (x), hence f (x) = −f (x) and f (x) = 0.
(2) =⇒ (3). If f : S n → S n−1 ⊂ Rn is odd, then there exists x ∈ S n with f (x) = 0, a
contradiction.
n
(3) =⇒ (4). Consider hemispheres D± = {(x0 , . . . , xn ) ∈ Sn | ± x0 ≥ 0} and the projection
p: D+ → D onto the last n coordinates which is a homeomorphism. If f : Dn → S n−1 is odd on
n n

the boundary, define a new map

(
n
f p(x) x ∈ D+
g: S n → S n−1 , g(x) = n
−f p(−x) x ∈ D−

Note that if x ∈ S n−1 = D+ n

∩ D−n
, then −f p(−x) = −f (−px) = f (px) as f is odd on the
boundary. This implies that g is well-defined and continuous. It is odd by definition and this
contradicts the assumption.
(4) =⇒ (5). If f is null homotopic, consider the corresponding homotopy h: S n−1 × I → S n−1 .
It is constant on S n−1 × {1}, hence factors through the cone

h̄: CS n−1 = (S n−1 × I)/(S n−1 × {1}) → S n−1 .

But CS n−1 ' Dn with the boundary ∂Dn = S n−1 × {0}. But f is odd on S n−1 , hence h̄ is odd
on the boundary. This contradicts our assumption.
(5) =⇒ (1). Assume that f : S n → Rn is such that f (x) 6= f (−x) for all x ∈ S n . Then the map
f (x) − f (−x)
g: S n → S n−1 , x 7→
kf (x) − f (−x)k
is antipodal. Its restriction g: S n−1 → S n−1 is odd and null homotopic (as it is defined on the
upper hemisphere of S n ), a contradiction.
(1) =⇒ (6). If some of the Fi are empty, we make them 1-point sets. Consider the map

f : S n → Rn , . . . , x 7→ (d(x, F1 ), , d(x, Fn ))
40 HOMOLOGY THEORY

using the Euclidean distance d. By assumption, there exists x ∈ S n with f (x) = f (−x) = y. If
yi = 0, then d(x, Fi ) = d(−x, Fi ), hence x, −x ∈ Fi as Fi is compact. If yi = 6 0 for all i, then
Sn
x, −x ∈ / i=1 Fi , hence x, −x ∈ Fn+1 .
(6) =⇒ (3). Assume there exists an odd map f : S n → S n−1 . There exists a closed covering
F1 , . . . , Fn+1 of S n−1 such that no Fi contains antipodal points. Indeed, consider a projection
of ∂∆n onto S n−1 and take the images of the facets of ∂∆n (there are n + 1 facets). If
x, −x ∈ f −1 (Fi ), then −f (x) = f (−x) ∈ Fi , hence Fi contains antipodal points, a contradiction.
This means that the closed covering f −1 (F1 ), . . . , f −1 (Fn+1 of S n contradicts our assumption. 

Remark 4.19. In particular, if S 2 is covered by 3 closed sets F1 , F2 , F3 then at least one of

them has antipodal points. The above proof shows that there exists a covering of S 2 by 4 closed
sets such that neither of them has antipodal points.

We will prove that statement (5) of the above theorem is satisfied.

Theorem 4.20. If f : S n → S n is odd, then deg f is odd. In particular, f is not null homotopic.

Proof. Recall that f∗ : Hn (S n ) → Hn (S n ) is a multiplication by deg f . Assuming that deg f is

even, we consider homology groups with coefficients in Z2 and obtain that the map

f∗ : Hn (S n , Z2 ) → Hn (S n , Z2 )

is zero. We will show that f∗ is actually an isomorphism. Let us write Hn (X) for Hn (X, Z2 )
and Cn (X) for Cn (X, Z2 ) from now on.
Consider the projection p: S n → P n = RP n which is a 2 : 1 covering. As f : S n → S n is odd,
it induces f¯: P n → P n such that f¯p = pf .
f
Sn Sn
p p
f¯
Pn Pn

For any simplex σ: ∆k → P n there exist exactly 2 lifts σ1 , σ2 : ∆k → S n (as ∆k is simply-

connected). Considering chain complexes with coefficients in Z2 , we have an exact sequence
τ p]
0 → Ck (P n ) →
− Ck (S n ) −
→ Ck (P n ) → 0

where τ (σ) = σ1 + σ2 (note that p] (σ1 + σ2 ) = 2σ = 0 because of the coefficient ring Z2 ). The
above maps commute with differentials, hence there is a long exact sequence

τ p∗ =0
0 → Hn (P n ) −∼
∗
→ Hn (S n ) −−−→ Hn (Pn ) −∼
→ Hn−1 (P n ) → 0 → . . .
→ Hi−1 (P n ) → 0 → . . .
· · · → 0 → Hi (Pn ) −∼
0
→ H0 (P n ) →
· · · → 0 → H1 (Pn ) −∼ − H0 (S n ) −∼
→ H0 (Pn ) → 0

To see that τ∗ : Hn (P n , Z2 ) → Hn (S n , Z2 ) is an isomorphism, we represent the generator of

Hn (S n , Z) ' Z as σ1 − σ2 , where σ1 , σ2 are hemispheres in S n . They both are mapped to the
same σ ∈ Hn (P n , Z). Therefore τ∗ : Hn (P n , Z2 ) → Hn (S n , Z2 ) is surjective. We also note that
Hi (S n ) = 0 for 0 < i < n. We conclude from the diagram that ∂: Hi (P n ) → Hi−1 (P n ) are
isomorphisms for 1 ≤ i ≤ n.
 HOMOLOGY THEORY 41

Next we note that there is a commutative diagram

τ p]
0 Ck (P n ) Ck (S n ) Ck (P n ) 0
f¯ f f¯
τ p]
0 Ck (P n ) Ck (S n ) Ck (P n ) 0
It induces a commutative diagram with long exact sequences as rows. In particular, we have
τ∗ p∗
Hk (P n ) Hk (S n ) Hk (P n ) ∼ Hk−1 (P n )
f¯∗ f∗ f¯∗ f¯∗
τ p∗
Hk (P n ) Hk (S n ) Hk (P n ) ∼ Hk−1 (P n )
Starting with the fact that f¯∗ : H0 (P n ) → H0 (Pn ) is an isomorphism, we inductively prove that
f¯∗ : Hn (P n ) → Hn (Pn ) is an isomorphism. But then also f∗ : Hn (S n ) → Hn (Sn ). This is what we
wanted. 
4.7. Lefschetz fixed point theorem. Let X be a space such that all homology groups
Hn (X, Q) are finite-dimensional and only finitely many of them are non-zero. Given a continuous
map f : X → X, define the Lefschetz number
X
τ (f ) = (−1)n tr(f∗ : Hn (X) → Hn (X)).
n

Note that if f = 1 is the identity, then tr(1∗ : Hn (X) → Hn (X)) = dim Hn (X), hence
τ (1) =
X
(−1)n dim Hn (X) = χ(X),
n
the Euler number of X.
Theorem 4.21. If X is a finite simplicial complex and f : X → X satisfies τ (f ) 6= 0, then f
has a fixed point.
Proof. Assume that f : X → X does not have fixed points. We use without a proof the fact
that in this case there exists a subdivision L of X, a subdivision K of L and a simplicial map
g: K → L homotopic to f such that g(σ) ∩ σ = ∅ for all σ (we mix abstract simplices and their
geometric realizations here). Let K n ⊂ |K| = X and Ln ⊂ |L| = X be the n-skeletons. Then
Ln ⊂ K n as K is a subdivision of L. Therefore g(K n ) ⊂ K n (we use the geometric realization
of g here). It induces the chain map of the cellular chain complex Cn = Hn (K n , K n−1 ). Then
X
τ (g) = (−1)n tr(g∗ : Cn → Cn ).
n

But the map g∗ : Hn (K , K ) → Hn (K n , K n−1 ) has trace zero since Hn (K n , K n−1 ) has a
n n−1

basis consisting of n-simplices and g(σ) ∩ σ = ∅ by our assumption. We conclude that

τ (f ) = τ (g) = 0. 
Corollary 4.22 (Brouwer fixed point theorem 3.37). A continuous map f : Dn → Dn has a
fixed point.
Proof. The space Dn is homotopic to a point, hence f is homotopic to the identity map.
Therefore τ (f ) = χ(Dn ) = 1 6= 0, hence f has a fixed point. 
Lemma 4.23. Given a map f : S n → S n , we have τ (f ) = 1 + (−1)n deg f .
42 HOMOLOGY THEORY

Appendix A. Preliminaries
A.1. Relations.

Definition A.1. Let X be a set.

(1) A (binary) relation on X is a subset R ⊂ X × X. We write xRy if (x, y) ∈ R.
(2) A relation ∼ on X is called an equivalence relation if we have
(1) x ∼ x (reflexivity).
(2) x ∼ y =⇒ y ∼ x (symmetry).
(3) x ∼ y, y ∼ z =⇒ x ∼ z (transitivity).
(3) A relation ≤ on X is called a partial order if we have
(1) x ≤ x (reflexivity).
(2) x ≤ y, y ≤ x =⇒ x = y (anti-symmetry).
(3) x ≤ y, y ≤ z =⇒ x ≤ z (transitivity).
A set X equipped with a partial order is called a poset (partially ordered set). We
write x < y if x ≤ y and x 6= y.
(4) A poset X is called a chain (or a totally ordered set) if for any x, y ∈ X we have
x ≤ y or y ≤ x.
(5) A map f : X → Y between posets is called order-preserving (or monotone) if
x ≤ y =⇒ f (x) ≤ f (y).
It is called (strictly) increasing (or strictly monotone) if
x < y =⇒ f (x) < f (y).
By an increasing map we will usually mean a strictly increasing map.

Definition A.2. Given an equivalence relation ∼ on X, we define an equivalence class of

x ∈ X to be [x] = {y ∈ X | y ∼ x}. The set of all equivalence classes of X is denoted by X/∼,
called the quotient set of X by ∼.

Example A.3. Let X = [0, 1] and let x ∼ y if x = y or {x, y} = {0, 1}. Then X/∼ is a circle.

Example A.4. Let G be a group acting on a set X, that is, we have a map (called an action)
G × X → X, (g, x) 7→ gx
such that g(hx) = (gh)x and ex = x for g, h ∈ G, x ∈ X and the identity e ∈ G. Define
an equivalence relation on X to be x ∼ y if y = gx for some g ∈ G. The equivalence classes
[x] = { gx | g ∈ G} = Gx are called orbits. The quotient set X/∼ is denoted by X/G, called
the orbit space.

Definition A.5. Given a relation R ⊂ X × X, define the equivalence relation generated

by R to be the minimal equivalence relation ∼ that contains R (the intersection of equivalence
relations that contain R). We have x ∼ y ⇐⇒ there exist elements x0 , . . . , xn with x0 = x,
xn = y and xi−1 Rxi or xi Rxi−1 for 1 ≤ i ≤ n. The quotient set X/∼ is also denoted by X/R.

Remark A.6. Similarly, we can define a partial order generated by a relation R ⊂ X × X. For
its existence we require that whenever x0 Rx1 R . . . Rxn Rx0 , we have x0 = x1 = · · · = xn (so that
anti-symmetry is satisfied).

Example A.7. Given A ⊂ X, let X/A denote the quotient set X/(a ∼ b : a, b ∈ A). It can be
decomposed as X/A = (X − A) ∪ {∗}, where ∗ is the equivalence class of all elements in A.
 HOMOLOGY THEORY 43

Example A.8. For example, let Dn = {x ∈ Rn | kxk ≤ 1} and S n−1 = {x ∈ Rn | kxk = 1}.
Then S n−1 ⊂ Dn and Dn /S n−1 ' S n . In particular, S 0 = {±1} ⊂ D1 = [−1, 1] and we have
D1 /S 0 = [−1, 1]/ {±1} ' S 1 . Similarly D2 /S 1 ' S 2 .
A.2. Quotient topology. By a space we always mean a topological space.
Definition A.9. Let X be a set and τ, τ 0 ⊂ 2X be two topologies on X. If τ ⊂ τ 0 , then τ is
called weaker (coarser, smaller) than τ 0 , and τ 0 is called stronger (finer, larger) than τ .
Definition A.10. Let X be a space and f : X → Y be a surjective map onto a set Y . Define
the quotient topology on Y with respect to f to be
τ = U ⊂ Y | f −1 (U ) is open in X .


It is the strongest topology on Y such that f : X → Y is continuous. The space Y equipped

with the quotient topology is called the quotient space. A surjective map f : X → Y between
spaces is called the quotient map if Y has the quotient topology.
Lemma A.11. Let f : X → Y be a quotient map between spaces. Then a map g: Y → Z into a
space Z is continuous ⇐⇒ gf is continuous.
Example A.12. If X is a topological space equipped with an equivalence relation ∼, then
we equip X/∼ with the quotient topology with respect to the projection p: X → X/∼. In
particular, if A ⊂ X is a subspace, then we equip X/A with the quotient topology.
Example A.13. Let f : A → X and g: A → Y be two continuous maps. Define their pushout
to be the quotient space
X tA Y = X t Y /(f (a) ∼ g(a) : a ∈ A).
The maps i1 : X → X tA Y and i2 : Y → X tA Y have the following universal property. If
continuous j1 : X → Z and j2 : Y → Z satisfy j1 f = j2 g, then there exists a unique continuous
u: X tA Y → Z making the following diagram commute
g
A Y
f i2
j2
i1
X X tA Y
u

Z
j1

Example A.14. In particular, let A ⊂ Y be a subspace and φ: A → X be a continuous map.

We consider the inclusion map i: A → Y and the corresponding pushout
X ∪φ Y = X t Y /(φ(a) ∼ a : a ∈ A),
called the adjunction space (or attaching space) obtained by attaching Y to X via φ
φ
A X
i j
Φ
Y X ∪φ Y

There is a decomposition X ∪φ Y = X t (Y − A). The map φ is called the attaching map.

The map Φ is called the characteristic map.
44 HOMOLOGY THEORY

X
0

X Y Y
1

Figure 1. Cone, mapping cylinder and mapping cone.

Example A.15. Given a space X, define its cone to be the quotient space (I = [0, 1])
X ×I
CX = .
X × {0}
Example A.16. Given a continuous map f : X → Y , define its mapping cylinder to be the
quotient space
X ×I tY
Mf = .
(x, 1) ∼ f (x)
Define its mapping cone to be the quotient space
CX t Y
Cf = = Mf /X × {0} .
(x, 1) ∼ f (x)

Definition A.17. Define a pointed space (X, x0 ) to be a pair consisting of a space X and a
distinguished point x0 ∈ X. Define the wedge sum of pointed spaces (X, x0 ) and (Y, y0 ) to be
the quotient space
X ∨ Y = X t Y /(x0 ∼ y0 )
with the distinguished point [x0 ] = [y0 ].
Remark A.18. Note that if A ⊂ X is a subspace, then X/A has a distinguished point – the
equivalence class of points in A.
Remark A.19. The wedge sum can be interpreted as a coproduct in the category of pointed
spaces.
Q
Definition A.20. Let (Xi )i∈I be a family of spaces, X = i Xi and let pi : X → Xi be
projections. The product topology on X is the weakest topology that contains sets p−1
i (U )
for open U ⊂ Xi . This topology consists of (arbitrary) unions of sets
\ Y Y
p−1
i (U i ) = U i × Xi ,
i∈J i∈J i∈I\J

where J ⊂ I is finite and Ui ⊂ Xi is open, for i ∈ J.

Q
Lemma A.21. The product topology on X = i Xi is the weakest topology such that pi : X → Xi
are continuous. For any family of continuous maps fi : Z → Xi , there exists a unique continuous
map f : Z → X such that pi f = fi , for all i
fi
f Q pi
Z i∈I Xi Xi .
 HOMOLOGY THEORY 45

Appendix B. Categories and functors

Definition B.1. A category A consists of the following data
(1) A family Ob A, whose elements are called objects of A.
(2) For all objects X, Y of A, a set Hom(X, Y ) = HomA (X, Y ), whose elements are called
morphisms from X to Y .
(3) For all objects X, Y, Z of A, a map
Hom(X, Y ) × Hom(Y, Z) → Hom(X, Z), (f, g) 7→ g ◦ f,
called the composition map.
This data should satisfy
(1) ∀X ∈ Ob A, ∃1X ∈ Hom(X, X) s.t. 1Y ◦ f = f ◦ 1X = f for any f ∈ Hom(X, Y ).
(2) The composition of morphisms is associative.

Remark B.2.
(1) The element 1X is unique for every X ∈ Ob A.
(2) We write f : X → Y for f ∈ Hom(X, Y ).
(3) A morphism f : X → Y is called an isomorphism if ∃g: Y → X such that gf = 1X and
f g = 1Y .

Example B.3.
(1) The category Mod A of modules over a ring A and homomorphisms between them.
(2) The category Set of sets and all maps between them.
(3) The category Top of topological spaces and continuous maps between them.
(4) The category Com of commutative rings and ring homomorphisms.
(5) The category Grp of groups and group homomorphisms.
(6) The category A• of abelian groups and group homomorphisms. It can be identified with
Mod Z.

Definition B.4. Let A and B be two categories. A (covariant) functor F from A to B

consists of the following data
(1) A map F : Ob(A) → Ob(B).
(2) A map F : HomA (X, Y ) → HomB (F X, F Y ) for all objects X, Y ∈ Ob(A).
This data should satisfy
(1) F (1X ) = 1F X for all X ∈ Ob(A).
(2) F (g ◦ f ) = F (g) ◦ F (f ).

Definition B.5.
(1) Given a category A, we define the opposite category Aop using the data
Ob(Aop ) = Ob(A), HomAop (X, Y ) = HomA (Y, X).
(2) A functor from Aop to B is called a contravariant functor from A to B.

Example B.6.
(1) Given a category A and an object X, there is a (covariant) functor
Hom(X, −): A → Set, Y 7→ Hom(X, Y ).
There is also a contravariant functor
Hom(−, X): A → Set, Y 7→ Hom(Y, X).
46 HOMOLOGY THEORY

(2) For the category Mod A and an A-module M , we have similar functors Hom(M, −) and
Hom(−, M ) from Mod A to Mod A.
(3) For any A-module N , there is a functor
− ⊗ N : Mod A → Mod A, M 7→ M ⊗ N.
Definition B.7. Let F, G be two functors from A to B. A morphism (or natural transfor-
mation) φ from F to G consists of the data
(1) Morphism φX : F X → GX for every object X ∈ Ob(B)
such that for every f ∈ HomA (X, Y ) the following diagram commutes
F (f )
FX FY
φX φY
G(f )
GX GY
Definition B.8. Let f : A → B be a ring homomorphism.
(1) Given a B-module M , we can consider it as an A-module by setting ax = f (a)x for
a ∈ A, x ∈ M . In this way we obtain a functor Mod B → Mod A, called a restriction
of scalars.
(2) Given an A-module M , we consider a B-module
MB = B ⊗A M, b(b0 ⊗ x) = bb0 ⊗ x, b, b0 ∈ B, x ∈ M.
In this way we obtain a functor B ⊗A −: Mod A → Mod B, called an extension of
scalars.
Definition B.9. Two functors F : A → B, G: B → A are called adjoint if there exist natural
bijections
HomB (F (X), Y ) ' HomA (X, G(Y )) ∀ X ∈ Ob(A), Y ∈ Ob(B).
In this case F is called a left adjoint functor to G and G is called a right adjoint functor
to F .
Example B.10. There is a Tensor-Hom adjunction
Hom(L ⊗ M, N ) ' Hom(L, Hom(M, N )).
for A-modules L, M, N . It implies that the functors
F : Mod A → Mod A, L 7→ L ⊗ M,
G: Mod A → Mod A, N 7→ Hom(M, N )
are adjoint.
 HOMOLOGY THEORY 47

Appendix C. Simplicial approximation

A simplicial map f : K → L between two simplicial complexes is a map f : K0 → L0 such
that, for every simplex σ ∈ K (which is a subset of K0 ), the set f (σ) ⊂ L0 is a simplex in L.
Such map induces a continuous map |f | : |K| → |L| between geometric realizations.
Recall, that for every simplicial complex K, we have its barycentric subdivision N (K)
which consists of all finite non-empty chains of simplices in K. There is a homeomorphism
|N (K)| → |K|.
Definition C.1. A simplicial approximation of a map F : |K| → |L| is a simplicial map
f : K → L such that, for every simplex τ ∈ L,
x ∈ |K| , F (x) ∈ ∆(τ ) =⇒ |f | (x) ∈ ∆(τ ).
Lemma C.2. The above definition is equivalent to the requirement that, for every v ∈ K0 ,
[ 
◦ [ ◦
F ∆(σ) ⊂ ∆(τ )
σ3v τ 3f (v)
◦
Proof. Assume that f is a simplicial approximation of F . Let x ∈ ∆(σ) for some σ 3 v. Then
◦ ◦
τ = f (σ) is a simplex in L and |f | (x) ∈ ∆(τ ). By assumption, if F (x) ∈ ∆(τ 0 ) (there always
exists such simplex), then |f | (x) ∈ ∆(τ 0 ), hence τ ⊂ τ 0 . But f (v) ∈ τ , hence f (v) ∈ τ 0 .
◦
Conversely, we can assume that F (x) ∈ ∆(τ ) and we need to show that |f | (x) ∈ ∆(τ ). Let
◦ ◦
x ∈ ∆(σ) for some σ ∈ K. By assumption, for every v ∈ σ, we have F (x) ∈ ∆(τ 0 ) for some
τ 0 3 f (v). But such simplex is unique and we get τ 0 = τ , hence f (v) ∈ τ . We conclude that
f (σ) ⊂ τ , hence |f | (x) ∈ |f | (∆(σ)) ⊂ ∆(τ ).

Proposition C.3. If f : K → L is a simplicial approximation to F : |K| → |L|, then |f | is
homotopic to F .
Theorem C.4 (Simplicial approximation). For any continuous map F : |K| → |L|, there exists
r ≥ 1 and a simplicial approximation f : N r (K) → L to F .
 Index
n-dimensional cube, 3 face, 3, 10, 12
n-dimensional disc, 3 face map, 7
n-dimensional sphere, 3 facet, 4
∆-set, 5 functor
contravariant, 45
action of a group, 42 covariant, 45
acyclic chain complex, see chain complex left adjoint, 46
adjoint functors, 46 right adjoint, 46
adjunction space, 43
affinely independent, 12 geometric realization, 8
attaching map, 33, 43 good pair, 30
attaching space, 43 group of chains, 15
augmented chain complex, 24
homology class, 19
barycenter, 14 homology group, 15, 19
barycentric coordinates, 11 cellular, 34
barycentric subdivision, 14 reduced, 24
boundary, 15, 19 relative, 27
boundary operator, 15, 19, 23 singular, 23
homotopic, 25
category, 45 homotopic chain maps, 20
opposite, 45 homotopy, 25
cell, 4, 33 homotopy equivalence, 20, 25
cell complex, 33 homotopy equivalent, 25
chain, 15, 19, 42 homotopy equivalent complexes, 20
chain complex, 15, 19 homotopy inverse, 20
acyclic, 19
increasing, 42
cellular, 34
exact, 19 Lefschetz number, 41
chain homotopy, 20 local degree, 35
chain map, 19
characteristic map, 33, 43 mapping cone, 44
coface, 4 mapping cylinder, 44
coface map, 7
cone, 44 natural transformation, 46
contractible, 20, 25 nerve, 7, 10
CW-complex, 33 null-homotopic, 20, 25
CW-decomposition, 33 odd map, 39
CW-space, 33 orbit, 42
cycle, 15, 19 orbit space, 42
order complex, 10
deformation retract, 30
order-preserving, 42
deformation retraction, 30
ordered complex, 7
degree, 35
differential, 15, 19, 23 pair, 27
partial order, 42
equivalence class, 42 partially ordered set, 42
equivalence relation, 42 pointed space, 44
generated by, 42 poset, 42
Euler characteristic, 2, 18 power set, 10
Euler’s formula, 2 product topology, 44
exact chain complex, see chain complex pushout, 43
exact sequence, 19
extension of scalars, 46 quotient map, 43
48
 HOMOLOGY THEORY 49

quotient set, 42
quotient space, 43
quotient topology, 43

relation, 42
restriction of scalars, 46
retract, 30
retraction, 30

semi-simplicial set, 5
short exact sequence, 19
simplex, 12
dimension, 10
of a ∆-set, 5
open, 4
singular, 23
standard, 3
simplicial approximation, 47
simplicial complex, 10
geometric, 12
ordered, 10
simplicial homology group, 15
simplicial map, 47
simplicial morphism, 9
simplicial polyhedron, 11
singular ∆-set, 23
singular chain, 23
singular chain complex, 23
skeleton, 33
strong deformation retract, 30
strong deformation retraction, 30

tangent space, 36
topological manifold, 38
topology
stronger, 43
weaker, 43
totally ordered set, 42
triangulated space, 4
triangulation, 4

underlying space, 12

vector field, 36
vertex, 3, 12
vertex map, 5

wedge sum, 44