ALGEBRAIC TOPOLOGY II

(Simplicial Homology)

BY

A. DEB RAY

DEPARTMENT OF PURE MATHEMATICS

UNIVERSITY OF CALCUTTA
35, Ballygunge Circular Road
KOLKATA - 700019, INDIA
e-mail : [email protected]
Chapter 1

Simplicial Complex and Simplicial

Homology groups

1.1 Geometric Simplex and Simplicial Complex

Algebraic Topology provides various techniques for associating algebraic objects to topolog-
ical spaces. In Algebraic Topology I, Homotopy Theory was introduced and studied. We
shall now see how a sequence of groups, known as homology groups, can be associated to a
given topological space. For doing so, we demand the space to be a ‘nice’ one. Of course, we
need to formalize the term ‘nice’ and see that there are plenty of such nice spaces around.
In this Chapter, we build the basics for developing simplicial homology theory.

Throughout this Chapter, we consider subsets of Euclidean n-space, Rn , unless stated oth-
erwise.

Definition 1.1.1. A set of (k + 1) points {a0 , a1 , . . . , ak } in Rn (k ≥ 2) is said to be

geometrically independent if {a1 − a0 , a2 − a0 , . . . , ak − a0 } is linearly independent in Rn . By
definition, any singleton is geometrically independent.

It is easy to see that the above definition does not depend on the chosen point a0 . i.e.,
{a1 −a0 , a2 −a0 , . . . , ak −a0 } is linearly independent if and only if {aj −ai : 0 ≤ j ≤ k, j ̸= i}
is linearly independent for all i = 0, 1, . . . k.

Any one dimensional subspace of Rn is represented by a straight line passing through

origin. So, any straight line in Rn is actually a translate of the one dimensional subspace
parallel to it. Using Mathematical terms, any straight line in Rn is a coset of the form x + V 1
where x ∈ Rn and V 1 is a one dimensional subspace of Rn . By similar arguments, a plane
in Rn is a coset of the form x + V 2 where x ∈ Rn and V 2 is a two dimensional subspace of
Rn . The following definition generalizes this idea :

Definition 1.1.2. A k-dimensional plane H k (k ≤ n) in Rn is a coset of the form x + V k ,

where x ∈ Rn and V k is a k-dimensional subspace of Rn . i.e., H k = x + V k , for some x ∈ Rn .

Observation 1.1.1.
1
A. Deb Ray, Pure Math, C.U. 2

ˆ Any set {a0 , a1 } is always geometrically independent.

ˆ A set {a0 , a1 , a2 } is geometrically independent if and only if the points a0 , a1 and a2

are non-collinear. i.e., There is no one dimensional plane containing these points.

ˆ A set {a0 , a1 , a2 , a3 } is geometrically independent if and only if the points a0 , a1 , a2 and

a3 are non-coplanar. i.e., There is no two dimensional plane containing these points.

The Observation 1.1.1 leads to a characterization of geometrically independent set of

points in terms of k-dimensional planes :

Theorem 1.1.1. A set of (k + 1) points {a0 , a1 , . . . , ak } (k ≥ 2) in Rn is geometrically

independent if and only if there does not exist any (k − 1)-dimensional plane containing the
set.

Proof. Suppose {a0 , a1 , . . . , ak } (k ≥ 2) is geometrically independent in Rn . If possible let

there be a (k − 1)-dimensional plane H k−1 containing the set. Since H k−1 = x + V k−1 , for
some x ∈ Rn and some (k − 1)-dimensional subspace V k−1 of Rn , each ai − x ∈ V k−1 , for
i = 0, 1, . . . k. Then ai −a0 = (ai −x)−(a0 −x) ∈ V k−1 , for i = 1, 2, . . . k, which is not possible.
Because, the definition of geometrically independent set says that {ai − a0 : i = 1, 2, . . . k}
is a linearly independent set of k vectors.
Conversely, let {a0 , a1 , . . . , ak } (k ≥ 2) be geometrically dependent in Rn . So, A = {a1 −
a0 , a2 − a0 , . . . , ak − a0 } is linearly dependent. Then the dimension of the linear span of A
is at most k − 1. Hence one may obtain a vector space V k−1 of dimension k − 1 containing
A. i.e., ai − a0 ∈ V k−1 , for each i = 1, 2, . . . k. Consequently, ai ∈ a0 + V k−1 , for each
i = 1, 2, . . . k and a0 = a0 + 0 ∈ a0 + V k−1 . Hence there exists a (k − 1) dimensional plane
H k−1 = a0 + V k−1 that contains the given set.
In the process of proving Theorem 1.1.1, we feel the existence of a k-dimensional plane
containing (k + 1) geometrically independent points. Next theorem shows that not only it
exists, it exists uniquely.

Theorem 1.1.2. If {a0 , a1 , . . . , ak } (k ≥ 2) is a set of (k + 1) geometrically independent

points in Rn then there exists a unique k-dimensional plane containing the set.

Proof. Suppose {a0 , a1 , . . . , ak } (k ≥ 2) is geometrically independent in Rn . Then the set

S = {a1 − a0 , a2 − a0 , . . . , ak − a0 } is linearly independent. So, the linear span of S is a
vector space of dimension k, say V k , that contains S. Therefore, H k = a0 + V k is indeed a
k-dimensional plane containing {a0 , a1 , . . . , ak }.
Let x+V k and y+W k , where x, y ∈ Rn , be two k-dimensional planes containing {a0 , a1 , . . . , ak }.
Then ai − a0 = (ai − x) − (a0 − x) ∈ V k and ai − a0 = (ai − y) − (a0 − y) ∈ W k , for each
i = 1, 2, . . . k. Since S is a linearly independent set of k vectors, it forms a basis of both V k
and W k . This leads to V k = W k . Hence, x + V k and y + V k are two cosets of the same sub-
space V k . It is well known that two cosets are either disjoint or equal. Since S ⊆ V k ∩ W k ,
x + V k = y + V k.
A set of (k + 1) geometrically independent points can be used as a reference frame for
representing points on the k-dimensional plane determined by the set.
A. Deb Ray, Pure Math, C.U. 3

Theorem 1.1.3. If H k is the k-dimensional plane containing a geometrically independent

set {a0 , a1 , . . . ak }, then for any x ∈ H k , there exist unique λ0 , λ1 , . . . , λk ∈ R such that
k
P Pk
x= λi ai and λi = 1.
i=0 i=0

Proof. Without loss of generality, we assume that H k = a0 + V k , where V k is the k-

dimensional subspace of Rn generated by the linearly independent set {a1 −a0 , a2 −a0 , . . . ak −
k
a0 }. Therefore, any vector in V k is a linear combination
P
λi (ai − a0 ), where λi ∈ R, for
i=1
each i = 1, 2, . . . k. Hence, any x ∈ H k is of the form
k
X
x = a0 + λi (ai − a0 )
i=1
k k
!
X X
= a0 + λ i ai − λi a0
i=1 i=1
k
! k
X X
= 1− λi a0 + λi ai
i=1 i=1
k
X
= λ i ai
i=0
 k
 k
P P
where λ0 = 1 − λi ⇒ λi = 1. Suppose, λi , ηi ∈ R, i = 0, 1, 2, . . . , k such that
i=1 i=0
k
P Pk k
P k
P
x= λi ai = i=0 ηi ai and λi = 1 = ηi . Then
i=0 i=0 i=0

k
X
(λi − ηi )ai = 0 (1.1)
i=0

k
X
(λi − ηi ) = 0 (1.2)
i=0
k
P
From (1.2), (λi − ηi )a0 = 0 and so, subtracting it from (1.1),
i=0

k
X
(λi − ηi )(ai − a0 ) = 0.
i=1

The linear independence of the set {a1 − a0 , a2 − a0 , . . . ak − a0 } implies that λi = ηi , for each
i = 1, 2, . . . k. Using (1.2), λ0 = η0 .
For each x ∈ H k , the k scalars (λ0 , λ1 , . . . , λk ) representing x are called barycentric
co-ordinates of x with respect to the geometrically independent set {a0 , a1 , . . . ak }. The
portion of the k-dimensional plane which contains points with non-negative barycentric co-
ordinates, is called a k-simplex.
A. Deb Ray, Pure Math, C.U. 4

Definition 1.1.3. If {a0 , a1 , . . . ak } is a geometrically independent set in Rn then the set

( k k
)
X X
σk = λ i ai : λi = 1, λi ∈ R, λi ≥ 0, i = 0, 1, . . . , k
i=0 i=0

is called a k-simplex, usually denoted by σ k = ⟨a0 , a1 , . . . ak ⟩. Each ai is called a vertex of

the k-simplex.

Definition 1.1.4. For a k-simplex σ k , the set

( k k
)
X X
λ i ai : λi = 1, λi ∈ R, λi > 0, i = 0, 1, . . . , k
i=0 i=0

is called the interior of σ k , denoted by int(σ k ).

Definition 1.1.5. An m-simplex (m ≤ k) τ is called a face or m-face of σ k , if the vertex
set of τ is a subset of the vertex set of σ k .
With the concepts developed so far, we define a simplicial complex as follows :
Definition 1.1.6. Let K be a collection of finitely many simplexes. K is called a (finite)
simplicial complex, if the following two conditions hold:
1. σ ∈ K and τ is a face of σ ⇒ τ ∈ K;
2. σ, τ ∈ K and σ ∩ τ ̸= ∅ ⇒ σ ∩ τ is a common face of σ and τ .
One may note that the figure on the left represents a simplicial complex while the figure
on the right does not.
A. Deb Ray, Pure Math, C.U. 5

It is to note that a simplicial complex is a set of simplexes, not points ofSRn . But if one
takes the union of all the points lying on the simplexes of K, i.e., |K| = σ, then |K| is
σ∈K
a subspace of the Euclidean space Rn . This space |K| is called the geometric carrier or
underlying space of K.
Observation 1.1.2. ˆ |K|, being a subspace of Rn , is a metric space.

ˆ Since any simplex is a closed and bounded subset of an Euclidean space it is compact.
|K|, being a finite union of compact sets, is also compact.
Definition 1.1.7. If σ = ⟨a0 , a1 , . . . , am ⟩ is a m-simplex then cl(σ) (read as, closure of σ)
is the simplicial complex consisting of all possible faces of σ and the skeleton of σ (denoted
by sk(σ)) is cl(σ) \ {σ}.
Exercise 1.1.1. Show that both closure and skeleton of a simplex σ are simplicial complexes.

1.2 Orientation of a simplicial complex

We begin this section by discussing possible definition of ‘orientation’ that comes naturally
in our mind.
Illustration 1.2.1. Suppose σ 1 = ⟨v0 , v1 ⟩ is a 1-simplex. The geometric carrier of σ 1 is a
line segment between v0 and v1 . Of course, there are two possible directions one may think
of : v0 to v1 and v1 to v0 . It is quite expected that if v0 to v1 is considered as positive then
the other direction, i.e., v1 to v0 is negative. Consequently, we may write +σ 1 = ⟨v0 , v1 ⟩. In
this case it is clear enough that ⟨v1 , v0 ⟩ is −σ 1 .

Illustration 1.2.2. Suppose σ 2 = ⟨v0 , v1 , v2 ⟩ is a 2-simplex. The geometric carrier of σ 2 is

the region, on and inside the triangle with vertices v0 , v1 and v2 . Our intuition suggests that
there are two possible orientations : anticlockwise and clockwise. If anticlockwise direction
is chosen as positive, then +σ 2 = ⟨v0 , v1 , v2 ⟩ = ⟨v1 , v2 , v0 ⟩ = ⟨v2 , v0 , v1 ⟩. Then clearly,
−σ 2 = ⟨v1 , v0 , v2 ⟩ = ⟨v0 , v2 , v1 ⟩ = ⟨v2 , v1 , v0 ⟩.
A. Deb Ray, Pure Math, C.U. 6

Illustration 1.2.3. Suppose σ 3 = ⟨v0 , v1 , v2 , v3 ⟩ is a 3-simplex. The geometric carrier of σ 3

is the region, on and inside the tetrahedron with vertices v0 , v1 , v2 and v3 . Now it is difficult
to illustrate the possible orientations of σ 3 , as there exist more than two degrees of freedom
for rotating the tetrahedron!

In fact, for any m-simplex with m ≥ 4, we are unable to visualize the geometry of
the simplex σ m . Hence, we need to formulate something algebraically without involving
the geometry of the said simplex. Also, the definition should be a generalization of the
aforementioned natural concept of orientation.
Before defining it formally, we introduce the ordering of vertices as follows:

v0 < v1 < ... < vm if and only if ⟨v0 , v1 , ..., vm ⟩ is a positively oriented simplex

Definition 1.2.1. Suppose σ m = ⟨v0 , v1 , ..., vm ⟩ is any m-simplex and v0 < v1 < ... < vm is
an ordering of vertices associated with it. These (m+1) vertices can be permuted in (m + 1)!
ways. Among these permutations, half are even and half are odd. If Sm+1 denotes the
symmetric group on the set {0, 1, ..., m} of m + 1 symbols, then for any τ ∈ Sm+1 ,
vτ (0) , vτ (1) , ..., vτ (m) is positively oriented, if τ is an even permutation;
vτ (0) , vτ (1) , ..., vτ (m) is negatively oriented, if τ is an odd permutation.

Looking back to the illustrations described above, we find that this definition works! One
needs to be very careful about the fact that orientation of a k-simplex is determined from
the ordering of its k + 1 vertices, not induced from the orientation of some k + 1 simplex of
which it is a face. The following serves as a working rule for determining the orientation of
an m-simplex,
If ⟨a0 , a1 , . . . , am ⟩ is positively oriented then the orientation of σ formed by vertices a0 , a1 , . . . , am
is positively (respectively, negatively) oriented if the number of interchange of vertices that
takes σ back to ⟨a0 , a1 , . . . , am ⟩ is even (respectively, odd).

Definition 1.2.2. A finite simplicial complex K is said to be an oriented simplicial complex,

if all its simplexes are oriented.

Definition 1.2.3. Suppose σ q and σ q+1 are any two simplexes in an oriented simplicial
complex K of dimensions q and q + 1 respectively. Let +σ q = ⟨v0 , v1 , . . . , vq ⟩ and v be the
extra vertex of σ q+1 . The incidence number is an integer associated to the pair σ q and σ q+1 ,
denoted by [σ q+1 , σ q ] and defined by

0,
 if σ q is not a face ofσ q+1
[σ q+1 , σ q ] = 1, if + σ q+1 = ⟨v, v0 , v1 , . . . , vq ⟩

−1, if − σ q+1 = ⟨v, v0 , v1 , . . . , vq ⟩


Example 1.2.1. Let K = cl (⟨v0 , v1 , v2 ⟩). If σ 2 = ⟨v0 , v1 , v2 ⟩, σ11 = ⟨v1 , v2 ⟩, σ21 = ⟨v0 , v2 ⟩,
and σ 0 = ⟨v0 ⟩, then [σ 2 , σ11 ] = 1, [σ 2 , σ21 ] = −1, [σ11 , σ 0 ] = 0 and [σ21 , σ 0 ] = −1.

We deduce a basic result that is required in the next section.

A. Deb Ray, Pure Math, C.U. 7

Theorem 1.2.1. In an oriented simplicial complex K, if σ q−1 is a (q − 1)-face of σ q+1 , then

X
[σ q+1 , σ q ][σ q , σ q−1 ] = 0.
σ q ∈K

Here the summation is taken over all q-simplexes of K.

Proof. Let +σ q−1 = ⟨v0 , v1 , . . . , v q−1 ⟩ and a, b be the extra vertices of σ q+1 such that +σ q+1 =
⟨a, b, v0 , v1 , . . . , v q−1 ⟩. The value of each product term [σ q+1 , σ q ][σ q , σ q−1 ] = 0, if either σ q
is not a face of σ q+1 or σ q−1 is not a face of σ q . So the only terms that survive in the
above summation are due to those σ q ∈ K which are faces of σ q+1 as well as contain σ q−1
as a face. Therefore, there are only two possibilities for σ q : σ1q = ⟨a, v0 , v1 , . . . , vq−1 ⟩ and
σ2q = ⟨b, v0 , v1 , . . . , vq−1 ⟩. Now consider the following cases :
Case 1. +σ1q = ⟨a, v0 , v1 , . . . , vq−1 ⟩ and +σ2q = ⟨b, v0 , v1 , . . . , vq−1 ⟩
[σ q+1 , σ1q ][σ1q , σ q−1 ] + [σ q+1 , σ2q ][σ2q , σ q−1 ] = (−1)(1) + (1)(1) = 0.
Case 2. +σ1q = ⟨a, v0 , v1 , . . . , vq−1 ⟩ and −σ2q = ⟨b, v0 , v1 , . . . , vq−1 ⟩
[σ q+1 , σ1q ][σ1q , σ q−1 ] + [σ q+1 , σ2q ][σ2q , σ q−1 ] = (−1)(1) + (−1)(−1) = 0.
Case 3. −σ1q = ⟨a, v0 , v1 , . . . , vq−1 ⟩ and +σ2q = ⟨b, v0 , v1 , . . . , vq−1 ⟩
[σ q+1 , σ1q ][σ1q , σ q−1 ] + [σ q+1 , σ2q ][σ2q , σ q−1 ] = (1)(−1) + (1)(1) = 0.
Case 4. −σ1q = ⟨a, v0 , v1 , . . . , vq−1 ⟩ and −σ2q = ⟨b, v0 , v1 , . . . , vq−1 ⟩
[σ q+1 , σ1q ][σ1q , σ q−1 ] + [σ q+1 , σ2q ][σ2q , σ q−1 ] = (1)(−1) + (−1)(−1) = 0.

1.3 Chain complex and Simplicial Homology groups

Throughout our discussion we deal with only finite simplicial complexes and so, for each
dimension q there are at most finitely many q-simplexes in K. Let S q denote the collection
of all q-simplexes of K, where 0 ≤ q ≤ dim(K). Again, each q-simplex σ q can be equipped
with two possible orientations, +σ q and −σ q . Considering +σ q and −σ q as two different
objects, we denote the collection of all oriented q-simplexes as S˜q . Certainly, the number of
elements in S˜q is twice that of S q , when q ̸= 0. Since conventionally, 0-simplexes are always
positively oriented, |S 0 | = |S˜0 |.
Definition 1.3.1. Let K be a finite simplicial complex and 0 ≤ q ≤ dim(K). For q > 0,
a map f : S˜q → Z is called a q-chain, if f (−σ q ) = −f (σ q ), for each σ q ∈ S˜q . Any map
f : S˜0 → Z is called a 0-chain. The collection of all q-chains is denoted by the set Cq (K),
for 0 ≤ q ≤ dim(K).
Since Cq (K) consists of certain Z-valued functions, we define + on Cq (K) in usual way,
i.e., for all σ q ∈ S˜q
(f + g)(σ q ) = f (σ q ) + g(σ q ) (1.3)
It is easy to check that (Cq (K), +) is an abelian group, for 0 ≤ q ≤ dim(K).
A. Deb Ray, Pure Math, C.U. 8

Definition 1.3.2. The group (Cq (K), +) is called the q-dimensional chain group of K,
for 0 ≤ q ≤ dim(K). It is customary to define the chain groups as trivial for all q < 0 or
q > dim(K).

We now observe that the following are certain distinguished members of q-dimensional
chain group (Cq (K), +), corresponding to each positively oriented σ q ∈ K.

q q
+1, if τ = σ

(σ q )(τ q ) = −1, if τ q = −σ q

0, otherwise


From the following theorem we now see that for complete description of the chain group
Cq (K), it is enough to know that these particular q-chains, known as elementary q-chains.
In other words, elementary q-chains generate the q-th chain group.

Theorem 1.3.1. For 0 ≤ q ≤ dim(K), Cq (K) = ⊕Zσ q .

Proof. Consider any f ∈ Cq (K). Since K is a finite simplicial complex, we may list the
positively oriented q-simplexes of K as σ1q , σ2q , ... , σtqP
. Let f (σiq ) = ni , i = 1, 2, ..., t. Then
q t q
by definition
Pt of a q-chain, f (−σi ) = −ni . Choose i=1 ni σ i ∈ Cq (K). We claim that
f = i=1 ni σ qi .
For any τ q ∈ S˜q , there are two possibilities:
(i) τ q is positively oriented. In such case, τ q = σjq for some j. Therefore,

t
!
X
ni σ qi (τ q ) = nj = f (τ q ).
i=1

(ii) τ q is negatively oriented. In such case, −τ q = σjq for some j. Therefore,

t
!
X
ni σ qi (τ q ) = −nj = f (τ q ).
i=1

Hence, f = ti=1 ni σ qi . Arbitrariness of f proves the claim.

P
Pt q Pt q
To prove
Pt the uniqueness of the description of f . If possible, let f = i=1 n i σ i = i=1 mi σ i .
t
Then i=1 (ni −mi )σ qi = 0. In particular, for each i = 1, 2, · · · , t, q q
P

i=1 (ni − mi )σ i (σi ) = 0
⇒ ni − mi = 0. i.e., ni = mi , for all i = 1, 2, · · · t.

Remark 1.3.1. Theorem 1.3.1 shows that Cq (K) is a direct sum of infinite cyclic groups
generated by the elementary q-chains. Since upto isomorphism, there is only one infinite
cyclic group, e.g., Z, Theorem 1.3.1 tells us that the q-th chain group is a direct sum of as
many copies of Z as the number of positively oriented q-simplexes in K.

Remark 1.3.2. Since there is a one-one correspondence σ q ↔ σ q , we may re-write Cq (K)

as Cq (K) ∼= ⊕Zσ q , thereby visualizing the elements of Cq (K) as a Z-linear combination of
positively oriented q-simplexes. So, in the rest of our discussion, we consider Cq (K) = ⊕Zσ q .
A. Deb Ray, Pure Math, C.U. 9

Definition 1.3.3. For 0 ≤ q ≤ dim(K), the q-th boundary map is the map ∂q : Cq (K) →
Cq−1 (K) defined on the generators σ q by
X
∂q (σ q ) = [σ q , σiq−1 ]σiq−1

(where the summation is taken over all positively oriented (q − 1)-faces of σ q ), and extended
to the entire chain group by linearity. i.e.,
q
t t
!
X X X
∂q ( ni σiq ) = ni [σ q , σiq−1 ]σiq−1
i=1 i=1 i=0

It is easy to see that ∂q as defined in Definition 1.3.3 is a homomorphism and so, they
are known as boundary homomorphisms. If q < 0 or q > dim(K), then ∂q are taken as
trivial homomorphisms.
Remark 1.3.3. It is to be noted that there are q + 1 vertices in a q-simplex and so, a (q-
1)-face can be obtained by deleting one vertex from the list of q + 1 vertices. Consequently,
the number of (q-1)-faces of a q simplex is q + 1 and hence the summation in Definition 1.3.3
varies from i = 0 to i = q. But one may take the sum over all possible (q − 1) simplexes of
K instead of (q − 1)-faces of σ q , because, for the rest of the summands, the incidence number
will be 0, thereby contributing nothing extra.
So far, we have constructed a doubly infinite sequence of chain groups and boundary
homomorphisms, called a chain complex of K:
∂q+2 ∂q+1 ∂q ∂q−1
· · · → Cq+1 (K) → Cq (K) → Cq−1 (K) → · · ·
As discussed earlier, only a finite portion of this sequence (0 ≤ q ≤ dim(K)) is non-trivial.
A pivotal result on composition of boundary maps follows from Theorem 1.2.1 :
Theorem 1.3.2. ∂q ◦ ∂q+1 : Cq+1 (K) → Cq (K) is trivial homomorphism. i.e., ∂q ◦ ∂q+1 = 0,
for each q.
Proof. It is enough to show that (∂q ◦ ∂q+1 )(σ q+1 ) = 0 for each generator σ q+1 of Cq+1 (K).
 
X
(∂q ◦ ∂q+1 )(σ q+1 ) = ∂q  [σ q+1 , σiq ]σiq 
σiq ∈K
X
= [σ q+1 , σiq ]∂q (σiq )
σiq ∈K
X X
= [σ q+1 , σiq ] [σiq , σjq−1 ]
σiq ∈K σjq−1 ∈K
 
X X
=  [σ q+1 , σiq ][σiq , σjq−1 ]
σjq−1 ∈K σiq ∈K

= 0
Hence by using linearity of boundary homomorphisms the result follows.
A. Deb Ray, Pure Math, C.U. 10

Definition 1.3.4. Let K be an oriented simplicial complex.

ˆ A q-chain zq ∈ Cq (K) is called a q-cycle, if zq ∈ Ker(∂q ).

ˆ A q-chain bq ∈ Cq (K) is called a q-boundary, if bq ∈ Im(∂q+1 )

As Cq (K) is an abelian group, Zq (K) = Ker(∂q ) and Bq (K) = Im(∂q+1 (K)) are normal
subgroups of Cq (K), known as the group of q-cycles and the group of q-boundaries re-
spectively.

Theorem 1.3.3. For each q, Bq (K) ⊆ Zq (K).

Proof. bq ∈ Bq (K) ⇒ ∃aq+1 ∈ Cq+1 (K) such that bq = ∂q+1 (aq+1 ) ⇒ ∂q (bq ) = ∂q (∂q+1 (aq+1 )) =
(∂q ◦ ∂q+1 )(aq+1 ) = 0 (Using Theorem 1.3.2)
∂q+2 ∂q+1 ∂q ∂q−1
Definition 1.3.5. Let · · · → Cq+1 (K) → Cq (K) → Cq−1 (K) → · · · be the chain complex
of an oriented simplicial complex K. Then
(
Zq (K)/Bq (K), 0 ≤ q ≤ dim(K)
Hq (K) =
0, otherwise

is called the q-th simplicial homology group of K.

Chapter 2

Computation of Simplicial Homology

Groups

2.1 Simplicial maps and Induced Homomorphisms

Our aim is to produce a homomorphism between the homology groups, for every q ∈ Z,
corresponding to a simplicial map between the simplicial complexes. For that purpose, we
define a simplicial map and see how it induces homomorphisms between the chain groups
for every q ∈ Z.

Definition 2.1.1. Let K and L denote two simplicial complexes. A function f : K → L

which maps vertices of K to vertices of L is said to be a simplicial map if whenever
⟨v0 , v1 , . . . , vq ⟩ is a simplex in K, f (v0 ), ..., f (vq ) are vertices of a simplex in L. i.e.,
⟨f (v0 ), ..., f (vq )⟩ is a simplex in L after dropping the repetition (if any).

A simplicial map carries a simplex of K to a simplex of L. To be more precise, for every

k, a simplicial map takes a k-simplex of K to a l-simplex of L, for some l ≤ k. In what
follows, K and L denote simplicial complexes without further mention.

Definition 2.1.2. Let f : K → L be a simplicial map. For each q ∈ Z with 0 ≤ q ≤ dim(K),

a map f# : Cq (K) → Cq (L) is given by
(
⟨f (v0 ), f (v1 ), . . . , f (vq )⟩ , if all f(vi )′ s are distinct
f# (⟨v0 , v1 , . . . , vq ⟩) =
0, otherwise

for any q-simplex ⟨v0 , v1 , . . . , vq ⟩ of K and extending the definition linearly on entire K. i.e.,
!
X X
f# ni σiq = ni f# (σiq )
i i

For q < 0 or q > dim(K), f# is defined as the zero map. The collection of maps {f# :
Cq (K) → Cq (L) : q ∈ Z} is called a chain map induced by f .
11
A. Deb Ray, Pure Math, C.U. 12

It is an easy task to verify that these maps are homomorphisms, at every level q. So, we
get a diagram as follows:

∂qK
... / Cq (K) / Cq−1 (K) / ...

f# f#
 ∂qL 
... / Cq (L) / Cq−1 (L) / ...

Each rectangle in this diagram is commutative, as seen next.

Theorem 2.1.1. Let f : K → L be a simplicial map. Then

f# ◦ ∂qK = ∂qL ◦ f#

where ∂qK : Cq (K) → Cq−1 (K) and ∂qL : Cq (L) → Cq−1 (L) are the corresponding boundary
homomorphisms.

Proof. For simplicity of notation, we may sometimes use ∂ to denote both ∂qK and ∂qL and
it will be understood from the context which we are referring to. Also, it is enough to prove
the result on any generator

σ of Cq (K).
So, consider any σ = ⟨v0 , v1 , . . . , vq ⟩ from Cq (K).
To show that f# ◦ ∂qK (σ) = ∂qL ◦ f# (σ).
Two possibilities may occur : either all f (vi )’s are distinct or they are not all distinct. In
any case, the dimension of τ = ⟨f (v0 ), f (v1 ), . . . , f (vq )⟩ is ≤ q. We consider three possible
cases separately:
Case 1 : dimension of τ is q.
Case 2 : dimension of τ is q − 1.
Case 3 : dimension of τ is ≤ q − 2.
Now we see that in Case 2 and 3,

∂qL ◦ f# (σ) = ∂qL (0) = 0.

In Case 1,

∂qL ◦ f# (σ) = ∂qL (⟨f (v0 ), f (v1 ), . . . , f (vq )⟩)

q D E
X
= (−1)i f (v0 ), f (v1 ), . . . , f[ (vi ), . . . , f (vq )
i=0
q
!
X
= f# (−1)i ⟨v0 , v1 , . . . , vbi , . . . , vq ⟩
i=0
f# ◦ ∂qK (σ)


=

Hence, in case 1, the equality holds.

Case 3 : Suppose, dimension of τ ≤ q − 2.
Then either one can find three vertices of σ, say vk , vs , vm , such that f (vk ) = f (vs ) = f (vm )
or there are at least two pairs of vertices, say, vk , vs and vt , vm such that f (vk ) = f (vs )
and f (vt ) = f (vm ). In both the cases we have f# (⟨v0 , v1 , . . . , vq ⟩) = 0. Consequently,
A. Deb Ray, Pure Math, C.U. 13

∂qL ◦ f# (⟨v0 , v1 , . . . , vq ⟩) = 0. Therefore, the equality holds for this case too.
Case 2 : Suppose, dimension of τ = q − 1.
This can happen only when there exists a pair of vertices vi and vj such that f (vi ) = f (vj ).
Without loss of generality assume that i < j. Then

f# ◦ ∂qK (σ)


q
!
X
= f# (−1)i ⟨v0 , v1 , . . . , vbi , . . . , vq ⟩
i=0
D E
= (−1)i f (v0 ), f (v1 ), . . . , f[
(vi ), . . . , f (vj ), . . . f (vq )
D E
j [
+ (−1) f (v0 ), f (v1 ), . . . , f (vi ), . . . , f (vj ), . . . , f (vq )
D E
= (−1)i (−1)j−(i+1) f (v0 ), f (v1 ), . . . , f (vj ), . . . , f[ (vi ), . . . f (vq )
D E
+ (−1)j f (v0 ), f (v1 ), . . . , f (vi ), . . . , f[
(vj ), . . . , f (vq )

Here f (vi ) = f (vj ). Since (−1)j and (−1)j−1 are always of opposite sign, the above sum
= 0. Consequently, in this case also, the equality holds.
Corollary 2.1.1. If f# : Cq (K) → Cq (L) is the chain map induced from a simplicial map
f : K → L then
(i) f# (Zq (K)) ⊆ Zq (L)
(ii) f# (Bq (K)) ⊆ Bq (L)
Proof. (i) Let z ∈ Zq (K). Then ∂(z) = 0 and hence (f# ◦ ∂)(z) = 0 ⇒ (∂ ◦ f# )(z) = 0
⇒ ∂(f# (z)) = 0, i.e., f# (z) ∈ Zq (L).

(ii) Let b ∈ Bq (K). Then there exists some a ∈ Cq+1 (K) such that ∂(a) = b. So,
f# (b) = f# (∂(a)) = ∂(f# (a)) ∈ Bq (L).
Using Corollary 2.1.1, we define a map f∗ : Hq (K) → Hq (L) by f∗ (z + Bq (K)) =
f# (z) + Bq (L). This map is well defined, because,

z1 + Bq (K) = z2 + Bq (K) ⇒ z1 − z2 ∈ Bq (K)

⇒ f# (z1 − z2 ) ∈ Bq (L)
⇒ f# (z1 ) + Bq (L) = f# (z2 ) + Bq (L)

Also, f∗ is a homomorphism:

f∗ ((z1 + Bq (K)) + (z2 + Bq (K))) = f∗ ((z1 + z2 ) + Bq (K))

= f∗ (z1 + z2 ) + Bq (L)
= f∗ (z1 + Bq (L)) + f∗ (z2 + Bq (L))

Definition 2.1.3. For any simplicial map f : K → L, at every q ∈ Z with 0 ≤ q ≤ dim(K),

the homomorphism f∗ : Hq (K) → Hq (L) given by f∗ (z + Bq (K)) = f# (z) + Bq (L) is called
the homomorphism induced by f .
A. Deb Ray, Pure Math, C.U. 14

Theorem 2.1.2. (i) For each q, 0 ≤ q ≤ dim(K), if f# : Cq (K) → Cq (L) and g# : Cq (L) →
Cq (M ) are the chain maps then g# ◦ f# = (g ◦ f )# .
(ii) For each q, 0 ≤ q ≤ dim(K), if I# : Cq (K) → Cq (K) is the map induced from the
identity map I : K → K, then I# is the identity homomorphism.
Proof. It is enough to prove the results on the generators σ ∈ Cq (K).
(i) Let f : K → L and g : L → M be simplicial maps that induce chain maps f# :
Cq (K) → Cq (L) and g# : Cq (L) → Cq (M ). Since g ◦ f : K → M is a simplicial map,
(g ◦ f )# : Cq (K) → Cq (M ) is also a chain map. Let σ = ⟨v0 , v1 , . . . , vq ⟩. Then three
possibilities may arise :
Case 1. All f (v0 ), . . . f (vq ) are not distinct. Then of course, g(f (v0 )), . . . g(f (vq )) are not
distinct too. So,

(g ◦ f )# (σ) = 0
= g# (0)
= (g# ◦ f# )(σ)

Case 2. All f (v0 ), . . . f (vq ) are distinct, but g(f (v0 )), . . . g(f (vq )) are not distinct. Then
(g ◦ f )(v0 ), . . . (g ◦ f )(vq ) are also not distinct. So,

(g ◦ f )# (σ) = 0
= g# (⟨f (v0 ), . . . f (vq )⟩)
= (g# ◦ f# )(σ)

Case 3. All f (v0 ), . . . f (vq ) and all g(f (v0 )), . . . g(f (vq )) are distinct. So all (g ◦ f )(v0 ), . . . (g ◦
f )(vq ) are distinct.

(g ◦ f )# (σ) = ⟨(g ◦ f )(v0 ), . . . (g ◦ f )(vq )⟩

= g# (⟨f (v0 ), . . . f (vq )⟩)
= (g# ◦ f# )(σ)

Hence, (g ◦ f )# = g# ◦ f# .
(ii) I# (σ) = ⟨I(v0 ), . . . , I(vq )⟩ = ⟨v0 , . . . , vq ⟩ = σ, as vi ’s are all distinct.
Corollary 2.1.2. (i) For each q, 0 ≤ q ≤ dim(K), if f∗ : Hq (K) → Hq (L) and g∗ : Hq (L) →
Hq (M ) are the homomorphisms induced by f and g respectively then g∗ ◦ f∗ = (g ◦ f )∗ .
(ii) For each q, 0 ≤ q ≤ dim(K), if I∗ : Hq (K) → Hq (K) is the induced by the identity map
I : K → K, then I∗ is the identity homomorphism.
Proof. Follows immediately from the theorem.

2.2 Computation of Homology groups

It is natural to ask whether or not the homology group of a simplicial complex depends on
the orientation of the simplicial complex. We see next that homology groups do not depend
on the orientation of the simplicial complex.
A. Deb Ray, Pure Math, C.U. 15

Theorem 2.2.1. If K1 and K2 are two oriented simplicial complexes obtained from a sim-
plicial complex K then for each q,
Hq (K1 ) ∼
= Hq (K2 )
Proof. Suppose, σ q ∈ K is a q-simplex. Let σ1q be the oriented q-simplex in K1 arose from σ q
and σ2q be the oriented q-simplex in K2 arose from σ q . Then q q q
( σ2 is either +σ1 or −σ1 . There-
1, if σ1q and σ2q are of same sign
fore, we get a function ϕ : K → {−1, 1} given by ϕ(σ q ) = .
0, otherwise.
Then σ2q = ϕ(σ q )σ1q .
Define f# : Cq (K1 ) → Cq (K2 ) by
!
X q
X
f# mi σ1i = mi ϕ(σiq )σ2i
q

i i

We show that the following diagram commutes:

K1
∂q
... / Cq (K1 ) / Cq−1 (K1 ) / ...
f# f#
 ∂q
K2 
... / Cq (K2 ) / Cq−1 (K2 ) / ...

It is enough to show the result on the generators.

K1
∂q
... / Cq (K1 ) / Cq−1 (K1 ) / ...
f# f#
 K
∂q 2 
... / Cq (K2 ) / Cq−1 (K2 ) / ...

Choose any generator from Cq (K1 ), say, σ1q .

 
X
(f# ◦ ∂qK1 )(σ1q ) = f#  [σ1q , σ1q−1 ]σ1q−1 
σ1q−1 ∈K1

X
= [σ1q , σ1q−1 ]ϕ(σ q−1 )σ2q−1
σ1q−1 ∈K1
X
= [ϕ(σ q )σ2q , ϕ(σ q−1 )σ2q−1 ]ϕ(σ q−1 )σ2q−1
σ1q−1 ∈K1
X
= ϕ(σ q ) [σ2q , σ2q−1 ]σ2q−1
σ1q−1 ∈K1

= ϕ(σ q )∂qK2 (σ2q )

= ∂qK2 (ϕ(σ q )σ2q )
= ∂qK2 (f# (σ1q ))
A. Deb Ray, Pure Math, C.U. 16

So, f# induces a homomorphism f∗ : Hq (K1 ) → Hq (K2 ).

Changing the role of K1 and K2 , we can define in a similar way, a chain map g# : Cq (K2 ) →
Cq (K1 ) given by !
X q
X
g# mi σ2i = mi ϕ(σiq )σ1i
q
.
i i

Consequently, we get a homomorphism g∗ : Hq (K2 ) → Hq (K1 ). It is now a routine exercise

to check that f∗ and g∗ are inverse of each other. Hence, f∗ becomes an isomorphism, as
desired.

In the last theorem, we have proved that the homology groups do not depend on the
orientation of the given complex. So, we can choose any suitable orientation on the simplicial
complex whose homology groups we compute.

Example 2.2.1. Consider K = {⟨a0 ⟩ , ⟨a1 ⟩ , ⟨a2 ⟩ , ⟨a0 , a1 ⟩ , ⟨a1 , a2 ⟩ , ⟨a0 , a2 ⟩}. The simplicial
homology groups of K are as follows:
(
Z, if q = 0, 1
Hq (K) =
0, otherwise

Define an ordering on the vertices of K as a0 < a1 < a2 , so that K becomes an oriented

simplicial complex. Since there are three 0-simplexes and three 1-simplexes, the nontrivial
portion of the chain complex is
∂
0 → C1 (K) →1 C0 (K) → 0

Therefore, H1 (K) = Z1 (K)/0 ∼

= Z1 (K), H0 (K) = C0 (K)/B0 (K) and all other Hq (K) = 0.
So, we need to compute Z1 (K) = Ker(∂1 ) and B0 (K) = Im(∂1 ). We also know that

ˆ C0 (K) = Z ⟨a0 ⟩ ⊕ Z ⟨a1 ⟩ ⊕ Z ⟨a2 ⟩

ˆ C1 (K) = Z ⟨a0 , a1 ⟩ ⊕ Z ⟨a1 , a2 ⟩ ⊕ Z ⟨a0 , a2 ⟩

Let z ∈ Z1 (K). Then z ∈ C1 (K) such that ∂1 (z) = 0 which implies that z = m0 ⟨a0 , a1 ⟩ +
m1 ⟨a1 , a2 ⟩ + m2 ⟨a0 , a2 ⟩ such that

∂1 (m0 ⟨a0 , a1 ⟩ + m1 ⟨a1 , a2 ⟩ + m2 ⟨a0 , a2 ⟩) = 0

⇒ (m0 ⟨a1 ⟩ − m0 ⟨a0 ⟩) + (m1 ⟨a2 ⟩ − m1 ⟨a1 ⟩)
+(m2 ⟨a2 ⟩ − m2 ⟨a0 ⟩) = 0
⇒ −(m0 + m2 ) ⟨a0 ⟩ + (m0 − m1 ) ⟨a1 ⟩
+(m1 + m2 ) ⟨a2 ⟩ = 0
⇒ m0 = m1 = −m2

Hence, z = m0 σ where σ = ⟨a0 , a1 ⟩ + ⟨a1 , a2 ⟩ − ⟨a0 , a2 ⟩ ∈ C1 (K). i.e., Z1 (K) = Ker(∂1 ) ∼

= Z.
Let w ∈ B0 (K). Then there exists some z ∈ C1 (K) such that w = ∂1 (z) which implies that
A. Deb Ray, Pure Math, C.U. 17

there exists z = m0 ⟨a0 , a1 ⟩ + m1 ⟨a1 , a2 ⟩ + m2 ⟨a0 , a2 ⟩ such that

w = ∂1 (m0 ⟨a0 , a1 ⟩ + m1 ⟨a1 , a2 ⟩ + m2 ⟨a0 , a2 ⟩)
⇒ w = (m0 ⟨a1 ⟩ − m0 ⟨a0 ⟩) + (m1 ⟨a2 ⟩ − m1 ⟨a1 ⟩)
+(m2 ⟨a2 ⟩ − m2 ⟨a0 ⟩)
⇒ w = −(m0 + m2 ) ⟨a0 ⟩ + (m0 − m1 ) ⟨a1 ⟩
+(m1 + m2 ) ⟨a2 ⟩
⇒ w = t0 ⟨a0 ⟩ + t1 ⟨a1 ⟩ + t2 ⟨a2 ⟩
where t0 + t1 + t2 = −(m0 + m2 ) + (m0 − m1 ) + (m1 + m2 ) = 0. Hence, w = t1 (⟨a1 ⟩ −
⟨a0 ⟩) + t2 (⟨a2 ⟩ − ⟨a0 ⟩), with ⟨a1 ⟩ − ⟨a0 ⟩, ⟨a2 ⟩ − ⟨a0 ⟩ ∈ C0 (K). i.e., B0 (K) = Im(∂1 ) ∼ =
Z(⟨a1 ⟩ − ⟨a0 ⟩) ⊕ Z(⟨a2 ⟩ − ⟨a0 ⟩).
Therefore, we get H0 (K) = C0 (K)/B0 (K) ∼ = (Z ⟨a0 ⟩ ⊕ Z ⟨a1 ⟩ ⊕ Z ⟨a2 ⟩)/(Z(⟨a1 ⟩ − ⟨a0 ⟩) ⊕
Z(⟨a2 ⟩ − ⟨a0 ⟩)) ∼= (Z ⊕ Z ⊕ Z)/(Z ⊕ Z) ∼ = Z. H1 (K) = Z1 (K)/B1 (K) = Z/0 ∼ = Z. Hence,
the homology groups of K are
(
Z, if q = 0, 1
Hq (K) =
0, otherwise

Exercise 2.2.1. Consider K = {⟨a0 ⟩ , ⟨a1 ⟩ , ⟨a2 ⟩ , ⟨a0 , a1 ⟩ , ⟨a1 , a2 ⟩ , ⟨a0 , a2 ⟩ , ⟨a0 , a1 , a2 ⟩}. Show
that the simplicial homology groups of K are as follows:
(
Z, if q = 0
Hq (K) =
0, otherwise

Example 2.2.2. Consider K = {⟨a0 ⟩ , ⟨a1 ⟩ , ⟨a2 ⟩ , ⟨a0 , a1 ⟩ , ⟨a1 , a2 ⟩ , ⟨a0 , a2 ⟩ , ⟨a0 , a1 , a2 ⟩ , ⟨b0 ⟩ , ⟨b1 ⟩ ,
⟨b0 , b1 ⟩}. Show that the simplicial homology groups of K are :
(
Z ⊕ Z, if q = 0
Hq (K) =
0, otherwise

2.3 Reduced Homology groups

We recapitulate a few concepts from algebra before introducing the reduced homology groups.
ˆ An abelian group G is a Z-module. An abelian group G is free if it has a basis (linearly
independent set of generators).
ˆ Some typical examples of free abelian groups are Z, Z ⊕ Z, Z ⊕ Z ⊕ . . . ⊕ Z, etc.

ˆ Any subgroup of a free abelian group is free.

ˆ Any free abelian group is torsion-free.

fn
ˆ Suppose {Mn } is a sequence of R-modules and Mn → Mn+1 (for each n) is a module
morphism. This sequence of modules and module morphisms is said to be exact if
Ker(fn+1 ) = Im(fn ), for each n.
A. Deb Ray, Pure Math, C.U. 18

f g
ˆ An exact sequence of the form 0 → N → M → P → 0 is called a short exact
sequence.
f g
ˆ A short exact sequence of the form 0 → N → N ⊕ P → P → 0 is called a split exact
sequence.
f g
ˆ A short exact sequence 0 → N → M → P → 0 is said to split if M ∼
= N ⊕ P.

Theorem 2.3.1. The following statements are equivalent:

f g
1. An exact sequence 0 → N → M → P → 0 splits;
′ ′
2. there exists a homomorphism f : M → N such that f (f (h)) = h, for all h ∈ N ;
′ ′

3. there exists a homomorphism g : P → M such that g g (k) = k, for all k ∈ P .

We introduce here the notion of Reduced homology groups, denoted by H

e q (K), for a simpli-
cial complex K.

Definition 2.3.1. Let K be a simplicial complex and the following a chain complex of K.
∂ ∂q+1 ∂q ∂ ∂
. . . → Cq+1 (K) → Cq (K) → Cq−1 (K) . . . →2 C1 (K) →1 C0 (K) → 0.

Define a map ϵ : C0 (K) → Z on the vertices of K by ϵ(v) = 1 for each vertexv of K  and
m
P
then extend the definition to the whole of C0 (K) by extending it linearly. i.e, ϵ ni v i =
i=1
m
P m
P
ni ϵ(vi ), i.e., = ni ). This map is called an augmentation map.
i=1 i=1

Clearly, an augmentation map ϵ has the following properties:

ˆ ϵ is a surjective homomorphism : That it is a homomorphism is clear (check) and for

surjectiveness, consider any n ∈ Z. Then for any vertex v of K, nv ∈ C0 (K) such that
n = ϵ(nv).
∂ ϵ
ˆ If we consider the sequence C1 (K) →1 C0 (K) → Z, then ϵ ◦ ∂1 = 0 : Since C1 (K) is
generated by 1-simplexes of K, it is enough to prove the result on the set of 1-simplexes
of K. So,
(ϵ ◦ ∂1 )(⟨v0 , v1 ⟩) = ϵ(⟨v1 ⟩ − ⟨v0 ⟩) = 1 − 1 = 0.

Definition 2.3.2. The 0-dimensional reduced homology group, denoted by H e 0 (K), of a

simplicial complex K is defined by H e 0 (K) = Kerϵ/Im(∂1 ), where ϵ : C0 (K) → Z is the
augmentation map.
For any q ̸= 0, we define H
e q (K) = Hq (K).

We now establish a formula that connects a 0-dimensional homology group with the corre-
sponding 0-dimensional reduced homology group.
A. Deb Ray, Pure Math, C.U. 19

Theorem 2.3.2. If K is a oriented simplicial cmplex then

H0 (K) ∼
=He 0 (K) ⊕ Z.
ι ϵ
Proof. Consider the short exact sequence 0 → Kerϵ → C0 (K) → Z → 0 and define the
following:
Kerϵ bι C0 (K) bϵ
0→ → → Z → 0,
Im∂1 Im∂1
where bι is the inclusion map given by bι(c + Im∂1 ) = c + Im∂1 and b ϵ(c + Im∂1 ) = ϵ(c). This
second sequence is also exact. Because, b ι is injective, b
ϵ is surjective and Kerbϵ = Imb ι.
Since Z is a free abelian group generated by {1}, define a homomorphism f : Z → C0 (K)
by setting f (1) = v, choosing any vertex v from K. Define fb : Z → C0 (K)/Im(δ  ) by
1
fb(n) = nv + Im(δ1 ), where v is that chosen vertex. Clearly, for all n ∈ Z, b ϵ ◦ fb (n)
=b ϵ(nv + Im(δ1 )) = ϵ(nv) = n. Therefore, by Theorem 2.3.1, the short exact sequence splits,
0 (K) ∼ Kerϵ
and so, CIm∂ 1
= Im∂1 ⊕ Z. Hence, H0 (K) ∼=H e 0 (K) ⊕ Z.

2.4 Zero dimensional Homology groups

Before discussing zero dimensional homology groups, we revisit a few elementary facts about
simplicial complexes and their geometric carriers.
Observation 2.4.1. We observe the following:
ˆ Any subset of a geometrically independent set is always geometrically independent.
(check!)
ˆ Suppose τ = ⟨v0 , v1 , . . . , vq ⟩ is a face of σ = ⟨v0 , v1 , . . . , vq , vq+1 , . . . , vk ⟩. If any point
Pq Pq
x ∈ τ ⊂ σ has the form x = αi vi , where αi ≥ 0, αi = 1 as a member of τ
i=0 i=0
k
P k
P
and x = βi vi , where βi ≥ 0, βi = 1 as a member of σ, then αi = βi for each
i=0 i=0
i = 0, 1, . . . , q and βi = 0 for all i = q + 1, . . . , k. (Check!)
ˆ Every simplex is a convex set. (check!)
Theorem 2.4.1. Let K be a simplicial complex. If σ and τ are two members of K such
that τ is not a face of σ, then σ ∩ int(τ ) = ∅.
Proof. There are two possibilities :
Case 1 : σ is a face of τ ;
Case 2 : σ is not a face of τ .
In case 1, each point x on σ has barycentric co-ordinates equal to 0 corresponding to each v
such that v is a vertex of τ but not of σ. Consequently, x ∈
/ int(τ ) and hence, σ ∩ int(τ ) = ∅.
Case 2 : If σ ∩ τ = ∅, then clearly, σ ∩ int(τ ) = ∅. So, let us assume that σ ∩ τ ̸= ∅ and
x ∈ σ ∩ τ . σ ∩ τ is a common face of σ and τ . By observation 2.4.1, there exists a vertex v
of τ which is not in σ ∩ τ such that the barycentric co-ordinate of x corresponding to v is 0.
Consequently, x ∈ / int(τ ). Hence, σ ∩ int(τ ) = ∅.
A. Deb Ray, Pure Math, C.U. 20

S
Theorem 2.4.2. If K is a simplicial complex then |K| = int (σ) and int (σ)∩int (τ ) = ∅
σ∈K
whenever σ ̸= τ . i.e., {int (σ) : σ ∈ K} partitions the geometric carrier |K| of K.
S
Proof. By definition, |K| = σ and int (σ) ⊆ σ, for each σ ∈ K. Hence, one side of the
σ∈K
desired equality is clear.
Let x ∈ |K|. Then there exists some σ ∈ K such that x ∈ σ. If x ∈ int(σ) then we get what
we need to prove. But if x ∈
/ int(σ) then x ∈ γ where γ is a face of σ. Again, if x ∈ int(γ)
then γ being a member of K, we obtain the result. But if x ∈ / int(γ) then it must be on
some face of γ. Repeating this argument finitely many times,Swe must get a simplexSβ (may
be of dimension 0) of K, such that x ∈ int(β). Hence, x ∈ int(σ). i.e., |K| = σ.
σ∈K σ∈K
Suppose σ ̸= τ in K. Then either there exists one vertex of σ which is not a vertex of τ or
one vertex of τ which is not a vertex of σ. Without loss of generality, we assume that there is
one vertex of σ which is not a vertex of τ . So, σ is not a face of τ . Hence, by Theorem 2.4.1,
int(σ) ∩ τ = ∅. So,
int(σ) ∩ int(τ ) ⊆ int(σ) ∩ τ = ∅.

One may prove the following:

Theorem 2.4.3. For every vertex v of K, ost(v) = ∪{int(σ k ) : v is a vertex of σ k } is an

open subset of |K|.

Proof. Exercise.

Theorem 2.4.4. Let K be a simplicial complex. If |K| has k many topological components
and vi , i = 1, . . . k are vertices of K taken exactly one from each component Ki of |K| then
H0 (K) is a free abelian group with a basis {[vi ] : i = 1, . . . k}, where [vi ] stands for the
homology class corresponding to vi ∈ C0 (K). Moreover, H0 (K) ∼ = ⊕ki=1 Z.

Proof. We prove this result in two steps.

Step 1. Consider the set of vertices of K, say vert(K). Define a relation ∼ on vert(K) as
follows:
v ∼ w if and only if there exists a sequence of (distinct) vertices a0 , a1 , . . . , an such that
a0 = v and an = w and ⟨ai , ai+1 ⟩ is a 1-simplex of K.
It is easy to check that ∼ is an equivalence
S relation on vert(K). (check!)
We show that for each vertex v, Cv = ost(w) is a topological component of |K|. One
w∼v
may easily see that (check!)

ˆ If v ∼ w then Cv = Cw .

ˆ If v ≁ w then Cv ∩ Cw = ∅.

ˆ Each Cv is open in |K|.

ˆ Each Cv is path connected.

A. Deb Ray, Pure Math, C.U. 21

Cv ’s are disjoint open connected subsets of |K| and therefore form components of |K|.
Step 2. Choose exactly one vertex vi from each component Cv and construct the set
B = {[vi ] : i = 1, 2, . . . , r} (Here [vi ] denotes the homology class vi + Im∂1 in H0 (K)). We
show that B is a basis of H0 (K).
ˆ B generates H0 (K) : Let [w] ∈ H0 (K). Then w ∈ C0 (K) and so w =
P
ni wi , where
i
wi ’s are 0-simplexes of K. Suppose wk ∈ Cvj . Then wk ∼ vj . i.e., There is a sequence
of vertices a0 , a1 , . . . an such that a0 = wk , an = vj and ⟨ai , ai+1 ⟩ is a 1-simplex of K.
n−1
P
Consider c = ⟨ai , ai+1 ⟩ ∈ C1 (K). Then
i=0

n−1
!
X
∂1 (c) = ∂1 ⟨ai , ai+1 ⟩
i=0
n−1
X
= (⟨ai+1 ⟩ − ⟨ai ⟩)
i=0
= ⟨vj ⟩ − ⟨wk ⟩
P r
P
So, vj − wk ∈ Im∂1 ⇒ [vj ] = [wk ]. Hence, [w] = ni [wi ] = mj [vj ]. Therefore, B
i j=1
generates H0 (K).
r r
ˆ B is linearly independent : Let
P P
mi [vi ] = 0. Then mi vi ∈ Im∂1 . There exists
i=1 i=1
r
P P
d ∈ C1 (K) such that ∂1 (d) = mi vi . Again d ∈ C1 (K) implies that d = dj , where
i=1 j
dj lie in one component Cvj , j = 1, 2, . . . , r. Hence, ∂1 (dj ) = mj vj . (check!)
If we consider the augmentation map ϵ : C0 (K) → Z given by ϵ(a) = 1 for each vertex
a ∈ K, then for each j = 1, 2, . . . r,

mj = ϵ(mj vj ) = ϵ∂1 (dj ) = 0

Therefore, H0 (K) is a free abelian group.

r
Define a map ψ : ⊕ri=1 Z → H0 (K) by ψ(m1 , m2 , . . . , mr ) =
P
mi [vi ]. That ψ is a homomor-
i=1
phism is a routine exercise (check!).
Since B generates H0 (K), ψ is onto. ψ is one-one follows from the fact that B is linearly
independent. Hence,
H0 (K) ∼
= ⊕ri=1 Z.

2.5 Homology groups of cl(σ n) and sk(σ n)

In Chapter 2, we have directly computed the homology groups of K = cl(σ 2 ) and have seen
that H0 (K) = Z and trivial for all other q ̸= 0. Similarly, we have computed the homology
A. Deb Ray, Pure Math, C.U. 22

groups of K1 = sk(σ 2 ) and have obtained that H0 (K) = H1 (K) = Z and trivial for all other
q ̸= 0, 1. It is natural to ask whether we can extend these results to achieve homology groups
of cl(σ n ) and sk(σ n ) for any n. In this section we answer this question in affirmative.
Definition 2.5.1. Let K be a simplicial complex and w ∈ Rn be such that any ray emanating
from w intersects |K| at atmost one point. Consider w ∗ K as the collection of all simplexes
of the form ⟨w, a0 , a1 , . . . , an ⟩, where ⟨a0 , a1 , . . . , an ⟩ is a simplex in K and all possible faces
of ⟨w, a0 , a1 , . . . , an ⟩. Then w ∗ K is a simplicial complex called a cone over K with vertex
w.
Exercise 2.5.1. (worked-out in class) Show that w ∗ K is a simplicial complex.
Exercise 2.5.2. (worked-out in class) The geometric carrier |w ∗ K| is path connected (and
hence, connected).
Observation 2.5.1. It is easy to observe, if σ 2 = ⟨a0 , a1 , a2 ⟩ and σ1 = ⟨a1 , a2 ⟩, then cl(σ 2 ) =
a0 ∗ cl(σ1 ).
We may generalize this observation for any n as follows: If σ n = ⟨a0 , a1 , a2 , . . . , an ⟩ and
σn−1 = ⟨a1 , a2 , . . . , an ⟩, then cl(σ n ) = a0 ∗ cl(σn−1 ).
Observation 2.5.2. Clearly, cl(σ 2 ) = a0 ∗ cl(σ1 ) (where σ1 has all the vertices of σ2 except
a0 ). Proceeding similarly, one may obtain cl(σ n ) is a cone over cl(σ n−1 ). Hence, for the
homology groups of cl(σ n ), we need to know the homology groups of a cone w ∗ K over K.
Theorem 2.5.1. Let K be an oriented simplicial complex. If w ∈ Rn is such that w ∗ K is
a cone over K with vertex w then for all q,
(
Z, if q = 0
Hq (w ∗ K) =
0, otherwise.

(i.e., Reduced homology groups of a cone is trivial for all q)

Proof. Let σ = ⟨a0 , a1 , . . . , aq ⟩ be an oriented q-simplex of K. We denote
P the oriented (q +1)-
simplex ⟨w, a0 , a1 , . . . , aq ⟩ of w ∗ K by [w, σ]. For any q-chain c = ni σi , define [w, c] by
P i
ni [w, σi ].
i
If σ is a 0-simplex, then ∂1 [w, σ] = ∂1 (⟨w, σ⟩) = σ − w. But if σ = ⟨a0 , a1 , . . . , aq ⟩ is a
q-simplex with q > 0, then

∂q+1 ([w, σ]) = ∂q+1 (⟨w, a0 , a1 , . . . , aq ⟩)

q
X
= σ+ (−1)i ⟨w, a0 , . . . , b ai−1 , . . . , aq ⟩
i=1
q
X
= σ− (−1)i ⟨w, a0 , . . . , b
ai−1 , . . . , aq ⟩
i=0
= σ − [w, ∂q ⟨a0 , a1 , . . . , aq ⟩]
= σ − [w, ∂q σ]
A. Deb Ray, Pure Math, C.U. 23

Therefore, more generally, for any q-chain cq (check),

(
cq − ϵ(cq )w, if q = 0
∂q+1 [w, cq ] =
cq − [w, ∂q cq ], if q > 0

In view of Theorem 2.4.4 and Exercise 2.5.2, H0 (w ∗ K) = Z. Moreover, by definition

of homology groups, Hq (w ∗ K) = 0 for all q < 0 or q > dim(w ∗ K). So, we consider
0 < q ≤ dim(w ∗ K). Suppose K1 = w ∗ K.
Let zq + Bq (K1 ) ∈ Hq (K1 ). Then zq is a q-cycle. In order to prove that Hq (K1 ) = 0, we
need to show that zq + Bq (K1 ) = Bq (K1 ), i.e., zq ∈ Bq (K1 ).
Consider zq ∈ Zq and let zq = cq + [w, dq−1 ], where cq is that portion of the chain that does
not involve the vertex w and hence a chain on K; and dq−1 is a q − 1 chain in K. Then

zq − ∂q+1 [w, cq ] = cq + [w, dq−1 ] − cq + [w, ∂q cq ] = [w, cq−1 ]

where cq−1 = dq−1 + ∂q (cq ) is a (q − 1)-chain in K. Then ∂q (zq − ∂q+1 [w, cq ]) = ∂q ([w, cq−1 ])
implies (since, ∂q+1 ◦ ∂q = 0 and ∂q (zq ) = 0),
(
cq−1 − ϵ(cq−1 )w, if q = 1
0=
cq−1 − [w, ∂q−1 cq−1 ], if q > 1

which implies that (

ϵ(cq−1 )w, if q = 1
cq−1 =
[w, ∂q−1 cq−1 ], if q > 1
Both the terms on the right side of the above expression contain the vertex w while cq−1
does not. This is possible only when cq−1 = 0. In such case, zq − ∂q+1 [w, cq ] = [w, cq−1 ] = 0
and consequently, zq = ∂q+1 [w, cq ] ∈ Bq (K1 ), as desired.

Corollary 2.5.1. For each n > 0,

(
Z, if q = 0
Hq (cl(σ n )) =
0, otherwise

Proof. For each n > 0, |cl(σ n )| = |a0 ∗ cl(σ n−1 )|, where σ n = ⟨a0 , a1 , . . . , an ⟩ and σ n−1 =
⟨a1 , . . . , an ⟩. By Theorem 2.5.1, the result is immediate.

Theorem 2.5.2. For each n > 0,

(
Z, if q = 0 or n
Hq (sk(σ n+1 )) =
0, otherwise.

Proof. Consider sk(σ n+1 ) and construct the chain complex as follows:

∂ ∂n−1 ∂ ∂
0 → Cn (sk(σ n+1 )) −→
n
Cn−1 (sk(σ n+1 )) −→ . . . C1 (sk(σ n+1 )) −→
1
C0 (sk(σ n+1 )) −→
0
0
A. Deb Ray, Pure Math, C.U. 24

The chain complex of clσ n+1 is

∂n+1 ∂ ∂n−1 ∂ ∂
0 → Cn+1 (cl(σ n+1 )) −→ Cn (cl(σ n+1 )) −→
n
Cn−1 (cl(σ n+1 )) −→ . . . −→
1
C0 (cl(σ n+1 )) −→
0
0

Cq (cl(σ n+1 )) = Cq (sk(σ n+1 )), for all q = 0, 1, . . . , n. Hence, the q-th homology group of
cl(σ n+1 ) and sk(σ n+1 ) are same for q = 0, 1, . . . n − 1. So, we obtain
(
Z, if q = 0
Hq (sk(σ n+1 )) =
0, q = 1, 2, . . . n − 1.

Certainly, Hq (Sk(σ n+1 )) = 0 for any q > n or q < 0. Therefore, we need to calculate the
homology group Hn (Sk(σ n+1 )). Consider the following portions of the chain complexes:

∂n+1 ∂n /
Zσ n+1 ∼
= Cn+1 (cl(σ n+1 )) / Cn (cl(σ3 n+1 )) = Cn (sk(σ n+1 )) Cn−1 (cl(σ n+1 )) = Cn−1 (sk(σ n+1 ))

Now, Hn (sk(σ n+1 )) = Ker∂n /Im(0) ∼ = Ker∂n and Hn (cl(σ n+1 )) = Ker∂n /Im(∂n+1 ). But
n+1
Hn (cl(σ )) = 0. Hence, Ker∂n = Im(∂n+1 ) = ∂n+1 (Zσn+1 ). Since homomorphic image of
a cyclic group is cyclic, Ker∂n becomes a cyclic subgroup of Cn (clσn+1 ). Also, Cn (clσn+1 )
being a free abelian group, it has no torsion element. Therefore, Ker∂n can not be a finite
cyclic group. (For if it is a finite cyclic group, a generator of it would have been of finite order
and hence a torsion element). So, Ker∂n is an infinite cyclic group and hence isomorphic to
Z. i.e., Hn (sk(σ n+1 )) ∼
= Z, as desired.
Chapter 3

Simplicial Approximations

In this chapter we observe that a simplicial map between two simplicial complexes induces a
continuous function between their corresponding geometric carriers |K| and |L|. We shall ex-
plore whether or not any continuous map can be approximated by a simplicial map. For that
matter, we need to make the term ‘approximation’ more precise and search for a satisfactory
answer to this question.

3.1 Continuous map induced by a simplicial map

We recapitulate the definition of a simplicial map :

Definition 3.1.1. Let K and L denote two simplicial complexes. A function f : K → L

which maps vertices of K to vertices of L is said to be a simplicial map if whenever
⟨v0 , v1 , . . . , vq ⟩ is a simplex in K, f (v0 ), ..., f (vq ) are vertices of a simplex in L. i.e.,
⟨f (v0 ), ..., f (vq )⟩ is a simplex in L after dropping the repetition (if any).

A simplicial map carries a simplex of K to a simplex of L. To be more precise, for every

k, a simplicial map takes a k-simplex of K to a l-simplex of L, for some l ≤ k.

Example 3.1.1. Let K = {⟨v0 ⟩ , ⟨v1 ⟩ , ⟨v2 ⟩ , ⟨v0 , v1 ⟩ , ⟨v1 , v2 ⟩} and

L = {⟨a⟩ , ⟨b⟩ , ⟨c⟩ , ⟨d⟩ , ⟨a, b⟩ , ⟨c, d⟩ , ⟨a, d⟩}. Then
(i) f : K → L given by f (v0 ) = a, f (v1 ) = b, f (v2 ) = a is a simplicial map.
(ii)f : K → L given by f (v0 ) = b, f (v1 ) = a, f (v2 ) = d is a simplicial map.
(iii) f : K → L given by f (v0 ) = a, f (v1 ) = b, f (v2 ) = c is not a simplicial map.
(iv) f : K → L given by f (v0 ) = a, f (v1 ) = a, f (v2 ) = c is not a simplicial map.

The following are easy consequences of the definition of simplicial maps:

Theorem 3.1.1. (i) If f : K → L and g : L → M are simplicial maps then g ◦ f : K → M

is a simplicial map, where K, L and M are simplicial complexes.
(ii) I : K → K given by I(v) = v, for each v ∈ ver(K) is a simplicial map.

Given a simplicial map f : K → L, we wish to define a map fb : |K| → |L| and establish
that such fb is continuous.
25
A. Deb Ray, Pure Math, C.U. 26

Theorem 3.1.2. Let K and L be two simplicial complexes. Each simplicial map f : K → L
induces a continuous function fb : |K| → |L|.

Proof. Let f : K → L be a simplicial map and take any x ∈ |K|. Without loss of generality,
m
P
we choose σ = ⟨a0 , a1 , . . . , am ⟩ such that x ∈ int(σ). Then x = λi ai where λi > 0, for all
i=0
m
P
i = 0, 1, . . . , m and define fb(x) = λi f (ai ). In view of an earlier observation (see the 2nd
i=0
part of Observation 2.4.1), the unique representation of x ensures that the function fb is well
defined. we need to show that fb is continuous.
Since each simplex σ is a closed set, it is enough to show that f |σ is continuous. Because,
whenever x ∈ σ ∩ τ , fb|σ (x) = fb|τ (x) (using Observation 2.4.1 2nd part, again) and hence,
by pasting lemma, fb is continuous on |K|.
Consider the barycentric representation (λ1 , λ2 , . . . , λm ) of x.
Case 1 : Assume that for all i = 0, . . . , m, f (ai )’s are all distinct.
Consider the following diagram, for each j = 0, . . . , m.

fb
(λ0 , λ1 , . . . , λm ) / (λ0 , λ1 , . . . , λm )
πj
πj ◦fb 
)
λj

For each j = 0, 1, 2, . . . , m, πj ◦ fb is a projection map and hence continuous. Therefore, fb is

continuous.
Case 2 : Assume that for all i = 0, . . . , m, f (ai )’s are not all distinct. Without loss of
generality we assume that f (ak ) = f (as ) and rest of the vertices are all distinct.
Then for j ̸= k, s, the composition πj ◦ fb is a projection map and hence continuous. In case
j = k, we have the following diagram:

/f(λ0 , λ1 , . . . , λk
b
(λ0 , λ1 , . . . , λm ) + λs , . . . λm )
πk
πk ◦fb * 
λk + λs

So πk ◦fb = πk +πs , i.e., the sum of two continuous maps and hence, continuous. Consequently,
fb is continuous.
It is now natural to ask whether a continuous function g : |K| → |L| is achievable from
a simplicial map K → L. Of course it is too ambitious to hope for getting a simplicial map
f : K → L such that the induced continuous function fb is exactly equal to the given function
g. But, one may relax the condition of equality by replacing it with homotopy and frame a
little more convincing question as follows :
“For any continuous function g : |K| → |L| can we get a simplicial map f : K → L, such
that fb is homotopic to g? ”
The answer to this question is still in negative. If such f exists then it is called a simplicial
A. Deb Ray, Pure Math, C.U. 27

approximation of g. In the section that follows next, we explore, under which condition(s)
such ‘simplicial approximations’ exist.

3.2 Simplicial Approximation

We first show that a continuous function g : |K| → |L| always possesses a simplicial approx-
imation when K and L are two simplicial complexes satisfying a special condition that they
are ‘star-related with respect to g’.

Definition 3.2.1. Let K and L be two simplicial complexes and g : |K| → |L| a continuous
function. K is said to be star-related to L with respect to g, if for each vertex v of K there
exists a vertex w of L such that g(ost(v)) ⊆ ost(w).

Lemma 3.2.1. A geometrically independent set of vertices {v0 , v1 , . . . , vk } of a simplicial

k
T
complex K forms a simplex ⟨v0 , v1 , . . . , vk ⟩ in K if and only if ost(vi ) ̸= ∅.
i=0

Proof. Suppose {v0 , v1 , . . . , vk } constitutes a set of vertices of a simplex σ ∈ K. i.e., σ =

⟨v0 , v1 , . . . , vk ⟩. Now, for each i = 0, 1, . . . , k,
[
ost(vi ) = {int(τ ) : vi is a vertex of τ } ⊇ int(σ)
τ

k
T
Therefore, ost(vi ) ⊇ int(σ) ̸= ∅.
i=0
k
T
Conversely, let {v0 , v1 , . . . , vk } be such that ost(vi ) ̸= ∅. Then there exists some x ∈
i=0
ost(vi ), for each i = 0, 1, . . . , k. So x ∈ int(σi ), for some simplex σi (i = 0, 1, . . . , k) of K
with vi as a vertex. Since {int(σ) : σ ∈ K} forms a partition of K, σ0 = σ1 = . . . = σk = τ
(say). Since vi is a vertex of σi for each i, it follows that v0 , v1 , . . . , vk are vertices of τ . Then
either ⟨v0 , v1 , . . . , vk ⟩ is τ or a face of τ . In either case, ⟨v0 , v1 , . . . , vk ⟩ ∈ K.

Theorem 3.2.1. Suppose K and L are simplicial complexes and g : |K| → |L|, a continuous
function. If K is star-related to L with respect to g, then there is a simplicial map f : K → L
such that the induced continuous function fb is homotopic to g. (in otherwords, g has a
simplicial approximation)

Proof. K and L are star-related with respect to g implies that for each v in K there exists
w ∈ L such that g(ost(v)) ⊆ ost(w). Define f : vert(K) → vert(L) as follows:
f (v) = w whenever g(ost(v)) ⊆ ost(w).
We claim that f is a simplicial map. For that, we take any σ = ⟨v0 , . . . , vk ⟩ from
 K.
Tk Tk
By Lemma 3.2.1, ost(vi ) ̸= ∅. For each i, ost(f (vi )) ⊇ g(ost(vi )) ⊇ g ost(vi ) ̸= ∅.
i=0 i=0
Applying the lemma again, ⟨f (v0 ), f (v1 ), . . . , f (vk )⟩ is a simplex in L. Hence, f is a simplicial
A. Deb Ray, Pure Math, C.U. 28

map.
Let x ∈ |K|. Then there exists σ = ⟨v0 , v1 , . . . , vn ⟩ ∈ K such that x ∈ int(σ). Then
n
P n
P n
P
x = λi vi , where each λi > 0 and λi = 1. Then fb(x) = λi f (vi ). If each f (vi ) is
i=1 i=1 i=1
distinct, then the coefficients of fb(x) (i.e., λi ’s) are all > 0. If for some i ̸= j f (vi ) = f (vj ),
then the i-th coefficient in fb(x) (i.e., λi + λj ) is > 0. So, in either case, barycentric co-
ordinates are > 0. Hence, fb(x) ∈ int (⟨f (v0 ), . . . , f (vn )⟩) (cancelling possible repetitions of
vertices). So, for each i = 0, 1, . . . , n, fb(x) ∈ ost (f (vi )). On the other hand,

x ∈ ost(vi ) ⇒ g(x) ∈ g(ost(vi )) ⊆ ost (f (vi )) .

Therefore, both g(x) and fb(x) are in ost(f (vi )), for each i = 0, 1, 2, . . . n. i.e., g(x) and fb(x)
n
T
are in ost(f (vi )), hence by Lemma 3.2.1, there exists some simplex τ with vertices f (vi ),
i=0
i = 0, 1, . . . , n, such that g(x), fb(x) ∈ τ . A simplex being a convex set, the line segment
tg(x)+(1−t)fb(x) ∈ τ , for all t ∈ [0, 1]. Therefore, we define a homotopy H : |K|×[0, 1] → |L|
by
H(x, t) = tg(x) + (1 − t)fb(x)
H
It is clear that g ∼ fb.
How far can we extend this result? For obtaining a simplicial approximation of any
continuous function g : |K| → |L|, we need to build some basics.
Definition 3.2.2. Let σ = ⟨v0 , v1 , . . . , vk ⟩ be a simplex. The barycentre of σ, denoted by σ̇
is the point on σ whose barycentric co-ordinates are all equal. i.e.,
k
X 1
σ̇ = vi
i=0
k+1

Observation 3.2.1. One may observe that

ˆ barycentre of a 0-simplex is itself;

ˆ barycentre of a 1-simplex is the midpoint of the 1-simplex;

ˆ barycentre of a 2-simplex is the centroid of the triangle formed by the 2-simplex, etc.
We construct a new simplicial complex K (1) from K obeying the rules mentioned below:
1. Take the barycentres of all the simplexes of K as the set of vertices (i.e., 0-simplexes);
2. If σ0 , σ1 , . . . , σn ∈ K, then take ⟨σ˙0 , σ˙1 , . . . , σ˙k ⟩ ∈ K (1) if and only if there is some
permutation t of {0, 1, . . . , n} such that σt(i) is a proper face of σt(i+1) in K.
K (1) is known as the first barycentric subdivision of K.

(1)
The first barycentric subdivision K (1) of K (1) is called the 2nd barycentric subdivision
of K. Hence, by induction, the n-th barycentric subdivision of K, denoted by K (n) is given

(1)
by K (n−1) , for all n ∈ N, where by K 0 we mean the original simplicial complex K.
A. Deb Ray, Pure Math, C.U. 29

Definition 3.2.3. Let K be a simplicial complex. The mesh of K, denoted by mesh(K), is

defined as follows:
mesh(K) = max{ sup ||x − y||}
σ∈K x,y∈σ

where ||.|| stands for the Euclidean norm.

The following lemma shows that for calculating the mesh of a complex it is enough to
calculate the lengths of the 1-simplexes of K and take the maximum.

Lemma 3.2.2. If σ is a k-dimensional simplex (k > 0) of K then there are vertices v, w of

σ such that diameter(σ) = ||v − w||.

k
P k
P
Proof. Let σ = ⟨v0 , v1 , . . . , vk ⟩ and x, y ∈ σ. Then x = λi vi and y = µi vi , where
i=0 i=0
k
P k
P
λi , µi ≥ 0, λi = µi = 1.
i=0 i=0

Xk k
X k
X
||x − y|| = ||1.x − y|| = ||( µi )x − µi vi || = || µi (x − vi )||
i=0 i=0 i=0

Therefore,

k
X k
X
||x − y|| ≤ µi ||x − vi || ≤ µi max ||x − vi || = max ||x − vi ||.
i i
i=0 i=0

Again, proceeding similarly, ||x − vi || ≤ max ||vj − vi ||. Hence,

j

||x − y|| ≤ max ||vj − vi || = ||vr − vs ||

i,j

for some vertices vr , vs of σ. Then the diameter of σ = sup ||x − y|| ≤ ||vr − vs ||, for some
x,y∈σ
vertices vr , vs of σ. Clearly, ||vr − vs || ≤ sup ||x − y||. So, diameter(σ) = ||vr − vs ||, as
x,y∈σ
desired.


Theorem 3.2.2. If K is a simplicial complex with dim(K) > 0 then lim mesh K (n) = 0.
n→∞

Proof. Suppose

dim(K) = m. By definition of mesh and applying the above Lemma,
(1)
mesh K = max ||σ̇ − τ̇ ||.
⟨σ̇,τ̇ ⟩
Since ⟨σ̇, τ̇ ⟩ ∈ K (1) , σ is a proper face of τ .
p
1
P
Let τ = ⟨v0 , v1 , . . . , vp ⟩, where p ≤ m. Then τ̇ = p+1
vi is a point on τ . Also, σ̇ is a point
i=0
on σ and hence a point on τ . So, proceeding as the proof of the lemma above, there exists
A. Deb Ray, Pure Math, C.U. 30

some vertex vr of τ such that

||τ̇ − σ̇|| ≤ ||τ̇ − vr ||

p
1 X
= || vi − vr ||
p + 1 i=0
p
1 X
≤ ||vi − vr ||
p + 1 i=0
p
≤ mesh(K)
p+1
m
≤ mesh(K)
m+1
m


Hence, mesh K (1) ≤ m+1 mesh(K).
(2)

m m

2
Similarly, mesh K ≤ m+1 mesh(K (1) ) ≤ m+1
mesh(K). Repeating this process, we
obtain for each n ∈ N  n
(n) m
mesh(K )≤ mesh(K).
m+1
m

n
As n → ∞, m+1
→ 0 and therefore, lim mesh(K (n) ) = 0.
n→∞

Theorem 3.2.3. (Simplicial Approximation Theorem) For any continuous function

g : |K| → |L|, there exists k ∈ N and a simplicial map f : K (k) → L such that the
continuous function fb induced from f is homotopic to g. (K (k) denotes the k-th barycentric
subdivision of K)

Proof. K and L are finite simplicial complexes and therefore, |K| and |L| are compact subsets
of Euclidean spaces. {ost(w) : w ∈ vert(L)} is an open cover of |L| and so, applying
Lebesgue covering theorem, it has a Lebesgue number, say η1 > 0. Using the definition of
Lebesgue number, for each y ∈ |L|, there is some ost(w) such that Bη (y) ⊆ ost(w), where
Bη (y) = {z : ||z − y|| < η} and η = η21 .
Since g is continuous on compact set it is uniformly continuous and hence, for η there exists
some δ > 0 such that whenever ||s − t|| < δ, ||g(s) − g(t)|| < η.
From Lemma 3.2.2 we get that lim meshK (n) = 0. So, for δ > 0, there exists some n0 ∈ N,
n→∞

such that for all n ≥ n0 , mesh K (n) < 2δ .
Choose any k > n0 . Let v ∈ vertK (k) and x, z ∈ ost(v). There exist σ, τ ∈ K (k) with v as a
common vertex and such that x ∈ int(σ) and y ∈ int(τ ). Then ||x−v|| < 2δ and ||z −v|| < 2δ ;
so that ||x − z|| < δ. So, ||g(x) − g(z)|| < η. i.e., g(x), g(z) ∈ Bη (g(x)). Hence, there is some
w ∈ vert(L) such that g(x), g(z) ∈ Bη (g(x)) ⊆ ost(w). In otherwords, for each v ∈ vertK (k) ,
g(ost(v)) ⊆ ost(w), for some w ∈ vert(L). i.e., K (k) and L are star related with respect to
g. Using Theorem 3.2.1, there exists a simplicial map f : K (k) → L such that the induced
continuous function fb is homotopic to g.
Chapter 4

Simplicial homology of Topological

Spaces

In Chapter 2, we have seen how simplicial homology groups can be computed for a (finite)
simplicial complex K. It has also been observed that homology groups do not depend on
the orientation of the complex and therefore, one may use any suitable orientation of K for
the purpose of computing its homology groups.
A topological space X is said to be triangulable if there exists some simplicial complex K
such that X is homeomorphic to |K|. It is quite clear that one topological space may possibly
have more than one triangulations. In such case, unless we could prove that Hq (K) ∼
= Hq (L),
for each q, where K and L are two triangulations of a space X, we won’t be able to define
the homology group of X by taking Hq (X) = Hq (K). In fact, we assume without proof the
statement of the following result for the time being,
Theorem 4.0.1. If a topological space X has two triangulations K and L then Hq (K) ∼ =
Hq (L), for each q.
(Note that we use the same notation ∼= to mean isomorphism of groups as well as homeo-
morphism of topological spaces. One may figure out from the context the associated meaning
of ∼
=)
In view of this theorem, we define homology groups of a topological space as follows:
Definition 4.0.1. If X is a topological space and K is any triangulation of X then for each
integer q,
Hq (X) = Hq (K).
So, for any topological space X if we can find a simplicial complex K whose homology
groups are known and |K| is homeomorphic to X then we are done. In what follows we find
triangulations of some well known spaces.

4.1 Triangulation and Triangulable spaces

Our target is to find those topological spaces which are topologically equivalent to the geo-
metric carrier of some (finite) simplicial complex. It is amply clear that we have to search
within the collection of compact metrizable spaces.
31
A. Deb Ray, Pure Math, C.U. 32

Definition 4.1.1. A topological space X is said to be triangulable if there exists a finite

simplicial complex K such that X is homeomorphic to the geometric carrier |K| of K.
Then K is called a triangulation of X and |K| is called a polytope of K and X is called
a polyhedron.

We discuss here triangulations of some known compact surfaces (i.e., compact topological
2-manifolds). A triangulated surface can be regarded as having been constructed by gluing
together the various triangles in a specific way. Triangulating a surface is thus similar to
fitting the jigsaw puzzle pieces together to get a nice picture! Two different triangles cannot
have the same set of vertices and therefore, we can specify completely a triangulation of a
surface by numbering the vertices, and then listing which triples of vertices form triangles.
Such a list of triangles completely determines the surface together with how they are joined
and hence give a triangulation up to homeomorphism.

Example 4.1.1. Consider X = {(cos θ, sin θ) : 0 ≤ θ ≤ π2 }. Choose a and b as the points

in X with θ = 0 and θ = π2 . Then K = {⟨a⟩ , ⟨b⟩ , ⟨a, b⟩} is a triangulation of X. Here the
homeomorphism ψ : X → |K| is given by (cos θ, sin θ) 7→ ( coscos θ
, sin θ ).
θ+sin θ cos θ+sin θ

Example 4.1.2. Consider a cylinder C = {(x, y, z) ∈ R3 : x2 + y 2 = 1 and − a ≤ z ≤ a}.

Then C is homeomorphic to the quotient of a rectangle {(x, y) : −1 ≤ x ≤ 1, −a ≤ y ≤ a}
with the identification (−1, y) ∼ (1, y), for all y ∈ [−a, a]. The following diagram does not
give a triangulation of the cylinder (why?):

Find a triangulation of the cylinder.

Example 4.1.3. A torus is achievable via the quotient of a rectangle {(x, y) : −1 ≤ x ≤

1, −a ≤ y ≤ a}, gluing its opposite edges, i.e., (1, y) ∼ (−1, y), for −a ≤ y ≤ a and
(x, a) ∼ (x, −a), for −1 ≤ x ≤ 1. So, a triangulation of a torus is described by the figure
below:
A. Deb Ray, Pure Math, C.U. 33

Example 4.1.4. Consider a mobius band M . M can be viewed as a quotient space of a

rectangle {(x, y) : −1 ≤ x ≤ 1, −a ≤ y ≤ a} with the identification (−1, y) ∼ (1, −y), for
all y ∈ [−a, a]. The following diagram gives a triangulation of the mobius band:

We can show (check) that the following is not a triangulation of a mobius band:

The following is a triangulation of the real projective plane RP 2 :

But the following is not a triangulation of RP 2 :

It is not hard to realize that one may have several triangulations for a particular topological
space. So, given a topological space X, one may get two simplicial complexes K1 and K2
such that X ∼ = |K1 | ∼
= |K2 |.
A. Deb Ray, Pure Math, C.U. 34

4.2 Homology groups of S n and Dn

Before finding out the homology groups of S n and Dn , for any n, we visualise the trian-
= cl(σ 2 ) and S 2 ∼
gulations of D2 and S 2 . It is not hard to realize that D2 ∼ = sk(σ 3 ). So,
n ∼ n n ∼ n+1
generalising this idea, we certainly get, D = cl(σ ) and S = sk(σ ). It is proved in
Chapter 2 that for each n > 0,
(
Z, if q = 0 or n
Hq (sk(σ n+1 )) =
0, otherwise.

and (
Z, if q = 0
Hq (cl(σ n )) =
0, otherwise
Hence, (
Z, if q = 0 or n
Hq (S n ) =
0, otherwise.
and (
Z, if q = 0
Hq (Dn ) =
0, otherwise

Exercise 4.2.1. Calculate homology groups of a cylinder and a möbius band.

ALGEBRAIC TOPOLOGY II
(Singular Homology)

BY

A. DEB RAY

DEPARTMENT OF PURE MATHEMATICS

UNIVERSITY OF CALCUTTA
35, Ballygunge Circular Road
KOLKATA - 700019, INDIA
e-mail : [email protected]
Chapter 5

Singular Homology

In the previous chapters we have seen how simplicial homology groups can be constructed
for ‘triangulable’ spaces. Though simplicial homology groups are easy to calculate, simplicial
theory has an inherent limitation. Let us look back and scrutinize the results that we have
obtained so far, in simplicial homology.

ˆ We have taken a space X such that X ∼

= |K|, for some simplicial complex K.

ˆ Hn (K), for each integer n, were defined.

ˆ Every simplicial map K → L, induces continuous function |K| → |L| very naturally.

ˆ It has been proved that, given a continuous function f : X → Y , where X ∼ = |K| and
Y ∼= |L|, we can construct (in view of ‘simplicial approximation theorem’) a typical
simplicial complex K (n) (known as nth barycentric subdivision of K) and a simplicial
map g : K (n) → L such that the induced continuous function |g| : |K (n) | = |K| ∼
=X→
∼
Y = |L| is homotopic to f .

ˆ A simplicial map g : K → L induces homomorphisms Hq (g) : Hq (K) → Hq (L), for

each q.

But the question remains : Does the continuous function f : X → Y induce any homomor-
phism Hq (f ) : Hq (X) → Hq (Y )?
Or, one may put the question more precisely : For a continuous function f , can we define
Hq (f ) by taking the homomorphisms Hq (g), induced by the simplicial approximation g of
f ? This, certainly, is non-trivial and therefore, not coming very naturally in our mind.
So, it is quite expected from the mathematicians to rethink the process of building homology
theory, by taking into consideration the topological space X itself, instead of its triangula-
tion. In doing so, we can get rid of the limitation of taking ‘triangulable spaces’ and may
construct homology groups of any topological space. However, computation may become
more complicated in this new setting.

In this chapter, we shall learn how the idea of simplicial homology has been gradually
extended to achieve singular homology theory.
1
A. Deb Ray, Pure Math, C.U. 2

5.1 Singular Homology Groups

We define a particular type of n-simplex known as standard n-simplex as follows:
Definition 5.1.1. A standard n-simplex ∆n ⊆ Rn+1 is an n-simplex with vertices (1, 0, . . . , 0),
(0, 1, . . . , 0), . . . , (0, 0, . . . , 1), i.e.,
( n
)
X
∆n = (x0 , x1 , . . . , xn ) ∈ Rn+1 : xi = 1, xi ≥ 0, i = 0, 1, . . . , n
i=0

If K is a (finite) simplicial complex, then any n-simplex can be viewed as a copy of ∆n , and
therefore, we may re-define an n-simplex by saying that it is a continuous injective function
∆n → |K|. In that case, K becomes a list of continuous injections from ∆n to |K|. We
generalize this idea to obtain the definition of a singular n-simplex :
Definition 5.1.2. Let X be a topological space. Any continuous function ∆n → X is called
a singular n-simplex. The collection of all singular n-simplexes on X is denoted by Sn (X).
From Definition 5.1.2 it is clear that there is no need for the topological space X to be
triangulable. Also we note that a singular n-simplex is not in general injective and so, such
a map need not preserve the topology of ∆n . For instance, the following are possible images
of singular1-simplexes, though they are not simplicial 1-simplexes.

We have seen earlier that, intuitively, geometric carrier of a simplicial complex K is achieved
via gluing the simplexes along their boundaries suitably. To extend this idea of formation
of |K| in singular case, we need to define boundaries of singular n-simplexes suitably.
We define for each i = 0, 1, . . . , n, di : Rn → Rn+1 by
di (x0 , . . . , xn−1 ) = (x0 , . . . , xi−1 , 0, . . . , xn−1 ).
We observe that di (∆n−1 ) ⊆ ∆n , and hence obtain di |∆n−1 : ∆n−1 → ∆n . In what follows,
we shall refer to this restriction map di |∆n−1 simply by di . Further, to avoid notational
complexity, we use the same notation di to refer to all the maps ∆n → ∆n+1 that inserts 0
in its i-th co-ordinate, for all n ≥ 0 and understand its domain from the context. Note that,
for di ◦ dj : ∆n−1 → ∆n+1 ,
(
(x0 , . . . , xj−1 , 0, xj , . . . , xi−2 , 0, xi−1 , . . . , xn−1 ) if i > j
di ◦ dj (x0 , x1 , . . . , xn−1 ) = (5.1)
(x0 , . . . , xi−1 , 0, xi , . . . , xj−1 , 0, xj , . . . , xn−1 ) if i ≤ j
Since a (n − 1)-face of ∆n consists of points all of which have 0 in exactly one specific co-
ordinate, we see that di (∆n−1 ) is a (n − 1)-face of ∆n and so, the boundary of ∆n , denoted
by ∂∆n is given by d0 (∆n−1 ) ∪ d1 (∆n−1 ) ∪ . . . ∪ dn (∆n−1 ). i.e., ∂∆n consists of all points in
∆n with at least one co-ordinate 0. So, for a singular n-simplex f : ∆n → X, we get (n+1)
different (n − 1)-simplexes f ◦ di : ∆n−1 → X. These observations lead to the definition of
singular chain groups and boundary maps.
A. Deb Ray, Pure Math, C.U. 3

Definition 5.1.3. Let Cn (X) be the free abelian group on the set Sn (X) of singular n-
simplexes (n ≥ 0). i.e., Cn (X) is the collection of all Z-linear combinations of singular
n-simplexes. The elements of Cn (X) are known as singular n-chains. If n < 0 then define
Cn (X) as the trivial group, denoted by 0.

Definition 5.1.4. For each n ≥ 1, define δn : Cn (X) → Cn−1 (X) by

n
X
δn (f ) = (−1)i f ◦ di
i=0

for each singular n-simplex f : ∆n → X and extend it by linearity on Cn (X). If n ≤ 0 then

δn ’s are taken as the trivial maps. For any n ∈ Z, these δn ’s are group homomorphisms,
called boundary homomorphisms.

The rest of the construction of homology groups is similar to that of simplicial homology
groups.

Theorem 5.1.1. For each n, δn ◦ δn+1 = 0.

Proof. It is enough to prove (δn ◦ δn+1 )(f ) = 0, for any singular (n + 1)-simplex f . From
(5.1) it follows that
(
dj ◦ di−1 , if i > j
di ◦ dj = (5.2)
dj+1 ◦ di , if i ≤ j

So using (5.2), we obtain,

n n+1
!
X X
δn ◦ δn+1 (f ) = (−1)j (−1)i f ◦ di ◦ dj
j=0 i=0
n X
X n+1
= (−1)i+j f ◦ (di ◦ dj )
j=0 i=0
j
n X n X
n+1
X X
i+j
= (−1) f ◦ (di ◦ dj ) + (−1)i+j f ◦ (di ◦ dj )
j=0 i=0 j=0 i=j+1
= 0

Corollary 5.1.1. For each n, Im(δn+1 ) ⊆ Ker(δn ).

Proof. If f ∈ Im(δn+1 ), there exists g such that f = δn+1 (g) which implies that δn (f ) =
(δn ◦ δn+1 )(g) = 0. Hence f ∈ Ker(δn ).

We define the terms chain complex, cycles, boundaries and finally, singular homology groups
exactly the same way as those defined for simplicial homology.
A. Deb Ray, Pure Math, C.U. 4

Definition 5.1.5. Let X be a topological space. Consider the doubly infinite sequence
{Cn (X), δn } of groups and boundary maps, called a singular chain complex,
δn+1 n δ
. . . Cn+1 (X) → Cn (X) → Cn−1 (X) → . . .

where Cn (X) is the free abelian group on Sn (X), for n ≥ 0 and trivial group otherwise.
Define for each n ≥ 0, Bn (X) = Im(δn+1 ) and Zn (X) = Ker(δn )

1. The Elements of Zn (X) are called cycles (or, n-cycles)

2. The elements of Bn (X) are called boundaries (or, n-boundaries)

3. (
Zn (X)/Bn (X), if n ≥ 0
Hn (X) =
0, otherwise
are called the singular homology groups.

Example 5.1.1. Suppose X is a singleton, i.e., X = {x}. Then for each n ≥ 0, Sn (X) =
{f : ∆n → X : f (t) = x, ∀t ∈ ∆n } = {Cx }, where Cx : ∆n → X is the constant map t 7→ x.
By Definition of Cn (X) (n ≥ 0), it is the free abelian group generated by Cx , and hence it
is ∼
= Z and trivial for all n ≤ 0.
We observe that Cx ◦ d0 = Cx ◦ d1 = . . . = Cx ◦ dn = constant function. Consequently, for
n ≥ 1, (
n
X Cx ◦ d0 , n is even
δn (Cx ) = (−1)i (Cx ◦ di ) =
i=0
0, n is odd
So, for each n ≥ 1, (
0, if n is even
Zn (X) = Ker(δn ) =
Z, if n is odd
and Z0 (X) = C0 (X). Also, for each n ≥ 0
(
0, if n is even
Bn (X) = Im(δn+1 ) =
{Cx ◦ d0 : Cx ∈ Cn+1 (X)} ∼
= Z, if n is odd

Hence, for all n > 0, Hn (X) = Zn (X)/Bn (X) = 0. But H0 (X) = C0 (X)/B0 (X) = Z/0 =
∼
= Z. Consequently, (
Z, if n = 0
Hn (X) =
0, otherwise

Example 6.2.1 can be generalized to any discrete space consisting of finitely many (say, r
many) points, as seen next.

Example 5.1.2. Suppose X = {x1 , . . . , xr }. Then for any n ≥ 0, Sn (X) = {f : ∆n → X :

f is continuous}. Using the fact that the continuous image of a connected space is connected
and that ∆n ’s are all connected, Sn (X) = {fx1 , . . . , fxr }, where each fxi is the constant map
A. Deb Ray, Pure Math, C.U. 5

t 7→ xi . Hence Cn (X) is a free abelian group of rank r, i.e., Cn (X) = ⊕ri=1 Zfxi ∼
= ⊕ri=1 Z. As
in Example 6.2.1, f ◦ d0 = f ◦ d1 = . . . = f ◦ dn , for each f ∈ Sn (X). So, for n ≥ 1,
n
(
X f ◦ d0 , n is even
δn (f ) = (−1)i (f ◦ di ) =
i=0
0, n is odd

Therefore, proceeding exactly as before, Ker(δn ) = Im(δn+1 ), for each n ≥ 1, giving all
Hn (X) = 0 except H0 (X), which in this case is a free abelian group of rank r. i.e.,
(
Z ⊕ Z ⊕ . . . ⊕ Z (r copies), if n = 0
Hn (X) =
0, otherwise

In Example 5.1.2, we observe that the underlying space X is a disconnected space with r
path components and the 0-th homology group is the free abelian group whose rank is ex-
actly the number of path components of X. The earlier example is just a special case of this
one, because, a singleton set has only one path component. These two examples inspire us to
expect that the 0-th singular homology group might count the number of path components
of the topological space X. Following theorem asserts that our guess is absolutely right. In
fact, the 0-th singular homology groups provide information about path components of X.

For 0-th Homology groups, the following facts are known:

ˆ Consider
δ1
. . . C1 (X) → C0 (X) → 0
where C0 (X) is the free abelian group on S0 (X) = {Cx : x ∈ X}, where Cx (1) = x.
(Note that ∆0 = {1}).
C1 (X) is the free abelian group on S1 (X). (Note that identifying ∆1 with [0, 1], we see
that the members of S1 (X) are the paths in X)

ˆ H0 (X) = Z0 (X)/B0 (X) = C0 (X)/B0 (X).

ˆ B0 (X) = Im(δ1 ) = {δ1 (f ) : f ∈ C1 (X)} = {f ◦ d0 − f ◦ d1 : f ∈ C1 (X)}.

Define augmentation map ϵ : C0 (X) → Z by

ϵ(s) = 1
P P
for s ∈ S0 (X) and extend it by linearity on C0 (X). Then ϵ(c) = ni , where c = ni s i ,
si ∈ S0 (X). Clearly, ϵ is a homomorphism. For any σ ∈ S1 (X), ϵ(δ1 (σ)) = ϵ(σ ◦d0 −σ ◦d1 ) =
ϵ(σ ◦ d0 ) − ϵ(σ ◦ d1 ) = 1 − 1 = 0.
Define augmentation map ϵ∗ : H0 (X) → Z by

ϵ∗ (z + B0 (X)) = ϵ(z)

Then ϵ∗ is well defined, because, z1 +B0 (X) = z2 +B0 (X) ⇒ z1 −z2 ∈ B0 (X) ⇒ ϵ(z1 −z2 ) = 0.
Also ϵ∗ is a homomorphism. In fact, it is a surjective homomorphism.
A. Deb Ray, Pure Math, C.U. 6

Let X be a path connected space. We show that ϵ∗ : H0 (X) → Z is an isomorphism. It is

seen that ϵ∗ is in general a surjective homomorphism. So, we now prove that Ker(ϵ∗ ) = 0.
Once we prove Ker(ϵ∗ ) = 0, we establish that

H0 (X) ∼
= Z.

Take z + B0 (X) ∈ Ker(ϵ∗ ).

ϵ∗ (z + BP0 (X)) = 0 ⇒ ϵ(z) = 0 ⇒ z ∈ Ker(ϵ).
Let z = nx Cx . Fix a point x0 ∈ X. Let λx be a path from x0 to x. Identifying the domain
of λx with ∆1 , we get λx : ∆1 → X is a continuous function such that λx (1, 0) = x0 and
λx (0, 1) = x.
δ1 (λx ) = λx ◦ d0 − λx ◦ d1 .

X X X
z − δ1 ( nx λx ) = nx Cx − nx δ1 (λx )
X
= nx (Cx − δ1 (λx ))

where (Cx − = Cx (1) − (λx ◦ d0 )(1) + (λx ◦ d1 )(1) =

Pδ1 (λx ))(1) P Px − x + x0 = x0 = Cx0 (1).
So, z − δ1 ( nx λx ) = ( nx )Cx0 = ϵ(z)Cx0 = 0. i.e., z = δ1 ( nx λx ) ∈ Im(δ1 ) = B0 (X).
Hence, z + B0 (X) = B0 (X).

Theorem 5.1.2. Let X be a topological space with path components Yj , j ∈ Λ (Λ is an

index set). For each n ≥ 0,
Hn (X) = ⊕j∈Λ Hn (Yj )

Proof. We first show that Cn (X) = ⊕m j=1 Cn (Yj ).

Let f ∈ Sn (X). Then f : ∆n → X is a continuous function. Since ∆n is path connected and
continuous image of a connected space is connected, f (∆n ) ⊆ Yj , for some j = 1, 2, . . . , m.
Pk
So, f ∈ Sn (Yj ). Take any s ∈ Cn (X). Then s = ni fi , where fi ∈ Sn (Yj ).
! i=1
m
P P m
P
i.e., s = ni fi , so that s = sj , where sj ∈ Cn (Yj ). Clearly, Sn (Yi ) ∩ Sn (Yj ) = ∅.
j=1 fi ∈Yj j=1
We claim that Cn (Yi ) ∩ Cn (Yj ) = {0}. P P
If possible, let x ∈ Cn (Yi ) ∩ Cn (Yj ). Then x = nk f k = tk gk where fk ∈ Sn (Yi ) and
k k
gk ∈P Sn (Yj ). P
i.e., nk fk + (−tk )gk = 0.
k k P P
Since 0 is in both the free abelian groups Cn (Yi ) and Cn (Yj ), we get nk fk + (−tk )gk ∈
k k
Cn (Yi ) ∩ Cn (Yj ). Hence, nk = 0 and tk = 0, for all k. So, Cn (Yi ) ∩ Cn (Yj ) = {0}.
Pm
We claim that z + Bn (X) = (zj + Bn (Yj )), where zj ∈ Z(Yj ), for all j = 1, 2, . . . , m. Since
j=1
A. Deb Ray, Pure Math, C.U. 7

m
P
z ∈ Zn (X) ⊆ Cn (X), z = zj , where zj ∈ Cn (Yj ), for each j = 1, 2, . . . , m.
j=1

Xm
δn (z) = 0 ⇒ δn ( zj ) = 0
j=1
m
X
⇒ δn (zj ) = 0
j=1
⇒ δn (zj ) = 0
⇒ zj ∈ Zn (Yj )
m
P
Now, b ∈ Bn (X) implies that b = δn+1 (a) for some a ∈ Cn+1 (X). Then a = aj , where
j=1
m
P m
P
aj ∈ Cn+1 (Yj ). So, δn+1 (a) = δn+1 ( aj ) = δn+1 (aj ).
j=1 j=1
i.e., b is uniquely expressible as linear sum of elements of Bn (Yj ). Hence, z + Bn (X) =
Pm
(zj + Bn (Yj )) uniquely. So,
j=1
Hn (X) = ⊕m
j=1 Hn (Yj ).

Corollary 5.1.2. If X is path connected then H0 (X) is Z.

Corollary 5.1.3. If X has m path components then

H0 (X) ∼
= ⊕m
i=1 Z.

5.2 Induced Homomorphism

Given any continuous function f : X → Y and any s ∈ Sn (X), the composite f ◦ s : ∆n → Y
is a continuous function and therefore f ◦s ∈ Sn (Y ). So, each continuous function f : X → Y
induces a map Sn (X) → Sn (Y ) given by s 7→ f ◦ s. Since Cn (X) and Cn (Y ) are free abelian
groups on Sn (X) and Sn (Y ) respectively, we get a homomorphism Cn (f ) : Cn (X) → Cn (Y )
described on the generators by Cn (f )(s) = f ◦ s, for s ∈ Sn (X) and extending it linearly on
the whole of Cn (X). i.e.,

Theorem 5.2.1. Let X and Y be topological spaces. Any continuous function f : X → Y

induces a homomorphism Cn (f ) : Cn (X) → Cn (Y ) given by
X X
Cn (f )( sj ) = f ◦ sj

Proof. Follows from the above discussion.

A. Deb Ray, Pure Math, C.U. 8

Theorem 5.2.2. Let X and Y be topological spaces and f : X → Y be a continuous

function. If {Cn (X), δn } and {Cn (Y ), ∂n } are the respective chain groups then the following
diagram commutes, i.e., Cn−1 (f ) ◦ δn = ∂n ◦ Cn (f ).

Cn (X) / Cn−1 (X)

δn
Cn (f ) Cn−1 (f )
 
Cn (Y ) / Cn−1 (X)
∂n

Proof. Since Sn (X) generates the free abelian group Cn (X) and Cn (f ), δn and ∂n are ho-
momorphisms, it is enough to prove the result on the set of generators. Thus, we take any
s ∈ Sn (X) and show that (Cn−1 (f ) ◦ δn )(s) = (∂n ◦ Cn (f ))(s) holds.

(∂n ◦ Cn (f ))(s) = ∂n (Cn (f )(s))

= ∂n (f ◦ s)
Xn
= (−1)i (f ◦ s) ◦ di
i=0
n
!
X
i
= Cn−1 (f ) (−1) (s ◦ di )
i=0
= Cn−1 (f ) ◦ δn (s).

Corollary 5.2.1. For each n,

(i) Cn (f )(Zn (X)) ⊆ Zn (Y ).
(ii) Cn (f )(Bn (X)) ⊆ Bn (Y ).

Proof. If n ≤ 0, the results are trivially true. So, we assume n > 0.

(i) Let z ∈ Zn (X). Then δn (z) = 0 and so by the above theorem, ∂n (Cn (f )(z)) = (∂n ◦
Cn (f ))(z) = (Cn−1 (f ) ◦ δn )(z) = 0. i.e., Cn (f )(z) ∈ Zn (Y ), as desired.
(ii) Let b ∈ Bn (X). Then there exists some c ∈ Cn+1 (X) such that δn+1 (c) = b. So, using
the theorem, Cn (f )(b) = (Cn (f ) ◦ δn+1 )(c) = (∂n+1 ◦ Cn+1 (f ))(c) = ∂n+1 (Cn+1 (f )(c)). i.e.,
Cn (f )(b) ∈ Bn (Y ).

Corollary 5.2.1 helps in defining a homomorphism between Hn (X) and Hn (Y ) corre-

sponding to a continuous function f : X → Y .

Theorem 5.2.3. A continuous function f : X → Y defines a homomorphism Hn (f ) :

Hn (X) → Hn (Y ) given by

Hn (f )(z + Bn (X)) = Cn (f )(z) + Bn (Y ).

for each n ≥ 0.
A. Deb Ray, Pure Math, C.U. 9

Proof. By (i) of Corollary 5.2.1 it is clear that image of any point of Hn (X) under the map
Hn (f ) is inside Hn (Y ). That Hn (f ) is well-defined, follows from Corolloary 5.2.1(ii), because,

z1 + Bn (X) = z2 + Bn (X) ⇒ z1 − z2 ∈ Bn (X)

⇒ Cn (f )(z1 − z2 ) ∈ Bn (Y )
⇒ Cn (f )(z1 ) + Bn (Y ) = Cn (f )(z2 ) + Bn (Y )

It is now to show that Hn (f ) is a homomorphism, for every n ≥ 0.

Hn (f )(z1 + Bn (X) + z2 + Bn (X)) = Hn (f )((z1 + z2 ) + Bn (X))

= Cn (f )(z1 + z2 ) + Bn (Y )
= (Cn (f )(z1 ) + Bn (Y )) + (Cn (f )(z2 ) + Bn (X))
= Hn (f )(z1 + Bn (X)) + Hn (f )(z2 + Bn (X))

The following theorem shows that the induced homomorphisms satisfy functorial prop-
erties:
Theorem 5.2.4. For each n ≥ 0,
(1) If I : X → X is the identity map then Hn (I) : Hn (X) → Hn (X) is the identity
homomorphism.
(2) If f : X → Y and g : Y → Z are continuous functions then Hn (g ◦ f ) = Hn (g) ◦ Hn (f ).
Proof. (1) By definition, Hn (I) : Hn (X) → Hn (X) is such that Hn (I)(z + Bn (X)) =
Cn (I)(z) + Bn (X) where Cn (I)(s) = I ◦ s for each s ∈ Sn (X). But (I ◦ s)(t) = s(t), for all
t ∈ ∆n and so, I ◦ s = s, for each s ∈ Sn (X). As Sn (X) generates Cn (X), Cn (I)(ϕ) = ϕ,
for all ϕ ∈ Cn (X). i.e., Hn (I)(z + Bn (X)) = z + Bn (X), proving Hn (I) is the identity
homomorphism.
(2) For continuous functions f : X → Y and g : Y → Z,

Hn (g ◦ f )(z + Bn (X)) = Cn (g ◦ f )(z) + Bn (Z)

= ((g ◦ f ) ◦ z) + Bn (Z)
= (g ◦ (f ◦ z)) + Bn (Z)
= Cn (g)(f ◦ z) + Bn (Z)
= Cn (g)(Cn (f )(z)) + Bn (Z)
= Hn (g)(Cn (f )(z) + Bn (Y ))
= Hn (g)(Hn (f )(z + Bn (X)))
= (Hn (g) ◦ Hn (f ))(z + Bn (X))

for each n ≥ 0. i.e., Hn (g ◦ f ) = Hn (g) ◦ Hn (f ).

A very important consequence of the properties (i) and (ii) follow next:
Theorem 5.2.5. If f : X → Y is a homeomorphism then Hn (f ) : Hn (X) → Hn (Y ) is an
isomorphism.
A. Deb Ray, Pure Math, C.U. 10

Proof. Since f is a homeomorphism, there exists a continuous function g : Y → X such that

g ◦ f = IX and f ◦ g = IY . Hence, using the functorial properties, for each n,

Hn (g) ◦ Hn (f ) = Hn (g ◦ f ) = Hn (IX ) = IHn (X)

and
Hn (f ) ◦ Hn (g) = Hn (f ◦ g) = Hn (IY ) = IHn (Y )
This shows that the homomorphism Hn (f ) (for each n) is invertible, i.e., Hn (f ) : Hn (X) →
Hn (Y ) is an isomorphism.
If we take a homotopy equivalence f : X → Y instead of a homeomorphism then is Hn (f ) :
Hn (X) → Hn (Y ) an isomorphism? To answer this question, we need to know what happens
to the induced homomorphisms corresponding to homotopic maps.

5.3 Homology respects Homotopy

Our aim in this section is to establish that homotopic maps induce the same homomorphism.
In other words, if f ∼= g : X → Y then Hn (f ) = Hn (g). Before this, we compute the
homology groups of any convex subset X.
k
Observation 5.3.1. Let z ∈ ∆k , k > 0. Then z = (x0 , . . . , xk ) with
P
xi = 1. If xk = 0
i=0
then z = (x0 , x1 , . . . , xk−1 , 0). If xk = 1 then z = (0, 0, . . . , 0, 1). If xk ̸= 0, 1, then
x0 x1 xk−1
z = t( , ,..., , 0) + (1 − t)(0, 0, . . . 1)
1 − xk 1 − xk 1 − xk
where 0 < t = 1 − xk < 1.
In other words, any z ∈ ∆k can be viewed as a point lying on a line joining a point
(y0 , y1 , . . . , yk−1 , 0) ∈ dk (∆k−1 ) and (0, 0, . . . , 1).
Example 5.3.1. Let X be a convex set. Then
(
Z, if n = 0
Hn (X) =
0, otherwise

Since X is a convex set, it is path connected and hence, H0 (X) = Z.

Let f ∈ Sk (X) and x ∈ X. For any point z ∈ ∆k+1 choose the end point (x0 , x1 , . . . , xk , 0) ∈
dk+1 (∆k ) such that z lies on the line joining (0, 0, . . . , 1) and (x0 , x1 , . . . , xk , 0). Define Define
Cx (f )(z) = (1 − t)x + tf (x0 , . . . , xk ). Then Cx (f ) : ∆k+1 → X is a continuous function. So,
we get an assignment f 7→ Cx (f ) from Sk (X) to Sk+1 (X). Extending additively we get a
homomorphism Cx : Ck (X) → Ck+1 (X). It is easy to see that

δk+1 (Cx (f )) = Cx (δk (f )) + (−1)k+1 (f ).

Therefore, f ∈ Zk (X) ⇒ δk (f ) = 0 ⇒ δk+1 (Cx (f )) = (−1)k+1 (f ). i.e., f ∈ Bk (X). Hence,

Hk (X) = Zk (X)/Bk (X) = 0, for all k ≥ 1.
A. Deb Ray, Pure Math, C.U. 11

As a special case of this example we get the following:

Example 5.3.2. For any n,
(
Z, if k = 0
Hk (Dn ) =
0, otherwise
We first prove two technical lemmas without explaining their background and it will be
eventually understood why we need them for establishing ‘homology respects homotopy’. In
what follows, we use the following notations:
ˆ id : I → I is the identity function from I = [0, 1] to itself;
ˆ Gn ∈ Cn (∆n × I) is defined by Gn = e1 − e0 , where e0 , e1 : ∆n → ∆n × I are the
singular n-simplexes e0 (z) = (z, 0) and e1 (z) = (z, 1), for each n ≥ 0.
Lemma 5.3.1. For each n ≥ 0, there exists αn : ∆n+1 → ∆n × I, such that when n > 0,
n
X
δn+1 (αn ) + (−1)j (dj × id) ◦ αn−1 = Gn
j=0

and when n = 0, δ1 (α0 ) = G0 .

Proof. Define α0 : ∆1 → ∆0 × I by
α0 (x0 , x1 ) = (1, x1 ).
Since (x0 , x1 ) ∈ ∆1 , it follows that 0 ≤ x1 ≤ 1. Also,
δ1 (α0 )(1) = α0 ◦ d0 (1) − α0 ◦ d1 (1)
= α0 (0, 1) − α0 (1, 0)
= (1, 1) − (1, 0)
= e1 (1) − e0 (1)
= G0 (1)
Hence, we have constructed α0 satisfying δ1 (α0 ) = G0 .
To get αn (n > 0) satisfying the given condition, we apply mathematical induction. We
observe that
1
! 1 1
!
X X X
δ1 (−1)j (dj × id) ◦ α0 = (−1)k (−1)j (dj × id) ◦ α0 ◦ dk
j=0 k=0 j=0
1 1
!
X X
= (−1)j (dj × id) (−1)k α0 ◦ dk
j=0 k=0
1
X
= (−1)j (dj × id) ◦ δ1 (α0 )
j=0
1
X
= (−1)j (dj × id) ◦ G0
j=0
A. Deb Ray, Pure Math, C.U. 12

Therefore, δ1 ( 1j=0 (−1)j (dj × id) ◦ α0 )(1) = (d0 × id)(G0 (1)) − (d1 × id)(G0 (1)) = (0, 1, 1) −
P
(0, 1, 0) − (1, 0, 1) + (1, 0, 0) δ1 (G1 (1)) = G1 (d0 (1)) − G1 (d1 (1)) = (0, 1, 1) − (0, 1, 0) −
 and P
(1, 0, 1) + (1, 0, 0). So, δ1 G1 − 1j=0 (−1)j (dj × id) ◦ α0 = 0. i.e., G1 − 1j=0 (−1)j (dj ×
P

id) ◦ α0 ∈ Ker(δ1 ). Since ∆1 × I is a convex set, Hk (∆1 × I) = 0 for all k > 0. So,
Ker(δ1 ) = Im(δ2 ) and consequently, there exists some α1 : ∆2 → ∆1 × I such that

1
X
δ2 (α1 ) + (−1)j (dj × id) ◦ α0 = G1 .
j=0

Suppose that we have already obtained αn−1 : ∆n → ∆n−1 × I satisfying the condition
n−1
X
δn (αn−1 ) + (−1)j (dj × id) ◦ αn−2 = Gn−1 .
j=0

Then
n
! n n
!
X X X
δn (−1)i (di × id) ◦ αn−1 = (−1)k (−1)i (di × id) ◦ αn−1 ◦ dk
i=0 k=0 i=0
n
X
= (−1)i (di × id) ◦ δn (αn−1 )
i=0
n n−1
!
X X
= (−1)i (di × id) ◦ Gn−1 − (−1)j (dj × id) ◦ αn−2
i=0 j=0
Xn
= (−1)i (di × id) ◦ Gn−1
i=0
n X
X n−1
− (−1)i+j (di × id) ◦ (dj × id) ◦ αn−2
i=0 j=0
n
X
= (−1)i (di × id) ◦ Gn−1
i=0
n X
X n−1
− (−1)i+j ((di ◦ dj ) × id) ◦ αn−2
i=0 j=0
n
X
= (−1)i (di × id) ◦ Gn−1 ,
i=0

zero. Again, ni=0 (−1)i (di × id) ◦ Gn−1 =

P
since,
Pn the isecond term of the expression becomes Pn i
i=0 (−1) (e1 − e0 ) ◦ di = δn (Gn ). So, Gn − i=0 (−1) (di × id) ◦ αn−1 ∈ Ker(δn ) = Im(δn+1 ),
n
as ∆ × I is convex and its n-th Pnhomology group is trivial. Hence, there exists some αn ∈
n i
Cn+1 (∆ × I) such that Gn − i=0 (−1) (di × id) ◦ αn−1 = δn+1 (αn ), which completes the
proof.
A. Deb Ray, Pure Math, C.U. 13

Lemma 5.3.2. If f , g are two continuous maps from X into Y which are homotopic, then
there exists a family {Φn : Cn (X) → Cn+1 (Y )}n≥0 of homomorphisms such that for n > 0

∂n+1 Φn + Φn−1 δn = Cn (g) − Cn (f )

and
∂1 Φ0 = C0 (g) − C0 (f ).
i.e., In diagrams,
δn+1 δn
Cn+1 (X) / Cn (X) / Cn−1 (X) and C0 (X)
0 / 0

y Φn  y Φn−1 z Φ0  } 0
Cn+1 (Y ) ∂ / Cn (Y ) / Cn−1 (Y ) C1 (Y ) / C0 (Y ) / 0
n+1 ∂n ∂1 0

Proof. Let F : X × I → Y be a homotopy from f to g. If σ ∈ Sn (X) ⊆ Cn (X), then

σ × id : ∆n × I → X × I is a continuous function. In view of Lemma 5.3.1, there is some
αn : ∆n+1 → ∆n × I, such that, for n > 0
n
X
δn+1 (αn ) + (−1)j (dj × id) ◦ αn−1 = Gn
j=0

and for n = 0, δ1 (α0 ) = G0 . So, for every n ≥ 0 we have the following composite map in
Cn+1 (Y ):
αn σ×id F
∆n+1 −→ ∆n × I −→ X × I −→ Y
For each n ≥ 0, define Φn : Cn (X) → Cn+1 (Y ) by Φn (σ) = F ◦ (σ × id) ◦ αn , for each
σ ∈ Sn (X) and extending it by linearity on the whole of Cn (X). Each Φn is indeed a
homomorphism. We claim that such a sequence satisfies the desired condition. If n > 0, for
any σ ∈ Sn (X),

(∂n+1 Φn + Φn−1 δn )(σ) = ∂n+1 (F ◦ (σ ◦ id) ◦ αn ) + F ◦ (δn (σ) × id) ◦ αn−1

n+1
X X n
= (−1)j (F ◦ (σ × id) ◦ αn ◦ dj ) + (−1)j F ◦ ((σ ◦ dj ) × id) ◦ αn−1
j=0 j=0
n+1 n
!
X X
= F ◦ (σ × id) ◦ (−1)j αn ◦ dj + (−1)j (dj × id) ◦ αn−1
j=0 j=0
n
!
X
= F ◦ (σ × id) ◦ δn+1 (αn ) + (−1)j (dj × id) ◦ αn−1
j=0
= F ◦ (σ × id) ◦ Gn
= F ◦ (σ × id) ◦ e1 − F ◦ (σ × id) ◦ e0

Similarly, if n = 0, then for any σ ∈ S0 (X), (∂1 Φ0 )(σ) = F ◦ (σ × id) ◦ G0 . For any z ∈ ∆n
(n ≥ 0), F (σ(z), 1) − F (σ(z), 0) = g(σ(z)) − f (σ(z)) so that for n > 0

(∂n+1 Φn + Φn−1 δn )(σ)(z) = g(σ(z)) − f (σ(z)) = (g ◦ σ)(z) − (f ◦ σ)(z).

A. Deb Ray, Pure Math, C.U. 14

and for n = 0

∂1 Φ0 (σ)(z) = g(σ(z)) − f (σ(z)) = (g ◦ σ)(z) − (f ◦ σ)(z).

i.e., ∂1 Φ0 (σ) = (C0 (g) − C0 (f ))(σ) and for n > 0, ∂n+1 Φn + Φn−1 δn = Cn (g) − Cn (f ), as
desired.

Definition 5.3.1. Let {Cn , δn }n≥0 and {Dn , ∂n }n≥0 be two chain complexes and ϕn , ψn :
Cn → Dn (n ≥ 0) be two chain maps. A family of homomorphisms {Φn : Cn → Dn+1 }n≥0 is
called a chain homotopy between {ϕn } and {ψn } if

∂n+1 Φn + Φn−1 δn = ψn − ϕn

for all n > 0 and

∂1 Φ0 = ψ0 − ϕ0 .

In view of this definition of chain homotopy, we may re-state Lemma 5.3.2 by saying that
two homotopic chain maps induce a chain homotopy between the given chain complexes.
Now we see how this existence of chain homotopy helps to prove our main result:

Theorem 5.3.1. (Homology respects Homotopy) If f, g : X → Y are continuous

functions such that f ∼
= g then Hn (f ) = Hn (g) for all n ≥ 0.

Proof. Since f and g are homotopic, there exists a chain homotopy {Φn : Cn (X) → Cn+1 (Y )}n≥0 .
If c ∈ C0 (X),

C0 (g)(c) − C0 (f )(c) = (∂1 Φ0 )(c) = ∂1 (Φ0 (c)) ∈ Im(∂1 ) = B0 (Y ).

Hence, H0 (g)(c + B0 (X)) = C0 (g)(c) + B0 (Y ) = C0 (f )(c) + B0 (Y ) = H0 (f )(c + B0 (X)), for

all c ∈ C0 (X). i.e., H0 (g) = H0 (f ).
Let n > 0 and choose z ∈ Zn (X) = Ker(δn ). Then

Cn (g)(z) − Cn (f )(z) = (∂n+1 ◦ Φn )(z) + (Φn−1 ◦ δn )(z)

= ∂n+1 (Φn (z)) + Φn−1 (0)
= ∂n+1 (Φn (z))
∈ Im(∂n+1 ) = Bn (Y )

Hence Hn (g)(z + Bn (X)) = Cn (g)(z) + Bn (Y ) = Cn (f )(z) + Bn (Y ) = Hn (f )(z + Bn (X))

and so, Hn (g) = Hn (f ), for all n > 0.

Finally, using Theorem 5.3.1 we show next that Hn (f ) is an isomorphism (for all n ≥ 0),
if f is a homotopy equivalence. In other words, topological spaces which are of the same
homotopy type have the same homology groups in each dimension.

Theorem 5.3.2. If f : X → Y is a homotopy equivalence then for all n ≥ 0, Hn (f ) :

Hn (X) → Hn (Y ) is an isomorphism.
A. Deb Ray, Pure Math, C.U. 15

Proof. Since f : X → Y is a homotopy equivalence, there exists g : Y → X such that

g◦f ∼= IX and f ◦ g ∼
= IY . By Theorem 5.3.1, Hn (g ◦ f ) = Hn (IX ) and Hn (f ◦ g) = Hn (IY ).
Applying functorial properties of induced homomorphisms, we get

Hn (g) ◦ Hn (f ) = Hn (g ◦ f ) = Hn (IX ) = IHn (X)

and
Hn (f ) ◦ Hn (g) = Hn (f ◦ g) = Hn (IY ) = IHn (Y )
So, Hn (f ) and Hn (g) are inverse of each other and hence, isomorphisms.
Chapter 6

Mayer-Vietoris Sequence and

Computation of Singular Homology
Groups

In this chapter we first prove Mayer-Vietoris Theorem and then compute homology groups
using Mayer-Vietoris sequence. This theorem guarantees the existence of a long exact se-
quence of homology groups for a topological space X which is expressible as union of two
of its open subspaces. Such a long exact sequence is known as Mayer-Vietoris sequence (in
short, M-V sequence).

6.1 Mayer Vietoris Theorem

In what follows, we use the following notations unless stated otherwise:

ˆ X stands for a topological space.

ˆ U , V are open subsets of X, with X = U ∪ V .

ˆ i : U → X, j : V → X, k : U ∩ V → U , l : U ∩ V → V are inclusion maps.

ˆ Hn (i) : Hn (U ) → Hn (X), Hn (j) : Hn (V ) → Hn (X), Hn (k) : Hn (U ∩ V ) → Hn (U )

and Hn (l) : Hn (U ∩ V ) → Hn (V ) are the induced homomorphisms from the respective
inclusion maps, for each n ≥ 0.

We define the following maps:

ˆ fn : Hn (U ) ⊕ Hn (V ) → Hn (X) by

fn (x, y) = Hn (i)(x) − Hn (j)(y)

ˆ gn : Hn (U ∩ V ) → Hn (U ) ⊕ Hn (V ) by

gn (x) = (Hn (k)(x), Hn (l)(x))

16
A. Deb Ray, Pure Math, C.U. 17

In what follows, for ease of notation, we shall denote these maps by f and g instead of fn
and gn and understand the domain from the context.

We state a homological analogue of Van Kampen’s theorem, without proof, that we shall
use frequently:
Lemma 6.1.1. (Chain Splitting) If s ∈ Cn (X) and X = U ∪ V where U and V are open
in X, then there are sU ∈ Cn (U ) and sV ∈ Cn (V ) such that s is homologous to sU + sV .
i.e., There is some chain t ∈ Cn+1 (X) such that s − sU − sV = δn+1 (t).
It is easy to observe that Hn (i) ◦ Hn (k) = Hn (j) ◦ Hn (l). In other words, the following
diagram commutes:
Hn (k)
Hn (U ∩ V ) / Hn (U )
Hn (j) Hn (i)
 
Hn (V ) / Hn (X)
Hn (l)

Hence,
f (g(x)) = f ((Hn (k)(x), Hn (l)(x))) = Hn (i)(Hn (k)(x)) − Hn (j)(Hn (l)(x)) = 0
Therefore, Im(g) ⊆ Ker(f ).

Now in this setting, we have the following results:

g f
Theorem 6.1.1. For each n ≥ 0, Hn (U ∩ V ) −→ Hn (U ) ⊕ Hn (V ) −→ Hn (X) is an exact
sequence.
Proof. For this sequence to be exact, we need to show Ker(f ) = Im(g). As discussed
above, Im(g) ⊆ Ker(f ). Choose any (c, d) ∈ Ker(f ). Then c ∈ Hn (U ) and d ∈ Hn (V )
such that f (c, d) = 0. Suppose, α ∈ Cn (U ) and β ∈ Cn (V ) are cycles that represent c, d
respectively. i.e., c = α + Bn (U ) and d = β + Bn (V ), where α ∈ Zn (U ) and β ∈ Zn (V ).
Now, f (c, d) = 0 ⇒ Hn (i)(c) − Hn (j)(d) is the null element of Hn (X), so that there exists
some t ∈ Cn+1 (X), with Cn (i)(α) − Cn (j)(β) = δn+1 (t). By chain splitting, there are chains
tU ∈ Cn+1 (U ) and tV ∈ Cn+1 (V ) such that t is homologous to tU + tV . So, t − tU − tV =
δn+2 (a) for some a ∈ Cn+2 (X). i.e., δn+1 (t − tU + tV ) = δn+1 (δn+2 (a)) = 0. Therefore,
δn+1 (t) = δn+1 (tU ) + δn+1 (tV ). Hence, δn+1 (tU ) + δn+1 (tV ) = Cn (i)(α) − Cn (j)(β) = α − β,
i.e.,
α − δn+1 (tU ) = β + δn+1 (tV ).
The n-chain on the left of the expression is a chain on U and the n-chain on the right of the
expression is chain on V . So the last equation suggests that α−δn+1 (tU ) and β +δn+1 (tV ) are
chains of U ∩V . Therefore, α−δn+1 (tU ) = β+δn+1 (tV ) = x, for some x ∈ Cn (U ∩V ) such that
δn (x) = 0 (i.e., x ∈ Zn (U ∩ V )). Consequently, c = α + Bn (U ) = (α − δn+1 (tU )) + Bn (U ) =
x + Bn (U ) = Hn (k)(x + Bn (U ∩ V )).
Similarly, d = Hn (l)(x + Bn (U ∩ V )). Hence, we have, (c, d) = (Hn (k)(x + Bn (U ∩
V )), Hn (l)(x + Bn (U ∩ V ))) = g(x + Bn (U ∩ V )). i.e., (c, d) ∈ Im(g).
This proves Ker(f ) ⊆ Im(g), as desired.
A. Deb Ray, Pure Math, C.U. 18

We wish to show that the sequence Hn (U ) ⊕ Hn (V ) → Hn (X) → Hn−1 (U ∩ V ) is also

exact. For that we begin to search for a homomorphism Hn (X) → Hn−1 (U ∩ V ).

Let c ∈ Cn (X) be such that c+Bn (X) ∈ Hn (X). Then c is a cycle and so, δn (c) = 0. By chain
splitting, there exists cU ∈ Cn (U ) and cV ∈ Cn (V ) such that c − cU − cV = δn+1 (t) for some
t ∈ Cn+1 (X). Therefore, δn (c − cU − cV ) = δn (δn+1 (t)) = 0. i.e., 0 = δn (c) − δn (cU ) − δn (cV )
which gives
δn (cU ) = −δn (cV )
The left side of this expression shows that it is a chain in U and the right side of the expression
shows that it is a chain in V . Hence, both are chains in U ∩ V . Also, δn (cU ) ∈ Cn−1 (U ∩ V )
with δn−1 (δn (cU )) = 0 implies that δn (cU ) ∈ Zn−1 (U ∩ V ). Hence, for each c + Bn (X) ∈
Hn (X), we find an element δn (cU ) + Bn−1 (U ∩ V ) ∈ Hn−1 (U ∩ V ). It is not hard to show
that such an element depends on c and not on the choice of cU .

Lemma 6.1.2. If c is homologous to cU + cV and also to bU + bV , where cU , bU ∈ Cn (U ) and

cV , bV ∈ Cn (V ) then δn (cU ) + Bn−1 (U ∩ V ) = δn (bU ) + Bn−1 (U ∩ V ).

Proof. c is homologous to cU +cV implies that there exists p ∈ Cn+1 (X) such that c−cU −cV =
δn+1 (p). Similarly, c is homologous to bU +bV implies that there exists q ∈ Cn+1 (X) such that
c − bU − bV = δn+1 (q). Therefore, subtracting one from the other we get, cU + cV − bU − bV =
δn+1 (q − p). By chain splitting on a = q − p, there exist aU ∈ Cn+1 (U ) and aV ∈ Cn+1 (V )
such that a − aU − aV = δn+2 (s), for some s ∈ Cn+2 (X). Therefore,

δn+1 (a − aU − aV ) = δn+1 (δn+2 (s)) = 0

So, δn+1 (a) = δn+1 (aU )+δn+1 (aV ) which in turn gives cU +cV −bU −bV = δn+1 (aU )+δn+1 (aV ).
i.e.,
(cU − bU ) − δn+1 (aU ) = (bV − cV ) + δn+1 (aV )
The left side expression shows that it is a chain in U and the right side shows that it is
a chain in V . Hence both are chains in U ∩ V with boundary δn (cU − bU − δn+1 (aU ) =
δn (cU ) − δn (bU ). Consequently, δn (cU ) − δn (bU ) ∈ Bn−1 (U ∩ V ). Hence,

δn (cU ) + Bn−1 (U ∩ V ) = δn (bU ) + Bn−1 (U ∩ V )

The following lemma defines a map Hn (X) → Hn−1 (U ∩ V ):

Lemma 6.1.3. For each n ≥ 1, there exists a map η : Hn (X) → Hn−1 (U ∩ V ), described by

c + Bn (X) 7→ δn (cU ) + Bn−1 (U ∩ V )

where cU ∈ Cn (U ) is as discussed before.

Proof. To show that such assignment is well-defined. Let c + Bn (X) = d + Bn (X) ∈ Hn (X).
Then c − d ∈ Bn (X). So there is some t ∈ Cn+1 (X), such that c − d = δn+1 (t). By chain
splitting on c and d, there are cU , dU ∈ Cn (U ) and cV , dV ∈ Cn (V ) such that c − cU − cV =
A. Deb Ray, Pure Math, C.U. 19

δn+1 (y) and d − dU − dV = δn+1 (z) for some y, z ∈ Cn+1 (X). Then (c − d) − (cU − dU ) −
(cV − dV ) = δn+1 (y − z). Hence,
(cU − dU ) + (cV − dV ) = δn+1 (t − y − z)
Using chain splitting on (t − y − z), we get tU ∈ Cn+1 (U ) and tV ∈ Cn+1 (V ) such that
(t − y − z) ∼ tU + tV . In particular, δn+1 (t − y − z) = δn+1 (tU ) + δn+1 (tV ).

Therefore, cU − dU − δn+1 (tU ) = dV − cV + δn+1 (tV ).

The equality above shows that each is a chain on U ∩ V with boundary
δn (cU − dU − δn+1 (tU )) = δn (cU ) − δn (cV ).
Hence, δn (cU ) − δn (cV ) is the image under δn of a chain in U ∩ V . Hence, δn (cU ) − δn (cV ) ∈
Bn−1 (U ∩ V ) and consequently, η(c + Bn (X)) = η(d + Bn (X)).
Before proving the result, we observe the following :
Observation 6.1.1. If c + Bn (X) ∈ Hn (X), then δn (c) = 0 and applying chain splitting on
c ∈ Cn (X), there exists cU ∈ Cn (U ) and cV ∈ Cn (V ) such that c − cU − cV = δn+1 (t) for some
t ∈ Cn+1 (X). So, δn (c−cU −cV ) = δn (δn+1 (t)) which in turn gives δn (c)−δn (cU )−δn (cV ) = 0.
Hence,
δn (cU ) = −δn (cV )
The left side expression is a chain on U and the right side expression is a chain on V .
Therefore, both are chains in U ∩ V . But we can not claim that both are in Bn−1 (U ∩ V ).
f η
Theorem 6.1.2. For each n ≥ 0, Hn (U ) ⊕ Hn (V ) −→ Hn (X) −→ Hn−1 (U ∩ V ) is an exact
sequence.
Proof. Choose any x ∈ Im(f ). Then there exists (c, d) ∈ Hn (U )⊕Hn (V ) such that f (c, d) =
x. Now,
η(x) = η(f (c, d)) = η(Hn (i)(c) − Hn (j)(d)).
Now Hn (i)(c) = Hn (i)(c + Bn (U )), where c ∈ Zn (U ) ⊆ Cn (U ). Therefore, Hn (i)(c) =
Cn (i)(c) + Bn (X) = i ◦ c + Bn (X) = c + Bn (X). Similarly, Hn (j)(d) = d + Bn (X). So, the
element (Hn (i)(c)−Hn (j)(d)) has a split (c+d)+Bn (X) and hence, applying the definition of
η, η(x) = η(Hn (i)(c) − Hn (j)(d)) = δn (c) + Bn−1 (U ∩ V ) = Bn−1 (U ∩ V ). (Since c ∈ Zn (U )).
i.e., Im(f ) ⊆ Ker(η).
For the reverse inclusion, we start with c + Bn (X) ∈ Ker(η). Using chain splitting on c,
cU + cV is homologous to c. Then δn (cU ) ∈ Bn−1 (U ∩ V ). Proceeding as in the previous
Observation, δn (cU ) = −δn (cV ) = δn (d) for some d ∈ Cn (U ∩ V ). Hence, cU − d ∈ Ker(δn )
and cV + d ∈ Ker(δn ).
Now, cU − d ∈ Cn (U ), cV + d ∈ Cn (V ) and cU + cV = (cU − d) + (cV + d) implies that c is
homologous to (cU − d) + (cV + d). Hence,
f ((cU − d) + Bn (U ), −(cV + d) + Bn (V ))
= Hn (i)(cU − d) − Hn (j)(−(cV + d))
= (Cn (i)(cU − d) + Cn (j)(cV + d)) + Bn (X)
= c + Bn (X).
A. Deb Ray, Pure Math, C.U. 20

Consequently, c + Bn (X) ∈ Im(f ). i.e., Ker(η) ⊆ Im(f ).

Finally, we prove,
η g
Theorem 6.1.3. For each n > 1, Hn (X) −→ Hn−1 (U ∩ V ) −→ Hn−1 (U ) ⊕ Hn−1 (V ) is an
exact sequence.
Proof. For any x + Bn (X) ∈ Hn (X),
(g ◦ η)(x + Bn (X))
= g(η(x + Bn (X)))
= g(δn (xU ) + Bn−1 (U ∩ V ))
= (Hn−1 (k)(δn (xU ) + Bn−1 (U ∩ V )), Hn−1 (l)(δn (xU ) + Bn−1 (U ∩ V )))
= (δn (xU ) + Bn−1 (U ), δn (xU ) + Bn−1 (V ))
= 0
Hence, Im(η) ⊆ Ker(g).
Let x = x + Bn−1 (U ∩ V ) ∈ Ker(g). Then
g(x) = (Bn−1 (U ), Bn−1 (V ))
⇒ (Hn−1 (k)(x), Hn−1 (l)(x) = (Bn−1 (U ), Bn−1 (V ))
⇒ (x + Bn−1 (U ), x + Bn−1 (V )) = (Bn−1 (U ), Bn−1 (V ))
⇒ x ∈ Bn−1 (U ) ∩ Bn−1 (V )
So, there are bU ∈ Cn (U ) and bV ∈ Cn (V ) such that x = δn (bU ) = δn (bV ). Then δn (bU −
bV ) = 0 and so, bU − bV ∈ Zn (X). i.e., η((bU − bV ) + Bn (X)) = δn (bU ) + Bn−1 (U ∩ V ) =
x + Bn−1 (U ∩ V ). Therefore, x + Bn−1 (U ∩ V ) ∈ Im(η), proving Ker(g) ⊆ Im(η).
Theorem 6.1.4. Given a topological space X and open subsets U and V of X such that
X = U ∪ V , there exists a long exact sequence
. . . → Hn (U ∩ V ) → Hn (U ) ⊕ Hn (V ) → Hn (X) → Hn−1 (U ∩ V ) → . . .
. . . → H1 (X) → H0 (U ∩ V ) → H0 (U ) ⊕ H0 (V ) → H0 (X) → 0
called Mayer-Vietoris (M-V) sequence.
Proof. Follows from Theorem 6.1.1, 6.1.2 and 6.1.3.

6.2 Computation of Homology groups

In this section, we shall learn how M-V sequence can be utilized for computing homology
groups of several well-known spaces. We recall the following example that we have established
in the previous chapter.
Example 6.2.1. Suppose X is a singleton, i.e., X = {x}. Then
(
Z, if n = 0
Hn (X) =
0, otherwise
A. Deb Ray, Pure Math, C.U. 21

Let X be any contractible space. Then X is of the same homotopy type with a point space
{a}. So, by a result proved in the previous chapter, Hn (X) ∼
= Hn ({a}). Using the example
above, we then have
Example 6.2.2. For any contractible space X,
(
Z, if n = 0
Hn (X) =
0, otherwise

We now calculate the homology groups of S 1 . (Note that S 1 is not contractible)

Example 6.2.3. Suppose X = S 1 . Then
(
Z, if n = 0, 1
Hn (S 1 ) =
0, otherwise.

If p = (0, 1) and q = (0, −1), then U = S 1 \ {p} and V = S 1 \ {q} are open subsets of X such
that X = U ∪ V . Certainly, U and V both are homeomorphic to R and R is contractible.
So, Hn (U ) = 0 = Hn (V ) for each n ̸= 0. Again U ∩ V is`the disjoint union of two spaces P
and Q, each of which is homeomorphic to R. So, H0 (P Q) = H0 (P ) ⊕ H0 (Q) = Z ⊕ Z.
Applying M-V sequence,

. . . → Hn (X) → 0 → . . . → 0 → H1 (X) → Z ⊕ Z → Z ⊕ Z → H0 (X) → 0

Since S 1 has one path component, H0 (X) = H0 (S 1 ) = Z. Since Hn (X) is sandwiched

between trivial groups, Hn (X) = 0, for all n > 1. So the only nontrivial portion of the M-V
sequence is
f1 f2 f3
0 → H1 (X) → Z ⊕ Z → Z ⊕ Z → Z → 0
Since f3 is surjective, Z⊕Z/Ker(f3 ) ∼
= Z and so, Ker(f3 ) ∼
= Z. By exactness of the sequence,
Im(f2 ) = Z.
f1 f2 f3
0 → H1 (X) → Z ⊕ Z → Z ⊕ Z → Z → 0
By a similar argument, Z ⊕ Z/Ker(f2 ) ∼ = Im(f2 ) ∼
= Z and so, Ker(f2 ) = Z. Therefore,
Im(f1 ) = Z. Ker(f1 ) being trivial, H1 (X) ∼
= H1 (X)/{0} ∼= Z. So,
(
Z, if n = 0, 1
Hn (S 1 ) =
0, otherwise.

Example 6.2.4. Suppose X = S 2 . Then

(
Z, if n = 0, 2
Hn (S 2 ) =
0, otherwise.

If p = (0, 0, 1) and q = (0, 0, −1), then U = S 2 \ {p} and V = S 2 \ {q} are open subsets of
X such that X = U ∪ V . U and V both are homeomorphic to R2 and R2 is contractible.
So, Hn (U ) = 0 = Hn (V ) for each n ̸= 0. Again U ∩ V is homeomorphic to R × S 1 , which is
A. Deb Ray, Pure Math, C.U. 22

homotopy equivalent to S 1 . Hence, Hn (S 1 ) = 0 for all n ̸= 0, 1 and Z otherwise. S 2 being

path connected, H0 (S 2 ) = Z. So, M-V sequence takes the form

. . . → 0 → Hn (X) → 0 → . . . → 0 → H2 (X) → Z → 0 ⊕ 0
→ H1 (X) → Z → Z ⊕ Z → Z → 0

Proceeding as in the previous example, H1 (X) = 0 and H2 (X) ∼

= Z. Consequently,
(
Z, if n = 0, 2
Hn (S 2 ) =
0, otherwise.

We may generalize this idea and prove inductively, for any m ≥ 1,

(
Z, if n = 0, m
Hn (S m ) =
0, otherwise.

Example 6.2.5. Suppose X is a figure-8 space. i.e., X = S 1 ∨ S 1 . Then


Z ⊕ Z,
 if n = 1
Hn (X) = Z, if n = 0

0, otherwise.


Consider an open set U obtained by deleting any point from the left circle (not lying on the
right circle) and an open set V obtained by deleting any point from the right circle (not
lying on the left circle).
Clearly, X = U ∪ V , with U and V both homotopy equivalent to S 1 and U ∩ V homotopy
equivalent to a point space, say, {p}. So, Hn (V ) = Hn (U ) = Hn (S 1 ) = Z, for n = 0, 1 and
trivial for any n > 1. Also, X being path connected, H0 (X) = Z. By M-V sequence,

. . . → Hn (S 1 ) ⊕ Hn (S 1 ) → Hn (X) → Hn−1 (p) → . . .

→ H1 (p) → H1 (S 1 ) ⊕ H1 (S 1 ) → H1 (X)
→ H0 (p) → H0 (S 1 ) ⊕ H0 (S 1 ) → H0 (X) → 0

Clearly, Hn (X) = 0, for all n > 1.

The nontrivial portion of the M-V sequence is
f1 f2 f3 f4 f5
0 → Z ⊕ Z → H1 (X) → Z → Z ⊕ Z → Z → 0.

From the exactness of the sequence, we get

ˆ f4 is surjective.
So, (Z ⊕ Z)/Ker(f4 ) ∼
= Im(f4 ) = Z and so, Ker(f4 ) = Z.
ˆ Im(f3 ) = Ker(f4 ) = Z.

ˆ Z/Ker(f3 ) ∼
= Im(f3 ) = Z and so, Ker(f3 ) = 0.
A. Deb Ray, Pure Math, C.U. 23

f1 f2 f3 f4 f5
0 → Z ⊕ Z → H1 (X) → Z → Z ⊕ Z → Z → 0.

ˆ Im(f2 ) = Ker(f3 ) = 0.

ˆ f2 = 0.

ˆ Ker(f2 ) = H1 (X) and hence, Im(f1 ) = Ker(f2 ) = H1 (X).

ˆ But, f1 is injective. So, Im(f1 ) = Z ⊕ Z.

Hence, H1 (X) = Z ⊕ Z.

One may also note that in some cases, M-V Theorem alone may not give the precise
homology groups of a space. For example, if we consider a torus and apply M-V sequence
then we shall get a list of probable candidates for its homology groups, thereby get an
estimate, not exact ones. Using other known results and facts we may of course determine
its homology groups precisely.

6.3 Applications
Among various applications of singular homology groups, we cite only a couple of very
important applications of these groups to resolve purely topological problems. We observe
how elegantly such non-trivial fundamental questions can be answered via homology groups.

Theorem 6.3.1. (Invariance of Dimension) For m ̸= n, S m and S n are not homeomor-

phic.

Proof. If possible let there be a homeomorphism f : S m → S n (m ̸= n). Then f −1 : S n → S n

is a continuous function such that f ◦ f −1 = IS n and f −1 ◦ f = IS m . Without loss of
generality we assume that n ̸= 0 (For otherwise, we have m ̸= 0). By functorial property,
Hn (f ) ◦ Hn (f −1 ) = IHn (S n ) . Since m ̸= n and n ̸= 0, Hn (S m ) = 0. So, we get

Hn (f −1 ) Hn (f )
Z −→ 0 −→ Z

Hence, the identity homomorphism is factoring through 0, which is impossible. Therefore,

S m is not homeomorphic to S n , when m ̸= n.

It has been observed before that a special case of Brouwer’s Fixed Point Theorem can
be achieved using Fundamental groups. In what follows next, we see that for any n > 0 this
theorem can be completely resolved via homology groups.

Theorem 6.3.2. (Brouwer’s Fixed Point Theorem) Every continuous map f : Dn →

Dn has a fixed point.

Proof. Suppose, f : Dn → Dn has no fixed point. i.e., f (z) ̸= z, for all z ∈ Dn . Consider the
ray (1−t)f (z)+tz (t ≥ 0) that starts from f (z) and passes through z. This ray will meet the
boundary S n−1 at some point, say g(z). Then we get a continuous function g : Dn → S n−1
A. Deb Ray, Pure Math, C.U. 24

given by z 7→ g(z) such that g(z) = z, for all z ∈ S n−1 .

Consider the following diagram, where i denotes the inclusion map.
i g
S n−1 → Dn → S n−1

Case 1 : n > 1.
This sequence of continuous functions induces another sequence of homomorphisms between
the respective (n-1)-th homology groups. As Dn is convex, Hm (Dn ) = 0 for all m. So, we
have
Hn (i) Hn−1 (g)
Z∼= Hn−1 (S n−1 ) −→ Hn−1 (Dn ) ∼
= 0 −→ Hn−1 (S n−1 ) ∼=Z
By functorial properties of fundamental groups,
Hn−1 (g) ◦ Hn−1 (i) = Hn−1 (g ◦ i) = Hn−1 (IS n−1 ) = IHn−1 (S n−1 ) showing that identity map on
Z is factoring through the trivial group 0, which is impossible. Hence, such a g can not exist
and so, there is some z ∈ Dn such that f (z) = z.

Case 2 : n = 1.
In this case, D1 = [−1, 1] and S 0 = {−1, 1}. As D1 is connected, image of D1 under the
continuous function g is either {1} or {−1}, but not the whole of S 0 . So, g(z) = z for all
z ∈ S 0 can not be satisfied - a contradiction to the construction of g.
Before we conclude, we point out that there is a theorem due to Eilenberg and Steenrod
that says, for any triangulable space, the singular homology groups are same as the simplicial
homology groups.
Therefore, one may raise a question: If Singular Homology is more general then why
we study simplicial homology at all?

This is indeed a pertinent question. But for computing homology groups of a triangulable
space one may use its simplicial structure, avoiding the structural complications of singular
homology, whereas the efficiency of singular homology theory lies in the results involving
induced homomorphisms, M-V sequence, etc. This justifies the study of both.
Chapter 7

Relation between H1(X) and π(X)

In this course of Algebraic Topology, we have studied fundamental groups of pointed topo-
logical spaces and have seen the construction of simplicial and singular homology groups of
topological spaces. A question certainly arises, whether any relation exists between funda-
mental group of a space and any of its homology group. We can not expect them to be
equal, because, homology groups are abelian but fundamental groups may not.

In this chapter, we shall learn that for path connected spaces, there always exists a ho-
momorphism from fundamental group onto its first homology group. This homomorphism
is known as Hurewicz homomorphism. Moreover, it can also be established that the
abelianization of the fundamental group of a path connected space is algebraically same
with the first homology group of the space.

We now discuss the technique of abelianization of any group.

7.1 Abelianization
Definition 7.1.1. Let G be a group and a, b ∈ G. The commutator of a and b is the element
aba−1 b−1 .
Let A = {aba−1 b−1 : a, b ∈ G} and let [G, G] be the subgroup of G generated by A.
[G, G] is called the derived or commutator subgroup of G.

ˆ G is commutative if and only if [G, G] = {e}.

If G is commutative, then A = {e} and so, [G, G] = {e}. Conversely, if [G, G] = {e},
then aba−1 b−1 = e for all a, b ∈ G. Therefore, ab = ba for all a, b ∈ G, i.e., G is
commutative.

Theorem 7.1.1. The derived subgroup [G, G] of a group G is a normal subgroup of G and
G/[G, G] is abelian.
Proof. By definition, [G, G] is a subgroup of G generated by the commutators aba−1 b−1 , for
all a, b ∈ G. It is enough to show that ∀a, b, g ∈ G, g(aba−1 b−1 )g −1 ∈ [G, G]. Now,

g(aba−1 b−1 )g −1 = (gag −1 )(gbg −1 )(ga−1 g −1 )(gb−1 g −1 ) = cdc−1 d−1

25
A. Deb Ray, Pure Math, C.U. 26

where c = gag −1 and d = gbg −1 .

So, g(aba−1 b−1 )g −1 is a commutator and hence belongs to [G, G].
To show that G/[G, G] is commutative. Let us denote [G, G] by H. i.e., To show that
(aH)(bH) = (bH)(aH), for any a, b ∈ G.
For a, b ∈ G, (ba)−1 (ab) = a−1 b−1 ab = a−1 b−1 (a−1 )−1 (b−1 )−1 ∈ H ⇒ (ab)H = (ba)H ⇒
(aH)(bH) = (bH)(aH), as desired.
More generally, one may show the following:
Theorem 7.1.2. Let [G, G] be the derived subgroup of a group G and H a subgroup of G.
Then H ⊇ [G, G] if and only if H is a normal subgroup of G and G/H is abelian.
Definition 7.1.2. If G is a group and [G, G] is the commutator subgroup of G then G/[G, G]
is called the abelianization of G. We denote abelianization of G by Gab , for simplicity of
notation.
It is clear that the abelianization of an abelian group G is G/{e} and hence isomorphic to
G itself.
Theorem 7.1.3. Let G and G1 be two groups such that G1 is abelian. A homomorphism
ϕ : G → G1 induces a homomorphism ψ : G/[G, G] → G1 , given by
ψ(x[G, G]) = ϕ(x)
Proof. To check that such a map ψ is well-defined.
x[G, G] = y[G, G]
⇒ x−1 y ∈ [G, G]
⇒ x−1 y = (a1 b1 a−1 −1 −1 −1
1 b1 ) . . . (an bn an bn )
⇒ ϕ(x−1 y) = ϕ(a1 )ϕ(b1 )ϕ(a1 )−1 ϕ(b1 )−1 . . . ϕ(an )ϕ(bn )ϕ(an )−1 ϕ(bn )−1
⇒ ϕ(x)−1 ϕ(y) ∈ [ϕ(G), ϕ(G)] = {e}
⇒ ϕ(x) = ϕ(y)
⇒ ψ(x[G, G]) = ψ(y[G, G]).
That ψ is a homomorphism follows from that ϕ is a homomorphism.

7.2 H1(X) is isomorphic to the abelianization of π(X)

We establish here the relationship between π(X) and H1 (X) for any path connected space.
We must emphasize the word ‘path connected’ before we proceed any further. Because,
unless the space is path connected, even if the fundamental group happens to be abelian, we
can not expect its sameness with the first homology group. (In fact, fundamental group in
such case very much depend on the base point and hence not unique). Therefore, by a space
in this section we always mean a path connected space.

Suppose X is a path connected space and x0 ∈ X. Here is a scheme that we wish to

follow:
A. Deb Ray, Pure Math, C.U. 27

ˆ To construct a surjective homomomorphism π(X, x0 ) → H1 (X).

ˆ This homomorphism will induce another surjective homomorphism π(X, x0 )ab → H1 (X).
To show that such homomorphism is actually an isomorphism, known as Hurewicz
isomorphism.

This will establish :

Theorem 7.2.1. Hurewicz Theorem for n=1 The abelianization of the fundamental
group of a path connected space is isomorphic to its 1st Homology group.

Suppose X is path connected. Choose any x0 ∈ X. We wish to construct a homomorphism

ϕ : π(X, x0 ) → H1 (X).

Observation 7.2.1. ˆ The closed interval I = [0, 1] can be considered as a subspace of

2
R , by treating the closed line segment joining (1, 0) and (0, 1), as a copy of I.

ˆ For each path α : I → X, we define a function α : ∆1 → X by α(s, t) = α(s). Note

that for each (s, t) ∈ ∆1 , we have s + t = 1, s, t ≥ 0. Therefore, t = 1 − s. Clearly, α
being continuous, α is continuous. Hence, it is a singular 1-simplex. In other words,
for each path α, we always get α ∈ S1 (X).
Conversely, suppose a singular one simplex β : ∆1 → X is given. Then defining
β : [0, 1] → X by β(s) = β(s, 1 − s) we get a path in X. Hence, there is a one-one
correspondence between the collection of paths in X and the members of S1 (X).

ˆ If α : I → X is a loop; i.e., α(0) = α(1) = x0 , then the 1-simplex α is a 1-cycle.

Conversely, the corresponding path constructed from a 1-cycle α must be a loop.,
To prove this we consider the following:

1δ
. . . −→ C1 (X) −→ C0 (X) −→ 0

δ1 (α)(1) = (α ◦ d0 )(1) − (α ◦ d1 )(1) = α(1, 0) − α(0, 1) = 0, because, α(1, 0) = α(0, 1) =

x0 . Hence, α ∈ Z1 (X) = Ker(δ1 ).

Theorem 7.2.2. If α, β : [0, 1] → X are two paths such that α(1) = β(0) then α+β−α ∗ β ∈
B1 (X).

Proof. Consider an isosceles triangle ∆ with vertices a0 , a1 , a2 and the midpoint of the base
[a0 , a1 ] (wherever we write [a, b], we mean the line segment joining a with b) of the triangle is
x0 . We join the vertex a2 with x0 . Suppose the line segment [a0 , a2 ] denotes the image of α
in X and the line segment [a2 , a1 ] denotes the image of β in X. Define a function F : ∆ → X
as follows:
(i) Choose any point (s, t) ∈ ∆ and draw a line l parallel to [a2 , x0 ] through (s, t).
(ii) If l intersects [a0 , a2 ], then assign F (s, t) = the value of α at that point of intersection.
(iii) If l intersects [a2 , a1 ], then assign F (s, t) = the value of β at that point of intersection.
Clearly, F is a continuous function, because α and β agree at a2 . Certainly, α = F ◦ <
a0 , a2 > and β = F ◦ < a2 , a1 >.
A. Deb Ray, Pure Math, C.U. 28

Consider the 2-simplex < a0 , a1 , a2 >: ∆2 → X. Then we claim that the 2-simplex F ◦ <
a0 , a1 , a2 >: ∆2 → X is such that δ2 (F ◦ < a0 , a1 , a2 >) = α + β − α ∗ β.

δ2 (F ◦ < a0 , a1 , a2 >)(s, t)
= (F ◦ < a0 , a1 , a2 > ◦d0 − F ◦ < a0 , a1 , a2 > ◦d1 + F ◦ < a0 , a1 , a2 > ◦d2 )(s, t)
= F ◦ < a1 , a2 > (s, t) − F ◦ < a0 , a2 > (s, t) + F ◦ < a0 , a1 > (s, t)
= −β(s, t) − α(s, t) + α ∗ β(s, t)

Hence, −β − α + α ∗ β ∈ B1 (X).
From Theorem 7.2.2, we get the following:

Theorem 7.2.3. 1. If Cx0 is a loop based at x0 then Cx0 ∈ B1 (X).

2. If α and β are two paths with α ∼ β then α − β ∈ B1 (X).

3. If α is a path then α + α−1 ∈ B1 (X).

Proof. 1. Consider an isosceles triangle ∆ with vertices a0 , a1 , a2 . Define F : ∆ → X

by F (s, t) = x0 . Then F ◦ < a0 , a1 , a2 >: ∆2 → X is a 2-simplex such that δ2 (F ◦ <
a0 , a1 , a2 >) = Cx0 . Hence, Cx0 ∈ B1 (X).

2. Suppose H : I × I → X is a path homotopy between α and β. We have to find some

τ ∈ C2 (X), such that δ2 (τ ) = α − β.
Suppose, a0 , a1 , a2 and a3 represent the vertices of the rectangle I ×I and < a0 , a1 , a2 >:
∆2 → I × I and < a0 , a2 , a3 >: ∆2 → I × I are continuous maps. So, the following
diagrams show that H◦ < a0 , a1 , a2 > and H◦ < a0 , a2 , a3 > are singular 2-simplexes
of X.
<a0 ,a1 ,a2 > <a0 ,a2 ,a3 >
∆2 /I ×I ∆2 /I ×I

H H
H◦<a0 ,a1 ,a2 >
#  H◦<a0 ,a2 ,a3 >
# 
X X
Therefore, we consider τ = H◦ < a0 , a1 , a2 > +H◦ < a0 , a2 , a3 >∈ C2 (X). We claim
that δ2 (τ ) = α − β. Viewing H◦ < a0 , a3 > and H◦ < a1 , a2 > as singular 1-simplexes,
we get (H◦ < a0 , a3 >)(s, t) = x0 = (H◦ < a1 , a2 >)(s, t).

δ2 (τ )(s, t) = τ ◦ d0 (s, t) − τ ◦ d1 (s, t) + τ ◦ d2 (s, t)

= τ (0, s, t) − τ (s, 0, t) + τ (s, t, 0)
= (H◦ < a0 , a1 , a2 >)(0, s, t) + (H◦ < a0 , a2 , a3 >)(0, s, t)
−(H◦ < a0 , a1 , a2 >)(s, 0, t) − (H◦ < a0 , a2 , a3 >)(s, 0, t)
+(H◦ < a0 , a1 , a2 >)(s, t, 0) + (H◦ < a0 , a2 , a3 >)(s, t, 0)
= (H ◦ (< a1 , a2 > − < a0 , a2 > + < a0 , a1 >))(s, t)
+(H ◦ (< a2 , a3 > − < a0 , a3 > + < a0 , a2 >))(s, t)
= (H◦ < a0 , a1 > +H◦ < a2 , a3 >)(s, t)
= α(s, t) − β(s, t).
A. Deb Ray, Pure Math, C.U. 29

3. If α is a path then by Theorem 7.2.2, α + α−1 − α ∗ α−1 ∈ B1 (X). Also, α ∗ α−1 ∼ Cx0 ,
α ∗ α−1 − Cx0 ∈ B1 (X). Since B1 (X) is a group, α + α−1 − Cx0 ∈ B1 (X). Therefore,
α + α−1 ∈ B1 (X).

From the above theorems we get (check) the following:

ˆ If [Cx0 ] is the identity element in π(X, x0 ), then Cx0 + B1 (X) = B1 (X) (= identity
element of H1 (X)).

ˆ For any loop α, α−1 + B1 (X) = −α + B1 (X).

In view of these observations, we define a map ϕ : π(X, x0 ) → H1 (X) as follows:

ϕ([α]) = α + B1 (X).
ϕ is well defined : Follows from the fact that α ∼ β implies α − β ∈ B1 (X).
ϕ is a homomorphism : If α and β are two loops based at x0 , then by Theorem 7.2.2,
(α ∗ β) + B1 (X) = (α + B1 (X)) + (β + B1 (X)). Hence, ϕ([α] ◦ [β]) = ϕ([α]) + ϕ([β]). i.e., ϕ
is a homomorphism.

So, we can define a homomorphism ψ : π(X)ab → H1 (X) with the help of ϕ as follows:
Denote the commutator subgroup of π(X, x0 ) by G. Then any element of π(X)ab is denoted
by [α]/G, where α is a loop based at x0 . So, α is a member of H1 (X). Hence,

ψ([α]/G) = α.

We define a homomorphism η : H1 (X) → π(X)ab and show that such η is the inverse of ψ,
proving ψ to be an isomorphism.

ˆ Choose any path ωx , starting from x0 and ending at x, whenever x ̸= x0 .

ˆ ωx0 is the constant loop at x0 .

ˆ For any singular 1-simplex σ, if σ(0, 1) = a and σ(1, 0) = b, then ωa ∗ σ ∗ ωb−1 is a loop
based at x0 .

ˆ If σ is a 1-cycle, then σ(1, 0) = σ(0, 1) = a, (say). In such case, ωa ∗ σ ∗ ωa−1 is a loop

based at x0 .

Hence, define λ : S1 (X) → π(X)ab by sending σ to the element of π(X)ab represented by the
loop ωa ∗ σ ∗ ωb−1 and extend it to C1 (X) by linearity.
i.e., Define λ : S1 (X) → π(X)ab by λ(σ) = [ωa ∗ σ ∗ ωb−1 ]/G and extend it by linearity to
C1 (X).
Of course, it is not hard to see that λ is a homomorphism. If δ2 : C2 (X) → C1 (X) and
λ(δ2 (S2 (X))) = 0 then we can define a homomorphism η : H1 (X) → π(X)ab in a very
natural way. So, we proceed to show that λ(δ2 (S2 (X))) = 0.
Let ϵ : ∆2 → X be a singular 2-simplex. We shall show that λ(δ2 (ϵ)) = 0. Let ϵ(1, 0, 0) =
a, ϵ(0, 1, 0) = b, ϵ(0, 0, 1) = c. One can easily check that (ϵ ◦ d0 )(s, t) =< b, c > (s, t),
A. Deb Ray, Pure Math, C.U. 30

(ϵ◦d1 )(s, t) =< a, c > (s, t) (and therefore, (−ϵ◦d1 )(s, t) =< c, a > (s, t)) and (ϵ◦d2 )(s, t) =<
a, b > (s, t)
If di : ∆1 → ∆2 are the maps that insert 0 in i-th place (i = 0, 1, 2), we get
λ(δ2 (ϵ)) = λ(ϵ ◦ d0 − ϵ ◦ d1 + ϵ ◦ d2 )
= λ(ϵ ◦ d0 ) ◦ λ(−ϵ ◦ d1 ) ◦ λ(ϵ ◦ d2 )
= [ωb ∗ (ϵ ◦ d0 ) ∗ ωc−1 ]/G ◦ [ωc ∗ (−ϵ ◦ d1 ) ∗ ωa−1 ]/G ◦ [ωa ∗ (ϵ ◦ d2 ) ∗ ωb−1 ]/G
= [(ωb ∗ (ϵ ◦ d0 ) ∗ ωc−1 ) ∗ (ωc ∗ (−ϵ ◦ d1 ) ∗ ωa−1 ) ∗ (ωa ∗ (ϵ ◦ d2 ) ∗ ωb−1 )]/G
= [(ωb ∗ (ϵ ◦ d0 ∗ −ϵ ◦ d1 ∗ ϵ ◦ d2 ) ∗ ωb−1 )]/G
= [Cx0 ]/G
Because, ϵ ◦ d0 ∗ −ϵ ◦ d1 ∗ ϵ ◦ d2 is a loop based at b lying in the path connected space ϵ(∆1 ), it
is homotopic to the constant loop at b. Hence, ωb ∗ (ϵ ◦ d0 ∗ −ϵ ◦ d1 ∗ ϵ ◦ d2 ) ∗ ωb−1 is homotopic
to a constant loop at x0 . So, Im(δ2 ) ⊆ Ker(λ).
Define η : H1 (X) → π(X)ab by
η(σ + B1 (X)) = λ(σ).
Since σ is a 1-cycle, σ(1, 0) = σ(0, 1) = a (say), and therefore, λ(σ) = [ωa ∗ σ ∗ ωa−1 ]/G.
We now show that the map η is well defined.
Let σ1 = σ2 in H1 (X). Then
σ1 − σ2 ∈ B1 (X) = Im(δ2 ) ⊆ Ker(λ)
⇒ λ(σ1 − σ2 ) = 0
⇒ [ωa ∗ (σ1 − σ2 ) ∗ ωa−1 ] ∈ G
⇒ [ωa ∗ σ1 ∗ ωa−1 ]/G = [ωa ∗ σ2 ∗ ωa−1 ]/G
⇒ λ(σ1 ) = λ(σ2 )
It can now be checked that such η is the inverse of ψ:
(η ◦ ψ)([α]/G) = η(ϕ([α])) = η(α + B1 (X)) = λ(α) = [ωx0 ∗ α ∗ ωx−10
]/G = [α]/G, since α is a
−1
loop based at x0 , it is homotopic to ωx0 ∗ α ∗ ωx0 .
(ψ ◦η)(α+B1 (X)) = ψ(λ(α)) = ψ([ωa ∗α∗ωa−1 ]/G), where α is the loop corresponding to α ∈
Z1 (X), based at a ∈ X. Then (ψ ◦ η)(α + B1 (X)) = ϕ([ωa ∗ α ∗ ωa−1 ]) = ωa ∗ α ∗ ωa−1 + B1 (X).
We now show that ωa ∗ α ∗ ωa−1 − α ∈ B1 (X) :
Applying Theorem 7.2.2, ωa ∗ α − α − ωa ∈ B1 (X) and ωa ∗ α ∗ ωa−1 − ωa ∗ α − ωa−1 ∈ B1 (X).
Since B1 (X) is a group, ωa ∗ α ∗ ωa−1 − ωa−1 − α − ωa ∈ B1 (X). Also. ωa−1 + ωa ∈ B1 (X) and
therefore, ωa ∗ α ∗ ωa−1 − α − ωa ∈ B1 (X). Hence, we arrive our desired result.

This shows that ψ : π(X)a b → H1 (X) is an isomorphism. Hence, for a path connected
topological space, 1st Homology group and the abelianization of its fundamental group are
same.

Hurewicz Theorem actually provides a connection between generalization of fundamental

groups, called homotopy groups, and the homology groups. In our course, we had no scope
for discussing the higher homotopy groups and so, could not state the general version of
Hurewicz Theorem here. However, Hurewicz Theorem is extremely useful and has immense
application in higher dimension.
A. Deb Ray, Pure Math, C.U. 31

7.3 Singular Homology groups of a Torus (Hn(T 2), for

all n)
In the last chapter, we have mentioned that we may not be able to find all the homology
groups only applying Mayer Vietoris sequence. We compute the homology groups of torus
now to see how Hurewicz theorem helps in this case:
Let X = T 2 = a torus. We delete all points on any verticle circle C1 drawn on a torus and
get the open set U = X \C1 . We delete all points on another verticle circle C2 , other than C1
and get the open set V = X \ C2 . Then X = U ∪ V , U ∩ V is disjoint union of two cylinders.
Clearly, U ∼ S 1 , V ∼ S 1 . So, Hn (U ) = Hn (S 1 ), Hn (V ) = Hn (S 1 ) and Hn (U ∩ V ) =
Hn (S 1 ) ⊕ Hn (S 1 ), for all n. Therefore, H1 (U ) = H1 (V ) = Z, H0 (U ) = H0 (V ) = Z and
H1 (U ∩ V ) = H0 (U ∩ V ) = Z, and all other homology groups are trivial. By Mayer Vietoris
sequence we now get:
Hn (X) = 0, for all n > 2. The nontrivial part of the sequence is as follows:
f1 f2 f3 f4 f5 f6
0 → H2 (X) → Z ⊕ Z → Z ⊕ Z → H1 (X) → Z ⊕ Z → Z ⊕ Z → H0 (X) → 0

Since X is path connected, H0 (X) = Z. f6 being surjective, Im(f6 ) = Z. By 1st isomorphism

theorem, Z ⊕ Z/Ker(f6 ) ∼= Im(f6 ) = Z. Then Ker(f6 ) = Z = Im(f5 ) (by exactness). By
1st isomorphism theorem again, Z ⊕ Z/Ker(f5 ) ∼ = Im(f5 ) = Z. Hence, Ker(f5 ) = Z. By
exactess at that point, Ker(f5 ) = Z = Im(f4 ). Therefore, H1 (X)/Ker(f4 ) ∼= Im(f4 ) = Z.
Now, By Hurewicz theorem, we know that H1 (X) = π(X) (Since for torus, π(X) is abelian)
= Z ⊕ Z. So, Ker(f4 ) = Z. Proceeding similarly, we get:

ˆ Im(f3 ) = Z

ˆ Ker(f3 ) = Z

ˆ Im(f2 ) = Z

ˆ Ker(f2 ) = Z

ˆ Im(f1 ) = Z

By 1st isomorphism theorem, H2 (X)/Ker(f1 ) ∼ = Im(f1 ) = Z. But f1 is injective and hence,

∼
Ker(f1 ) = 0. i.e., H2 (X) = Z. So, the homology groups of torus are as follows:

Hn (T 2 ) = Z ⊕ Z, if n = 1
= Z, if n = 0, 2
= 0, otherwise