INTRODUCTION

Hypergraphs are systems of finite sets and form, probably, the most gen-

eral concept in discrete mathematics. This branch of mathematics has

developed very rapidly during the latter part of the twentieth century,

influenced by the advent of computer science. Originally, developed in

France by Claude Berge in 1960, it is a generalization of graph theory in

which an edge can join any number of vertices. The basic idea consists in

considering sets as generalized edges and then in calling hypergraph the

family of these edges (hyperedges).

Hypergraphs model more general types of relations than graphs do. In

the past decades, the theory of hypergraphs has proved to be of a major

interest in applications to real-world problems. These mathematical tools

can be used to model networks, biology networks, data structures, process

scheduling, computations and a variety of other systems where complex

relationships between the objects in the system play a dominant role.

My work includes four chapters. First chapter contains basic defini-

tions in hypergraphs like in graph theory. Second chapter provides the

first properties such as the Helly property, the König property, and so on.

In third chapter, the classical notions of colorings are addressed. Fourth

chapter deals with some applications of hypergraph theory.

1
 PRELIMINARIES

Definitions

Graph

A graph is an ordered triple G = (V (G), E(G), IG ) where V (G) is a nonempty

set, E(G) is a set disjoint from V (G) and IG is an "incidence" relation that

associates with each element of E(G) an unordered pair of elements (same

or distinct) of V (G). Elements of V (G) are called the vertices(or nodes

or points) of G; and elements of E(G) are called the edges (or lines) of

G.V (G) and E(G) are the vertex set and edge set of G, respectively. If, for

the edge e of G, IG (e) = {u, v} we write IG (e) = uv.

Example:

2
 Here V (G) = {v1 , v2 , v3 , v4 , v5 } E(G) = {e1 , e2 , e3 , e4 , e5 , e6 } and IG is

given by

IG (e1 ) = {v1 , v5 } , IG (e2 ) = {v2 , v3 },

IG (e3 ) = {v2 , v4 } , IG (e4 ) = {v2 , v5 },

IG (e5 ) = {v2 , v5 } , IG (e6 ) = {v3 , v3 }

Vertices u and v are adjacent to each other in G if and only if there is

an edge of G with u and v as its ends. Two distinct edges e and f are said

to be adjacent if and only if they have a common end vertex.

The number of edges incident at v in G is called the degree of the

vertex v in G denoted by d(v).

A walk in a graph G is an alternating sequence W : v0 e1 vl e2 v2 . . . ep vp

of vertices and edges beginning and ending with vertices in which vi−1

and vi are the ends of ei. . A walk is called a trail if all the edges appearing

in the walk are distinct. It is called a path if all the vertices are distinct. A

cycle is a closed trail in which the vertices are all distinct. The length of a

walk is the number of edges in it.

3
 A Graph G is said to be connected if every pair of vertices u, v are con-

nected by a path. Otherwise G is said to be disconnected. The components

of G are the maximal connected subgraphs of G

Bipartite graph

A graph is bipartite if its vertex set can be partitioned into two nonempty

subsets X and Y such that each edge of G has one end in X and the other

in Y . The pair (X,Y ) is called a bipartition of the bipartite graph.

Example:

4
 Let G be a loopless graph. We construct a graph LG) in the following

way: The vertex set of L(G) is in 1-1 correspondence with the edge set

ofG and twO vertices of L(G) are joined by an edge if and only if the

corresponding edges of G are adjacent in G: The graph L(G) (which is

always a simple graph) is called the line graph or the edge graph of G.

Example:

5
MATRIX REPRESENTATION OF THE GRAPH

Adjacency Matrix

Let G be a graph with " n " vertices listed as v1 , v2 , . . . vn . The adjacency

matrix of G is the n × n matrix A(G) = [ai j ] where ai j is the number of

edges joining the vertex vi to the vertex v j

Example:

Incidence Matrix

Let G be a graph with n vertices v1 , v2 , . . . , vn and m edges e1 , e2 , . . . , em

then the incidence matrix of G is the n × m matrix M(G) = [mi j ] where

mi j is the number of times the vertex vi is incident with the edge e j . i.e,

6
 0 if vi is not an end vertex of ej

mi j = {1 if vi is an end vertex of non loop edge ej

2 if ej is a loop with vi as end vertex

Example:

7
 CHAPTER 1

HYPERGRAPHS: BASIC CONCEPTS

Hypergraph


A hypergraph H is a pair H = V ; E = (ei )i∈I where V is a set of elements

called vertices and E is a set of nonempty subsets of V called hyperedges.

Therefore E is a subset of P(V )\0/ where P(V ) is the power set of V .

Example:

Fig. 1.1 Hypergraph

8
 In this hypergraph H, V = {x1 , x2 . . . x11 } and E = {e1, e2 , e3 , e4 , e5 } =
{{x1 , x2 , x3 , x4 } , {x4 , x6 , x7 , x8 } , {x5 , x6 } , {x1 , x5 , x8 } , {x10 }}

Order of a hypergraph

The order of the hypergraph H = (V ; E) is the cardinality of V , i.e.

If V = {x1 , x2 . . . xn } then /V |= n; In fig 1.1 , | V| = 11.

Size of a hypergraph

The size of a hypergraph H = (V ; E) is the cardinality of E. i.e.

if E = {e1, e2 , e3 , e4,... , em } then |E| = m. In fig 1.1, |E| = 5.

Empty hypergraph

The empty hypergraph is the hypergraph such that:

• V = 0/

• E = 0/

9
Trivial hypergraph

A trivial hypergraph is a hypergraph such that:

• V ̸= 0/

• E = 0/

Isolated vertex
S
If i∈I ei = V the hypergraph is without isolated vertex, where a vertex x

is isolated if x ∈ V\
S
i∈l ei. In fig 1.1x11 , x9 are isolated vertices.

Loop

A hyperedge e ∈ E such that |e| = 1 is a loop.

Adjacent vertices

Two vertices in a hypergraph are adjacent if there is a hyperedge which

contains both vertices. In particular, if {x} is a hyperedge then x is adjacent

to itself.

10
Incident hyperedges

Two hyperedges in a hypergraph are incident if their intersection is nonempty.

Star

The star H(x) centered in x is the family of hyperedges (e j ) j∈J containing

x.

Degree of a hypergraph

The number of hyperedges containing the vertex x is called degree of x

denoted by d(x) and for a loop {x} the degree d(x) = 2

If the family of hyperedges is a set, i.e. if i ̸= j ⇔ ei ̸= e j , we say that H

is without repeated hyperedge.

If the hypergraph is without repeated hyperedge the degree is denoted by

d(x) = |H(x)|, excepted for a loop {x} where the degree d(x) = 2.

The maximal degree of a hypergraph is denoted by ∆(H).

11
Regular hypergraph

If each vertex has the same degree, we say that the hypergraph is regular,

or k-regular if for every

x ∈ V, d(x) = k

Rank r(H)

The rank r(H) of H is the maximum cardinality of a hyperedge in the

hypergrah:

r(H) = max |ei |

i∈l

Co-rank cr(H)

The minimum cardinality of a hyperedge is the co-rank cr(H) of H :

cr(H) = mini∈I |ei |

Uniform hypergraph

If r(H) = cr(H) = k the hypergraph is k-uniform or uniform

12
Simple hypergraph

A simple hypergraph is a hypergraph H = (V ; E) such that: ei ⊆ ej ⇔ i = j.

A simple hypergraph has no repeated hyperedge.

Linear hypergraph

A hypergraph is linear if it is simple and ei ∩ e j ≤ 1 for all i, j ∈ I where

i ̸= j.

Example:

13
 The hypergraph H above has 11 vertices; 5 hyperedges; 1 loop: e5 ; 2

isolated vertices: x11 , x9 . The rank r(H) = 4, the co-rank cr(H) = 1. The

degree of x1 is 2. Here x1 and x2 are adjacent vertices and e1 and e4 are

incident hyperedges. H is simple and linear.

Examples of hypergraph

Example 1:

Let M be a computer science meeting with k ≥ 1 sessions; S1 , S2 , S3 , . . . Sk .

Let V be the set of people at this meeting. Assume that each session is

attended by one person at least. We can build a hypergraph in the following

way:

• The set of vertices is the set of people who attend the meeting;

• The family of hyperedges (ei )i∈{1,2,...k} is built in the following way:

ei , i ∈ {1, 2, . . . k} is the subset of people who attend the meeting Si .

Example 2:

Fano Plane

14
 The Fano plane is the finite projective plane of order 2, which have

the smallest possible number of points and lines, 7 points with 3 points

on every line and 3 lines through every point. To a Fano plane we can

associate a hypergraph called Fano hypergraph:

• The set of vertices is V = {0, 1, 2, 3, 4, 5, 6};

Fig. 1.2 Fanno Hypergraph

• The set of hyperedges is E = {013, 045, 026, 124, 346, 235, 156}

• The rank is equal to the co-rank which is equal to 3, hence, the Fano

hypergraph is 3-uniform.

15
Path in a hypergraph

Let H = (V ; E) be a hypergraph without isolated vertex; a path P in H

from x to y, is a vertex- hyperedge alternative sequence:

x = x1 , e1 , x2 , e2 , . . . , xs , es , xs+1 = y such that

• x1 , x2 , . . . , xs , xs+1 are distinct vertices with the possibility that x1 =

xs+1 ;

• e1 , e2 , . . . , es are distinct hyperedges;

• xi , xi+1 ∈ ei,(i = 1, 2, . . . , s .

– If x = x1 = xs+1 = y the path is called a cycle.

– The integer s is the length of path P

– if there is a path from x to y there is also a path from y to x. In this

case we say that P connects x and y.

Connected hypergraph

A hypergraph is connected if for any pair of vertices, there is a path which

connects each pair of vertices; it not connected otherwise. In this case we

may also say that it is disconnected.

16
Distance

The distance d(x, y) between two vertices x and y is the minimum length of

a path which connects x and y. If there is a pair of vertices x, y with no path

from x to y (or from y to x ), we define d(x, y) = ∞(H is not connected).

Connected component

Let H = (V, E) be a hypergraph, a connected component is a maximal set

of vertices X ⊆ V such that, for all x, y ∈ X, d(x, y) ̸= ∞

Diameter

The diameter d(H) of H is defined by d(H) = max{d(x, y)/x, y ∈ V }.

Example:

The hypergraph above has 2 connected components, C1 ,C2 .P = x10 e4 x5 e3 x6 e2 x4 e1 x3

is a path from x10 to x3 , P′ = x10 e4 x1 e1 x3 is also a path from x10 to x3 and

the distance d (x10 , x3 ) = 2 is the length of P′ . Notice that the distance

d (x10 , x3 ) is also the length of the path P′′ = x10 e4 x2 e1 x3 .

17
Complete hypergraph

A hypergraph H is complete if H = (V ; E = P(V )\{0}).

/

For n = |V |, a complete k-uniform hypergraph on n ≥ k ≥ 2 vertices is a

hypergraph which has all k-subsets of V as hyperedges, i. e. E = Pk (V ),

where Pk (V ) is the set of all k-subset of V ; it is denoted by Knk .

Matrix representation of hypergraphs

Incidence matrix

Let H = (V ; E) be a hypergraph, without isolated vertex, V = {v1 , v2 , . . . , vn }

and E = (e1 , e2 , . . . , em ). Then H has an n × m incidence matrix A = ai,j

18
where 
 0 if v j ∈ e j

(ai, j ) =
 1 otherwise


Example:

x1 x2 x3 x4 x5
 
e1  1 1 1 0 0
 
0 0
e2  1 1 0
 
e3 1 1 0 0 1

19
Adjacency matrix

It is a square matrix which rows and columns are indexed by the vertices

of H and for all x, y ∈ V, x ̸= y the entry ax,y = |{e ∈ E : x, y ∈ e}| and

ax,x = 0.

Example:

 
 x1 x2 x3 x4 x5 
 

 x1 0 2 1 0 1  
 
x2 2 0 1 0 1 
 

 
 

 x3 1 1 0 1 0  
 

 x4 0 0 1 0 0  
 
x5 1 1 0 0 0

20
Induced Subhypergraph


Let H = V ; E = (ei )i∈I be a hypergraph. The induced subhypergraph

H (V ′ ) of the hypergraph H where V ′ ⊆ V is the hypergraph H (V ′ ) =

(V ′ , E ′ ) defined as E ′ = {V (ei ) ∩V ′ ̸= 0;
/ ei ∈ E and either ei is a loop or
|V (ei ) ∩V ′ | ≥ 2}.

Example:

21
 Consider the above hypergraph H.

H V ′ = V ′ = {x1 , x4 , x6 , x8 , x10 } ; e′1 = e1 ∩V ′ = {x1 , x4 } ;

e′2 = e2 ∩V ′ = {x4 , x6 , x8 } ; e′4 = e4 ∩V ′ = {x1 , x8 } ;

e′5 = e5 ∩V ′ = {x10 } is an induced subhypergraph.

Notice that e′3 = e3 ∩ V ′ = {x6 } is not an hyperedge for this induced

hypergraph.

22
Subhypergraph
 
Let H = V ; E = (e j ) j∈J be a hypergraph. Given a subset V ′ ⊆ V , the
 
′ ′ ′ ′
subhypergraph H is the hypergraph H = V , E = (e j ) j∈J such that for

all e j ∈ E ′ : e j ⊆ V ′ .

Example:

For the hypergraph H, H ′ = (V ′ = {x1 , x2 , x3 , x4 , x7 } , E = {e1 }) is a sub-

hypergraph with 1isolated vertex: x7 .

Partial hypergraph
 
Let H = V ; E = (e j ) j∈J be a hypergraph. A partial hypergraph gen-
 
′ ′ ′
erated by J ⊆ I, H of H is a hypergraph H = V , (e j ) j∈J , where

U j∈J e j ⊆ V ′ . Notice that we may have V ′ = V .

23
Example

For the hypergraph H, H ′ = (V ; {el , e2 }) is a partial hypergraph generated

by J = {1, 2}

24
 CHAPTER 2

HYPERGRAPHS: FIRST
PROPERTIES

Line graph


Let H = V ; E = (ei )i∈I be a hypergraph such that E ̸= 0.
/ The line-

graph (or representative graph) of H is the graph L(H) = (V ′ ; E ′ ) such

that: 1. V ′ := I or V ′ := E when H is without repeated hyperedge; 2.

{i, j} ∈ E ′ (i ̸= j) if and only if ei ∩ ej ̸= 0.

/

Example

Figure above shows a hypergraph H = (V ; E), where V = {x1 , x2 , x3 , . . . , x9 , },

E = {e1 , e2 , e3 , e4 , e5 }, and its representative. The vertices of L(H) are the

black dots and its edges are the curves between these dots.

Lemma

The hypergraph H is connected if and only if L(H) is connected.

25
Proposition

Any nontrivial graph Γ is the line-graph of a linear hypergraph

Proof

Let Γ = (V, E) be a graph with V = {xI , x2 , x3 , . . . xn }. Without losing of

generality we suppose Γ is connected. (Otherwise, we treat connected

components one by one). We can construct a hypergraph H = (W ; X) in

the following way:

• The set of vertices is the set edges of Γ i.e. W = E. It is possible

since Γ is simple;

26
 • The collection of hyperedges X is the family of Xi where Xi is the set

of edges of Γ having xi as incident vertex.

So we can write; H = (E; (X1 , X2 , . . . . . . ..Xn ) with Xi = {e ∈ E; xi ∈ e |

where i ∈ {1, 2, 3, . . . n}.

Notice that if Γ has only one edge then V = {x1 , x2 } and X1 = X2 .It is the

only case where H has a repeated hyperedge.

If |E| > 1, if i ̸= j and Xi ∩ X j ̸= φ ; there is exactly one, (since Γ is a simple

graph ) e ∈ E suchthat e ∈ Xi ∩ X j with e = (xi , x j ). It is clear that Γ is a

linear graph of H.

27
Dual of a hypergraph

Let H = (V ; E) be a hypergraph which is without isolated vertex. A dual

H ∗ = (V ∗ ; E ∗ ) of H isa hypergraph such that:

• the set of vertices, V ∗ = {x1 ∗, x2∗ , . . . , xm

∗ } is in bijection f with the set

of hyperedges E;

• the set of hyperedges is given by: e1 ∗ = X1 , e2 ∗ = X2 , . . . , en ∗ = Xn ,

where e∗j = X j = f (ei ) = xi∗ : x j ∈ e j . Without loss of general-



ity, we identify V ∗ with E. Hence e∗j = X j = ei : x j ∈ e j , for j ∈



{1, 2, . . . n}.

• So there is a bijection g from the hyperedges E of H to the vertices

V ∗ of H ∗ .

28
Example

Left side of the above figure represents a hypergraph H = (V ; E)

with V = {x1 , x2 , x3 , x4 , x5 } and set of hyperedges E = {e1 , e2 , e3 }. Right

side represents the dual H ∗ = (V ∗ ; E ∗ ) with V ∗ = {e1 , e2 , e3 } and E ∗ =

(Xi )i∈{1,2,3,4,5} .

We notice that H is without repeated hyperedge but its dual has one.

Proposition

The dual H ∗ of a linear hypergraph without isolated vertex is also linear.

29
Proof

Let H be a linear hypergraph. Assume that H ∗ is not linear. There is

two distinct hyperedges Xi and X j of H ∗ which intersect with at least two

vertices e1 and e2 . The definition of duality implies that xi and x j belong

to e1 and e2 (the hyperedges of H standing for the vertices e1 , e2 of H ∗

respectively) so H is not linear. Contradiction.

2-Section of a hypergraph

Let H = (V ; E) be a hypergraph, the 2-section of H is the graph, denoted

by [H2 , which vertices are the vertices of H and where two distinct vertices

form an edge if and only if they are in the same hyperedge of H. Example

30
Degree of a hyperedge

Let H = (V ; E) be a hypergraph, the degree of a hyperedge, e ∈ E is its

cardinality, that is d(e) = |e|.

Incidence graph

The incidence graph of a hypergraph H = (V ; E) is a bipartite graph IG(H)

with a vertex set S = VUE, and where x ∈ V and e ∈ E are adjacent if and

only if x ∈ e.

31
Proposition

Let H = (V ; E) be a hypergraph, we have: ∑x∈V d(x) = ∑e∈E d(e).

Proof :

Let IG(H) be the incidence graph of H. We sum the degrees in the part

E and in the part V in IG(H). Since the sum of the degrees in these two

parts are equal we obtain the result.

32
Intersecting Families


Let H = V ; E = (ei )i∈I be a hypergraph. A subfamily of hyperedges

(e j ) j∈J , where J ⊆ I is an intersecting family if every pair of hyperedges

has a nonempty intersection.

Example:

The maximum cardinality of |J| (of an intersecting family of H ) is

denoted by ∆0 (H). An intersecting family with 3 hyperedges e1 , e2 , e3 and

e1 ∩ e2 ∩ e3 = 0/ is called a triangle.

33
Helly Property and Strong Helly Property

A hypergraph has the Helly property if each intersecting family has a non-

empty intersection. Itis obvious that if a hypergraph contains a triangle it

has not the Helly property. A hypergraph having the Helly property will

be called Helly hypergraph.

A hypergraph has the strong Helly property if each partial induced

subhypergraph has the Helly property.

Example :

The hypergraph above has the Helly property but not the strong Helly

property because the induced subhypergraph on Y = V \ {x4 } contains the

triangle:

e′1 = e1 ∩Y, e′2 = e2 ∩Y, e′3 = e3 ∩Y

34
35
 This hypergraph has no helly property since, intersecting family e1 , e3 , e4

has an empty intersection, that is, e1 ∩ e3 ∩ e4 = 0.

/

36
 CHAPTER 3

HYPERGRAPH COLORING

Coloring


Let H = V ; E = (ei )i∈I be a hypergraph and k ≥ 2 be an integer. A

k-coloring of the vertices of H is an allocation of colors to the vertices

such that:

(i) A vertex has just one color.

(ii) We use k colors to color the vertices.

(iii) No hyperedge with a cardinality more than 1 is monochromatic.

From this definition it is easy to see that any coloring induces a partition

of the set of vertices in k - classes: (Cl ,C2 ,C3 , . . . ,Ck ) such that for e ∈

E(H), |e| > 1 then e not a subset of Ci for all i ∈ {1, 2, . . . k}.

Example

The figure below shows a colored hypergraph H where (r) is red and (b) is

blue.

37
Chromatic number χ(H)

The chromatic number χ(H) of H is the smallest k such that H has a

k-coloring. Here we have χ(H) = 2

Example:

If H is the hypergraph which vertices are the different waste products of

a chemical production factors, and which hyperedges are the dangerous

combination of these waste products. The chromatic number of H is the

smallest number of waste disposal sites that the factory needs in order to

avoid any dangerous situation.

38
Proposition

For any hypergraph H = (V ; E) with an order equal to n, we have: χ(H) ·

α(H) ≥ n.

Proof

Let (C1 ,C2 ,C3 , . . . ,Ck ) be a k - coloring of H with k = χ(H), we know

that for e ∈ E , |e| > 1 then e ← Ci , for all i ∈ {1, 2, 3, . . . k}, consequently

Ci is a stable for all i ∈ {1, 2, 3, . . . k}. Hence |Ci| ≤ α(H), for all i ∈

{1, 2, 3, . . . k}. We have: n = ∑ki=1 |ci | ≤ k · α(H) = χ(H) · α(H).

Proposition

If H = (V ; E) is a hypergraph with an order equal to n, we have: χ(H) +

α(H) ≤ n + 1

Proof

Assume that S is a stable with |S| = α(H). We can color any vertex of S

with the same color andusing n − α(H) other colors to color the set V − S

39
with different colors. Hence we have :

χ(H) ≤ n − α(H) + 1

that leads to

χ(H) + α(H) ≤ n + 1.

Particular Colorings

Strong Coloring

Let H = (V ; E) be a hypergraph, a strong k-coloring is a partition (C1 ,C2 , . . . ,CK )

of V such that the same color does not appear twice in the same hyperedge.

In another word: |e ∩Cl | ≤ 1 for any hyperedge and any element of the

partition.

The strong chromatic number denoted by χ ′ (H) is the smallest k such

that H has a strong k-coloring.

Equitable Coloring

Let H = (V ; E) be a hypergraph, an equitable k-coloring is a k-partition

(C1 ,C2 , . . ., CK ) of V such that, in every hyperedge e, all the colors

{1, 2, . . . , k} appear the same number of times, to within one, if k does not

40
divide |ei |. That is for all e ∈ E,

  
|e| |e|
| ≤ |e ∩ Ci |≤ i ∈ {1, 2, 3 . . . k}
k k

Example:

The figure shows a hypergraph H = (V ; E) and an equitable 2- coloring

of it, where (r) is red and (b)is blue.

Good Coloring

Let H = (V ; E) be a hypergraph, a good k-coloring is a k-partition (C1 ,C2 , . . . ,CK )

of V such that every hyperedge e contains the largest possible number

41
of different colors, i.e. for every e ∈ E, the number of colors in e is

min{|e|; k}.

Lemma

Let H = (V ; E) be a hypergraph (with m = |E|, and C = (C1 ,C2 , . . . ,CK )

be a good k-coloring of H, we have:

(i) if k ≤ c r(H),(c r(H) is the co-rank of H) then C is a partition in k

transversal sets;

(ii) if k ≥ r(H) then the good coloring C is a strong coloring

Proof:

Assume that k ≤ cr(H). By definition of a good coloring, if Ci is a set of

vertices with color i, we must have:

Ci ∩ e j ̸= 0,
/ ∀ j ∈ {1, 2, . . . , m}.

Hence Ci is a transversal of H. Assume now that k ≥ r(H). Let e ∈ E, then

k ≥| e, any two vertices belonging to e have different colors. Consequently,

by definition of a strong coloring, the good coloring C is a strong coloring.

42
Uniform Coloring

Let H = (V ; E) be a hypergraph with |V | = n. A uniform k-coloring is a

k-partition: (C1 ,C2 , . . . ,CK ) of V such that the number of vertices of the

same color is always the same, to within one, if k does not divide n, i.e

   
|n| |n|
≤ |Ci | ≤ , i ∈ {1, 2, 3 . . . k}
k k

Hyperedge Coloring

Let H = (V ; E) be a hypergraph, a hyperedge k-coloring of H is a coloring

of the hyperedges such that:

(i) A hyperedge has just one color.

(ii) We use k colors to color the hyperedges.

(iii) Two distinct intersecting hyperedges receive two different colors

Chromatic index

The size of a minimum hyperedge k-coloring is the chromatic index of H.

We will denote it by q(H).

43
Lemma

Let H be a hypergraph. We have: q(H) ≥ ∆0 (H) ≥ ∆(H). Where ∆0 (H)

is the maximum cardinality of the intersecting families and ∆(H) is maxi-

mum cardinality of the stars.

Proof

Assume that ∆0 (H) = l. We need l distinct colors to color an intersecting

family with at least l hyperedges. Hence

q(H) ≥ ∆o (H) ≥ ∆(H)

Hyperedge coloring property

A hypergraph has the hyperedge coloring property if q(H) = ∆(H). For

instance a star has the hyperedge coloring property.

44
 CHAPTER 4

APPLICATION OF HYPERGRAPH
THEORY
Like in most fruitful mathematical theories, the theory of hypergraphs has

many applications. Hypergraphs model many practical problems in many

different sciences. We find this theory in psychology, genetics, etc. but

also in various human activities. Hypergraphs have shown their power as

a tool to understand problems in a wide variety of scientific field.

Moreover it well known now that hypergraph theory is a very use-

ful tool to resolve optimization problems such as scheduling problems,

location problems and so on. This chapter shows some possible uses

hypergraphs in Applied Sciences.

Chemical Hypergraph Theory

The graph theory is very useful in chemistry. The representation of molec-

ular structures by graphs is widely used in computational chemistry. But

the main drawback of the graph theory is the lack of convenient tools to

represent organometallic compounds, benzenoid systems and so on.

45
 A hypergraph H = (V, E) is a molecular hypergraph if it represents

molecular structure, where x ∈ V corresponds to an individual atom, hy-

peredges with degrees greater than 2 correspond to polycentric bonds and

hyperedges with deg(x) = 2 correspond to simple covalent bonds.

Hypergraphs appear to be more convenient to describe some chemical

structures. Hence the concept of molecular hypergraph may be seen as a

generalization of the concept of molecular graph. Hypergraphs have also

shown their interest in biology.

Hypergraph Theory for Telecommunication

A hypergraph theory can be used to model cellular mobile communication

systems. A cellular system is a set of cells where two cells can use the

same channel if the distance between them is at least some predefined

value D. This situation can be represented by a graph where:

(a) Each vertex represents a cell.

(b) An edge exists between two vertices if and only if the distance be-

tween the corresponding cells is less than the distance called reuse

distance and denoted by D.

A forbidden set is a group of cells all of which cannot use a channel

simultaneously. A minimal forbidden set is a forbidden set which is

46
minimal with respect to this property, i.e. no proper subset of a minimal

forbidden set is forbidden. From these definitions it is possible to derive a

better modelization using hypergraphs. We proceed in the following way:

(a) Each vertex represents a cell.

(b) A hyperedge is minimal forbidden set.

Hypergraphs and Constraint Satisfaction Prob-

lems

A constraint satisfaction problem, P is defined as a tuple: P = (V, D, R1 (S1 ) , . . . , Rk

(Sk )) where:

• V is a finite set of variables.

• D is a finite set of values which is called the domain of P.

• Each Ri (Si ) is a constraint.

– −Si is an ordered list of ni variables, called the constraint scope.

– −Ri is a relation over D of arity ni , called the constraint relation.

To a constraint satisfaction problem one can associate a hypergraph in

the following way:

47
 (a) The vertices of the hypergraph are the variables of the problem.

(b) There is a hyperedge containing the vertices v1 , v2 , . . . , vt when there

is some constraint Ri (Si ) with scope Si = {v1 , v2 , . . . vt }.

Hypergraphs and Database Schemes

Hypergraphs have been introduced in database theory in order to model

relational database schemes.

A database can be viewed in the following way:

• A set of attributes.

• A set of relations between these attributes.

Hence we have the following hypergraph:

• The set of vertices is the set of attributes.

• The set of hyperedges is the set of relations between these attributes.

48
 CONCLUSION
A hypergraph is defined to be a family of hyperedges which are sets of

vertices of cardinality not necessarily 2 (as for graphs). We have discussed

basic definitions, some properties, colorings, and certain applications of

hypergraph theory. Hypergraphs model practical situations in different

sciences in a much more general setting than graphs do. Hypergraph theory

is used in chemistry, the engineering field, telecommunication, database

schemes, etc.

49
 BIBLIOGRAPHY
1. Hypergraph Theory - An Introduction by Alain Bretto

2. Introduction to Graph and Hypergraph Theory by Vitaly [Link]

3. A Textbook of Graph Theory - Second Edition by R. Balakrishnan

and K. Ranganathan

4. A First Look at Graph Theory by John Clark and Derek Allen Holton

50
DON BOSCO ARTS & SCIENCE
COLLEGE ANGADIKADAVU

DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

NETS AND FILTERS IN TOPOLOGY

 DON BOSCO ARTS & SCIENCE COLLEGE
ANGADIKADAVU
DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

NETS AND FILTERS IN TOPOLOGY

Dissertation submitted in the partial

Fulfillment of the requirement for the award of
MSc Degree in Mathematics of
Kannur University
Name : RAHINA
Roll No. : C1PSMM1907

Examiners:
1.
2.
 KANNUR UNIVERSITY

BONAFIDE CERTIFICATE

Certified that this project report “NETS AND FILTERS IN TOPOL-

OGY” is the bonafide work of RAHINA who carried out the project
work under my supervision.

Signature of Head Signature of Supervisor

 DECLARATION

I, RAHINA hereby declare that the Project work entitled NETS

AND FILTERS IN TOPOLOGY has been prepared by me and sub-
mitted to Kannur University in partial fulfillment of requirement for
the award of Master of Science is a record of original work done by
me under the supervision of Mr. ANIL M V, Assistant Professor,
Department of Mathematics, Don Bosco Arts & Science College,
Angadikadavu. I, also declare that this Project work has not been
submitted by me fully for the award of any Degree, Diploma, Title
or recognition before any authority.

Place: Angadikadavu RAHINA

Date: (C1PSMM1907)
 ACKNOWLEDGEMENT

Introduction is the proper place to begin. But first I bow my head

before the Almighty who is always with me. Also I must express
my deepest gratitude to people along the way. No words can ad-
equately express the sense of gratitude, still I try to express my
heartfelt thanks through words. The outset, I am deeply indebted
to my project supervisor Mr. ANIL M V Assistant Professor, De-
partment of Mathematics, Don Bosco Arts and Science College, An-
gadikadavu, for the invaluable guidance, loving encouragement and
meticulous care towards me throughout my career. I express my deep
sense of gratitude to all the faculty members of the Department of
Mathematics, Don Bosco Arts and Science College, Angadikadavu.
I can never forget the support and encouragement rendered by the
Principal and the Staff of Don Bosco Arts & Science College, An-
gadikadavu. I could not name many who sincerely supported and
helped for the successful completion of this Project. It is my plea-
sure and duty to thank each and everyone of them who walked with
me.

RAHINA
 CONTENTS

1. INTRODUCTION ........................................................... 01

2. PRELIMINARY RESULTS ...........................................03

3. TOPOLOGICAL PROPRTIES .................................... 08

AND SEQUENCES

4. NETS IN TOPOLOGY....................................................12
NETS AND ITS CONVERGENCE ......................................12
TOPOLOGICAL PROPERTIES AND NETS........................17

5. FILTERS IN TOPOLOGY ...............................................26

FILTERS AND ITS CONVERGENCE...................................26
CHARECTERISATION OF FILTERS...................................32
ULTRAFILTERS....................................................................36

6. CONCLUSION....................................................................41

7. BIBLIOGRAPHY ...............................................................42
 INTRODUCTION
When we are working on metric spaces ( or first countable topologi-
cal spaces) sequences are sufficient to describe almost all topological
properties. This dissertation mainly deals with the generalisation
of sequences called nets and the concept of filters. The goal of the
first chapter is to justify via examples the fact that sequences are
not enough to describe general topological spaces. Here some basic
definitions and results are included which we use further.
The second chapter deals with the definition of nets and basic proper-
ties of nets which ensure that nets are the generalisation of sequences.
Later we characterise some of the topological properties; such as
Hausdorff property, continuity of a function, closure of set, etc. Last
part of this chapter gives you an interesting proof of Tychonoff the-
orem via nets that proves the importance of nets in analysis.
The last chapter describes the use of filters and its interesting applica-
tions. Here first section talks about the definition and convergence of
filters with some examples while the next section characterise topo-
logical properties which is more easier compared to that of nets.
Coming to the end, we are introducing the concept of ultrafilters
and gives another proof of Tychonoff theorem using ultrafilters.

1
 PRELIMINARY

Definition
A Topological space is a pair (X,τ ) where X is a non-empty set and
τ is a family of subsets of X satisfying:
(i) ϕ ∈ τ and X ∈ τ .
(ii) τ is closed under arbitrary union
(iii)τ is closed under finite intersection.
The members of τ are called open sets and the compliment of open
sets are called closed sets.

Definition: Sequences and its convergence

A sequence {xn } in a topological space (X,τ ) is a function from N to

X. A sequence {xn } is said to be converge to a point x if for every
open set U containing x, there exists a positive N such that for every
integer n ≥ N, xn ∈ U.

Example

Let X be any non-empty set.

1) Indiscrete topology : The collection, τ ={ϕ, X} is the smallest
topology on X. This topology is called as Indiscrete topology.
In this topology every sequence converges to every point.

2
2) Discrete topology : If P(X) is the power set of X, then τ = P(X)
is a topology on X and is called as Discrete topology on X.
In this topology only eventually constant sequence is convergent and
it converges to the repeating point of the sequence.
3) Co-finite topology : Let

τcf = {A ⊆ X : Ac is f inite set}

is a topology on X. It is called as Co-finite topology.

In this topology only eventually constant sequence and sequence with
infinitely many distinct points are convergent. First sequence con-
verges to the repeating point and the second sequence converges to
every point of X.
4) Co-countable topology : The collection of subsets A of X such that
Ac is countable is a topology on X and it is called as Co-countable
topology.

ie, τcc = {A ⊆ X : Ac is countable}

In this topology, only eventually constant sequence converges and

it converges to the repeating point .

3
 Definition: Hausdorff space

A space X is said to be Hausdorff space if for every distinct points x,

y ∈ X there exists disjoint open sets U,V in X such that x ∈ X and
y ∈ X.

Definition: Limit Point of a Set

Let A be a subset of a topological space X and y ∈ X. Then y is said

to be a limit point of A if every open set U containing y contains
atleast one point of A other than y.
i.e, (A / {y}) ∩ U ̸= ϕ.

Definition: Closure of a Set

The closure of a subset A of a topological space X is defined as the

intersection of all closed sets containing A. It is denoted by A.
i.e, Closure of a set A is the union of A and its derived set, where
derived set of A is the collection of all limit points of A.
Clearly, A is the smallest closed set containing A.
Equivalently, a set A is closed if it contains all of its limit points, i.e,
if A = A.

Definition: Function

Let X and Y be two sets. A function from X to Y is a subset of X

× Y with the property that for each x ∈ X, there is a unique y ∈

4
Y such that (x, y) ∈ X × Y .The set X, Y and f are called domain,
codomain and graph of (X, Y, f) respectively. We also denote (X, Y,
f) by f : X → Y .

Definition: Continuity of a Function

Let X and Y be topological spaces and f : X → Y be a function.

Then f is said to be continuous on X if f −1 (U) is open in X, for every
open set U in Y.

Theorem: In a Hausdorff space X, every sequence converges to at

most one limit

proof

Let X be a Hausdorff space.

Let (xn ) be a any sequence converges to x and y.
Assume x ̸= y.
Since X is Hausdorff, ∃ opensets U and V ∋ x ∈ U and y ∈ V and
U ∩ V = ϕ.
xn → x ⇒ ∃ m1 ∈ N ∋ xn ∈ U n ≥ m1 .
xn → y ⇒ ∃ m2 ∈ N ∋ xn ∈ V n ≥ m2 .
Then xn ∈ U ∩ V n ≥ max{m1 , m2 } which is a contradiction to the
fact that U ∩ U = ϕ.
Hence x = y.

5
Definition: Cover of a set

A cover of a subset A of X is the collection A = {Ai : i ∈ I} of

S
subsets of X such that A ⊆ i∈I Ai .
Any subfamily of A which covers A is called a subcover of A.

Definition: Compact set

A subset A of a space X is said to be a compact subset if every cover

of A by open sets of X has a finite subcover [subcover containing fi-
nite number of sets]. A space is said to be compact if it is a compact
subset of itself.

Definition: Finite Intersection Property(f.i.p)

A family F of subsets of a set X is said to have the finite intersection

property if for any n ∈ N and F1 ,F2 , ..., Fn ∈ F, the intersection
Tn
i=1 Fi is non-empty.

Theorem: A topological space is compact if and only if every family

of closed subsets of it, which has f.i.p, has non-empty intersection.

6
 proof

Let X be any topological space.

Suppose X is compact.
Let C be any family of closed sets having f.i.p.
T
Claim: C∈C C ̸= ϕ
Suppose
T
C∈C C=ϕ
( C∈C C)c = X
T

c
S
C∈C C =X

C is closed implies C c is open.

∴ {C c } for C ∈ C ,is an open cover for X.
Sn
Since X is compact, there exists C1c , C2c ....... Cnc such that i=1 Cic
=X
⇒ ni=1 Ci = ϕ, a contradiction to the choice of C .
T
T
Hence C∈C C ̸= ϕ
The converse part is almost similar only use Demorgan’s law and the
definition,complement of open set is closed.

7
 CHAPTER 1

Topological properties and sequences

Example 1

Let X be an uncountable set with co-countable topology.

We know that the only converging sequence are eventually constant
sequences and the limit is unique.
But X is not Hausdorff space.
For let x and y be any two distinct points.
Let U be any open set of X containing x.
If we want X to be Hausdorff we need an open set contained in Uc .
ie, an open set containing countable elements, since U is open.
Then Vc will be uncountable, as X is uncountable.
∴ ∄ any open V ∋ y ∈ V and U ∩ V = ϕ.
Hence, X is not a Hausdorff space.
So, uniqueness of limit of converging sequence is not enough
to characterise the Hausdorff property.

8
 Example 2

Let X be an uncountable set.

Let (X, σ) be the discrete topology and (X, τ ) be the co-countable
topology.
We know that in both topology, the only convergent sequence are
the eventually constant sequences.
Let us define the identity function, I : (X, τ ) → (X, σ).
Since τ and σ have same convergent sequences, the function, I is
sequentially continuous.
However since τ → σ, the function I is not continuous.
Hence, sequences do not characterise continuity of a func-
tion.

Example 3

Let X be an uncountable set and fix x0 X.

/ A) or (x0 ∈ A and Ac is countable)}.
Define ω = {A ⊆ X : (x0 ∈
Then ω is a topology for X.
(a) ϕ, X ∈ ω.
(b) A, B ∈ ω ⇒ A ∩ B ∈ ω.
S
(c) A ∈ ω, ∈ J implies that α (Aα ) ∈ ω.
If x0 ∈ Aα0 then ( α Aα )c = α Acα ⊆ Aα0 , which is [Link]
S T
S
α Aα ∈ ω.

9
Claim: xn → x if xn = x eventually.
Case 1: x ̸= x0 ,
then {x} ∈ ω, then by definition of convergence,xn → x if and only
if v = x eventually.
Case 2: x = x0 .
Assume xn ̸= x0 for infinitely many n.
Define F = {xn : xn ̸= x0 }.
Then x0 ∈ F c and (F c )c = F, countable ⇒ F c ∈ ω.
ie, F c is an open neighbourhood of x0 which fails to hold xn ∈ F c ∀
n ≥ n0 for any n0 ∈ N.
Hence xn ̸= x0 is wrong.
⇒ xn = x0 = x eventually.
Conversely, if xn ↛ x0 , then ∃ O ∈ ω , x0 ∈ O ∋ xn ∈
/ O infinitely
often.
ie, xn ̸= x0 infinitely often.
Hence xn → x0 if and only if xn = x0 eventually.
Let A = ({x0 })c
Then x0 ∈ A, but there does not exist a sequence {xn } in A such
that xn → x0 in ω because any sequence in A satisfy xn ̸= x0 n ≥ 1.
Hence by above result, xn does not converge to x0 .
But for any open set O such that x0 ∈ O, we have O ∩ A ̸= ϕ,
∴ Oc is countable.

10
∴Every neighbourhood of x0 intersect A implies x0 ∈ A.
ie, Sequences do not characterise points of closure.
Note : A metric space is compact if and only if it is sequentially
compact.
But we cannot generalise compactness using sequential compactness.
Moreover there exist compact space which is not sequentially com-
pact and vice versa.

11
 CHAPTER 2

Nets in topology
2.1 Nets and its convergence

Definition [Link] set

A directed set is a pair (D, ≥) where D is a non-empty set and ≥ is
a binary relation on D satisfying:
(1) For all m, n, p ∈ D, m ≥ n and n ≥ p imply m ≥ p.
(2) For all n ∈ D, n ≥ n.
(3) For all m, n ∈ D, there exists p ∈ D suchthat p ≥ n and p ≥ m.

Example

1) For x ∈ X,ηx denote the set of all neighbourhood system [ Collec-

tion of neighbourhood of x ] in the topological space, X. We define
U ≥ V to mean U ⊆V in ηx . Then clearly ηx is directed set.
2) A partition of unit interval [0, 1] is a finite sequence sequence
P={a0 ,a1 ,.....an } such that 0 = a0 < a1 < ......< an =1 If Q is a
partition of [0, 1] which contains P, then Q is said to be a refinement
of P.
Define P ≥ Q if P is refinement of Q. Then the collection of all par-
titions of [0, 1]is a directed set.

12
Definition 2.1.2 Nets

A net in a set X is a function S : D → X, where D is a directed set

Example

Riemann net: Let f be a bounded real valued function from [0, 1] and
D be the set of all pairs (P, ξ) where P is a partition of [0, 1]. Say P
= {a0 ,a1 ,.......an } and ξ = (ξ1 , ξ2 , ..., ξn ) is finite sequence such that
ξi ∈ [ai1, ai] for i = 1, 2, ..., n. And define (P, ξ) (Q, η) of D if P is
a refinement of Q. Now the Riemann net S : D → R is defined by,
S(P, ξ) = ni=1 f (ξi )(ai -ai+1 )
P

Definition 2.1.3 Convergence of a net

Let (X,τ ) be a topological space and let S : D → X be a net. Then

S is said to be converging to a point x ∈ X, if given any open set
U containing x, there exists m ∈ D such that for all n ∈ D, n ≥ m
implies that S(n) ∈ U.
Note: As the partition becomes finer and finer, the limit of Riemann
net becomes the Riemann integral.

Definition 2.1.4 Eventual subset

A subset E of a directed set D is said to be eventual if there exists

m ∈ D such that for all n ∈ D, n ≥ m implies that n ∈ E.
A net S : D → X is said to be eventually in a subset A of X if the

13
set S 1 (A) is an eventual subset of D.
ie, a subset E is eventually if and only if it contains all elements in
D ’after a certain stage’ and a net S : D → X eventually in A if and
only if A contains all its terms after a certain stage

Definition 2.1.5 Cofinal subset

Let (D, ≥) be a directed set. A subset F of D is said to be a cofinal

subset of D if for every m ∈ D, there exists n ∈ F such that n ≥ m.
A net S : D → X is said to be frequently in a subset A of X if S 1 (A)
is a cofinalsubset of D.
• Every eventual subset is a cofinal subset of D.
• Converse is not true: In N every infinite subset is cofinal but it is
not necessarily eventual.
• Cofinal set in the neighbourhood system of a point in a topological
space is a local base at that point.
• In the set of all partitions of the unit interval [0, 1], the set of par-
titions of the form {0, n1 , n2 ...... n−1
n ,1} for n ∈ N is not a cofinal subset.

If we direct the set of all partitions in terms of norm [the length

of the longest subinterval ] such that P follows Q if and only if the
norm of P is less than or equal to that of Q, then it is a cofinal subset.

14
 Definition 2.1.6 Cluster point

Let S : D → X be a net. A point x ∈ X is said to be a cluster point

of S if for every neighbourhood U of x in X, and m ∈ D, there exists
n ∈ D such that n ≥ m and S(n) ∈ U.
Equivalently, x is a cluster point of S if and only if for every neigh-
bourhood U of x, S is frequently in U.

Proposition 2.1.1 Suppose S : D → X is a net and F is a cofinal

subset of [Link] S|F : F → X converges to a point x in X, then x is a
cluster point of S.

proof

Let U be a neighbourhood of x in X.
Let m ∈ D be given.
S|F → x ⇒ ∃ m1 ∈ F ∋ ∀ n ∈ F, n ≥ m, ⇒ Sn ∈ U
Choose n1 ∈ D such that n1 ≥ m and n1 ≥ m1 .
Since F is cofinal subset we can find n ∈ F ∋ n ≥ m1 .
Then n ≥ n1 ≥ m and S(n) ∈ U. So S is frequently in U.
Since U was arbitrary, x is a cluster point of S.

Definition 2.1.7 Subnet

Let S : D → X and T : E → X be nets. Then T is said to be a subnet

of S if there exists a function N : E → D such that,

15
(i) T = S ◦ N.
(ii) for any n ∈ D, there exists p ∈ E such that for all m ∈ E, m ≥
p implies N(m) ≥ n in D.

Theorem 2.1.1 Let S : D → X be a net in a topological space and

let x ∈ [Link] x is a cluster point of X if and only if there exist a
subnet of X which converges to x.

proof

Suppose T : E → X is a subnet of S converging to x.

claim: x is a cluster point of S.
Let N : E → D be the function such that T = S ◦ N.
Let U be any neighbourhood of x in X and let m1 D be given. From
the definition of subnet (ii), there exists p ∈ E such that for all m ∈
E , m ≥ p implies N(m) ≥ m1 .
Since T converges to x, there exists q ∈ E such that for all m ∈ E ,
m ≥ q implies T(m) ∈ U, ie, SN(m) ∈ U.
Choose n1 ∈ E such that n1 ≥ p and n1 ≥ q. Let N(n1 ) = n0 . Then
n0 ≥ m1 and S(n0 ) ∈ U.
Since m1 and U were arbitrary, we get x is a cluster point of S.
Conversely, suppose x is a cluster point of S
Let ηx be a neighbourhood system of the point x in X. Define,
E = {(n, U) ∈ D × ηx : S(n) ∈ U}

16
For (n, U) , (m, V ) ∈ E we let (n, U) ≥ (m, V ) means n ≥ m and
U⊆V.
It is easy to show that E is a directed set with the defined binary
relation.
Define T : E → X by T(n,U) = S(n) for any (n, U) ∈ E.
For N : E → D suchthat N(n, U) = n we get that T = SN(n, U) =
S(n) and for any m ∈ D, for any U’ ∈ ηx m1 ∈ D ∋ m1 ≥ m and
S(m1 ) ∈ U’.
Then (m1 , U’) ∈ E. Now for (n, U) ≥ (m1 , U’) n ≥ m2 ≥ m.
Hence T is a subnet of S.
Now we will prove that T → x ∈ X.
Let G be any open set of x in X. Since x is a cluster point of S, we
get S is frequently in G.
Fix n0 ∈ D ∋ S(n0 ) ∈ G. Then (n0 , G) ∈ E.
Now (m, G) ∈ E , (m, U) ≥ (n0 , G) ⇒ T(m, U) = Sm ∈ U ⊆ G.
Thus T converges to x in X.
Hence the proof.

2.2 Topological properties and nets

In this section we prove that every topological property can be char-
acterised in terms of convergence of nets. In order to do this it
suffices to obtain a characterisation of open sets because everything

17
in topology depends ultimately on open [Link] suffices to characterise
closure property in terms of convergence of nets since closed sets and
open sets are characterised in terms of closures. Here we begin with
the characterisation of Hausdorff property.

Characterisation of hausdorff property

Theorem 2.2.1 A topological space is Hausdorff if and only if limits

of all nets in it are unique.

Proof Suppose X is a Hausdorff space, S : D → X is a net in X and

S → x and S → y in X.
claim : x = y.
If x ̸= y, then there exists open sets U, V ∋ x ∈ U , y ∈ V and U ∩
V = ϕ. But by the definition of convergence,

S → x implies ∃ m1 ∈ D ∋ sn ∈ U ∀ n ≥ m1 and n ∈ D.
S → y implies ∃ m2 ∈ D ∋ sn ∈ U ∀ n ≥ m2 and n ∈ D.

m1 , m2 ∈ D then by property (iii) of directed set ∃ p ∈ D ∋ p ≥ m1

and p ≥ m2 .

⇒ Sn ∈ U ∩ V ∀ n ≥ p and n ∈ D.

Since U ∩ V = ϕ we are getting a contradiction.

Hence x = y
Conversely, suppose limits of all nets are unique.

18
Suppose X is not Hausdorff.
Then there exists two points x and y with the property that the open
sets of x and y are not disjoint.
Let ηx and ηy be the neighbourhood system of x and y respectively.
Let D = ηx × ηy and (U1 , V1 ) , (U2 , V2 ) ∈ D.
Define (U1 , V1 ) ≥ (U2 , V2 ) if U1 ⊆ U2 and V1 ⊆ V2 .
Clearly D is a directed set.
Define S : D → X ∋ S(U, V ) be any point in U ∩ V .
We will prove that S is converging to both x and y.
Let G be any open set of x. Then (G, X) ∈ D.
Now if (U, V ) ≥ (G, X) in D then U ⊆ G and so S(U, V ) ∈ U ∩
V ⊆ U ⊆ G.
Thus S converges to x. Similarly S converges to y.
Hence S converges to both x and y which is a contradiction to our
hypothesis.
So X is Hausdorff

Characterisation of limit point

Theorem 2.2.2 Let A be a subset of a space X and let x ∈ X. Then

x ∈ A if and only if there exists a net in A which converges to x in X.

19
 proof

Suppose S : D → A is a net which, when regarded as a net in X

converges to x.
Let U be any neighbourhood of X.
Since S → x in X ⇒ ∃n ∈ D m≥ n and m ∈ D ⇒ S(m) ∈ U.
But Sm ∈ A, m ∈ D. So A ∩U ̸= ϕ.
U is arbitrary, by the definition of limit point x∈ A.
Conversely suppose x∈ A. Then every neighbourhood of x meets A.
Let ηx be the neighbourhood system of x in X with the usual binary
relation.
Define S : D → A by S(U) =any point in U ∩ A. Then S is a net in
A.
claim : S → x in X.
Let G be any open set in X containing x.
For any U ∈ , U ≥ G ⇒ U ⊆ G.
Hence S(U) ∈ U ⊆ G.
So U ⊆ G ⇒ S(U) ∈ G ⇒ S → x in X.
Hence there exists a net in A which converges to x.

Corollary 2.2.1 A subset A of X is closed if and only if limits of

nets in A are in A

Corollary 2.2.2 A subset B of a space X is open if and only if no

20
net in the complement X/B converges to a point in X

Corollary 2.2.3 Let τ1 ,τ2 be two topologies on a set X such that a

net in X converges to a point with respect to τ1 if and only if it does
so with respect to τ2 . Then τ1 = τ2 .

Proof

Let B in τ1 we will prove that B ∈ τ2 .

By Corollary 2.2.2., B is open ⇒ no net in X / B converges to a
point in B with respect to τ1 .
Since convergence with respect to τ 1 is identical that with respect
to τ2
⇒ no net in X / B converges to a point in B in τ2 .
⇒ B is open by Corollary 2.2.2.
⇒ B ∈ τ2 .
So τ1 ⊆ τ2 .
Similarly τ2 ⊆ τ1 .
So we get τ1 = τ2 .

Characterisation of continuity of a function

Theorem 2.2.3: Let X, Y be topological spaces , x0 ∈ X and f :

X → Y a function. Then f is continuous at x0 if and only if when-
ever a net S converges to x0 in X, the net f ◦ S converges to f(x0 ) in Y.

21
 proof

Suppose f is a continuous at x0 and S : D → X is net converging to

x0 .
Let V be any neighbourhood of f(x0 ) in Y. Then f −1 (V ) is a neigh-
bourhood of x0 in X.
Since S → x0 , ∃ m ∈ D n ∈ D with n ≥ m ⇒ Sn ∈ f−1 (V ).
⇒ n ≥ m , f(Sn ) ∈ V .
Thus f ◦ S converges to f(x0 ) in Y.
Conversely suppose whenever a net S converges to x0 in X, the net f
◦S converges to f(x0 ) in Y.
Suppose f is not continuous at x0 .
Then there exists a neighbourhood V of f(x0 ) such that f− 1(V ) is
not a neighbourhood of x0 .
x0 ∈ f −1 (V ) is not a neighbourhood of x0 ⇒ every neighbourhood
of x0 intersects X / f −1 (V ).
Now ηx0 be the neighbourhood system of x0 with usual binary op-
eration.
Define S : ηx0 → X by S(N) = any point in N ∩ (X/f −1 (V )).
Then S converges to x0 in X. Then for any neighbourhood U of x0 ,
U ∈ ηx0 .
Then N ≥ U ⇒ N ⊆ U ⇒ S(N) ∈ N ⊆U
But f◦S takes values on Y/[Link] V is a neighbourhood of f(x0 ) which

22
contains no point of net f◦S.
Thus f◦S ↛ f(x0 ),a contradiction.
Hence f is continuous

Characterisation of compactness

Theorem 2.2.4: For a topological space X, the following state-

ments are equivalent:
(1) X is compact.
(2) Every net in X has a cluster point.
(3) every net in X has a convergent subnet in X.

proof

(2) ←→ (3) is proved previously. So we will prove that (1) ←→ (2).

Assume (1) and let S : D → X be a net in X.
Suppose S has no cluster point in X.
ie, No point x ∈ X is a cluster point of S.
ie, for each x, ∃ a neighbourhood of x, Nx and mx ∈ D n ∈ D, n ≥
mx ⇒ S(n) ∈ X / Nx .
Cover X by {Nx } x ∈X.
Since X is compact there exists x1 , x1 , ..., xk ∈ X X =∪ki=1 Nxi .
Let m1 , m2 , ..., mk be corresponding elements in D such that n ≥
mi
⇒ S(n) ∈ X / Nxi .

23
Since D is a directed set ∃ m ∈ D m ≥ mi i = 1, 2, ..., k.
Now by assumption S(m) ∈ ∩ki=1 (X/N xi) = X / ∪ki=1 N xi which is
not possible.
So S has atleast one cluster point in X.
Conversely assume (2) holds.
We will use Theorem 1.1.2. to prove X is compact.
Let C be a family of closed sets of X having finite intersection prop-
erty.
D be the family of all finite intersections of members of C.
Note that D itself closed under finite intersection property, and C ⊆
D.
Define D ≥ E for D, E ∈ D if D ⊆ E.
Thus D is a directed set.
If D, E ∈ D then D ∩ E ∈ D and D ∩ E ≥ D , D ∪ E ≥ E.
Clearly ϕ ̸= D , since C has finite intersection property
Define S : D → X , S(D) =any point in D.
By the hypothesis, S has a cluster point, say x in X.
T
Claim: x ∈ C∈C C
If not there exists C ∈ C such that x ∈ C. Then X / C is a neigh-
bourhood of x, since C is closed set belongs to C.
By the definition of cluster point, for C ∈ D there exists D ∈ D such
that D ≥ C and S(D) ∈ X / C.

24
D ≥ C ⇒ D ⊆ C so X / C ⊆ X / D.
But S(D) ∈ X / C ⊆ X / D = Dc , which is a contradiction since
S(D) is a point in D.
T
So x ∈ C∈C C.

25
 CHAPTER 3

Filters in topology
3.1 Filters and its convergence

Definition [Link]
A filter on (or in) a set X is a non-empty family F of non-empty
subsets of X suchthat:
(i) F is closed under finite intersections ,
(ii) if B ∈ F and B ⊆ A then A ∈ F for all A, B ⊆ X.
Clearly, X is an element of filter F always.

Example

(1)Fix some non-empty subset A of X. Then the collection of all

supersets of A (in X) is a filter on X. Such a filter is known as an
atomic filter, the set A being called the atom of the filter.
(2) In case X is infinite, the family F of all co-finite subsets of X is
a filter on [Link] a filter is called as a co-finite filter.
(3)Suppose τ is a topology on X. Then for any x ∈ X, the neighbour-
hood system ηx at x is a filter. It is called the τ -neighbourhood
filter at x. It depends on both x and τ .

26
 Definition [Link] of a filter
Let F be a filter on a set X. Then a sub-family B of F is said to
be a base of F (or a filter base) if for any A ∈ F there exists B ∈
B such that B ⊆ A.

Note: If B is a base of filter F then every member of F is a superset

of some members of B.

Proposition 3.1.1. Let B be a family of non-empty subsets of

a set X. Then there exists a filter on X having B as a base if and
only if B has the property that for any B1 , B2 ∈ B, there exists B3
∈ B such that B1 ∩ B2 ⊃ B3 .

proof

Suppose there exists a filter F on X having B as a base.

Then B ⊂ F and ϕ ∈
/ F since ϕ ∈
/B.
Also let B1 , B2 ∈.
Then B1 ∈ F (since B ⊆ F ) and so B1 ∩ B2 ∈ F as F is closed
under finite intersections.
So by the definition of a base , there exists B3 ∈ B1 ∩ B2 ⊇ B3 .
Conversely suppose B satisfies the given condition.
Let F be the family of all supersets of members of B. The condition
(ii) in the definition of a filter automatically holds for F .
The empty set cannot be supersets of any set.

27
 / F as ϕ ∈
Hence ϕ ∈ /B.
It only remains to show that F is closed under finite intersections.
So suppose A1 , A2 ∈ F .
Then there exists B1 , B2 ∈ B such that B1 ⊂ A1 and B2 ⊂ A2 .
By assumption there exists B3 ∈ B such that B3 ⊆ B1 ∩ B2 .
But A1 ∩ A2 is superset of B1 ∩ B2 hence it is a superset of B3 ∈
B.
So A1 ∩ A2 ∈ F .
Thus F is a filter on X and B is a base for it by the construction.

Corollary 3.1.1. Any family which does not contain the empty set
and which is closed under finite intersection is a base for a unique
filter.

Definition [Link]-base of a filter

Let F be a filter on a set X. Then a subfamily S of F is said to be
a sub-base for F if the family of all finite intersections of members
of S is a base for F . We also say S generates F .

Proposition 3.1.2. Let S be a family of subsets of a set X. Then

there exists a filter on X having S as a sub base if and only if S
has the finite intersection property.

proof

If there exists a filter F on X, containing S then F has finite inter-

28
section property and so does every subfamily of F .
Conversely suppose S has the finite intersection property.
Let B be the family of finite intersections of members of S .
/ B and B is closed under finite intersections.
Then ϕ ∈
So by the Corollary 3.1.1., B is a base for a filter F on X and thus
S is a sub base for F .

Definition [Link] of filter

A filter F converges to a point p if F ⊃ ηp . i.e, every neighbourhood
of p is contained in F .

Definition [Link] point

A point p is a limit point of a filter F if for every neighbourhood U
of p and every A ∈ F we have A ∩ U ̸= ϕ. p is also called as cluster
point of F .

Definition 3.1.6. Subfilter

Let F be filter on X. Then a filter G is said to be a subfilter of F
if F ⊆ G .

There is a canonical way of converting net to filters and vice versa.

Let S : D → X be a net. For each m ∈ D, let Bm = {(Sn ) : n ∈ D,
n ≥ m}. Let
F = {A ⊆ X : A ⊇ Bm , forsome m ∈ D}
In other words, F is the collection of all super sets of sets of the

29
form Bm for m ∈ D.
Using the fact that D is a directed set, we can show that F is a filter
on X.
This filter is called the filter associated with the net S.
Conversely let F be a filter on X. Let
D = {(x, F) ∈ X × F : x ∈ F}
For (x, F), (y, G) ∈ D define (x, F) ≥ (y, G) if F ⊆ G.
It is clear that D is a directed set because F is closed under finite
intersections.
Define S : D → X by S(x, F) = x.
Now this net is called the net associated with F .

Proposition 3.1.3. Let S : D → X be a net and F be the filter

associated with it. Let x ∈ X. Then S converges to x as a net if and
only if F converges to x as a filter. Also x is a cluster point of the
net S if and only if x is a cluster point of the filter F .

proof

Assume S converges to x.
Let U be any neighbourhood of x in X.
Then there exists m ∈ D such that Bm ∈ U, where Bm = {S(n) : n
∈ D, n ≥ m}.
But this means U ∈ F by the definition of F .

30
So ηx ⊂ F ,
i.e, F converges to x.
Conversely suppose F converges to x.
Let U be any neighbourhood of x.
Then U ∈ F . By the construction of F , there exists m ∈ D such
that Bm ⊂ U, where Bm is defined above.
⇒ S(n) ∈ U n ∈ D , n ≥ m.
Thus S → x in X.
Now for the second part , suppose x is a cluster point of the net S,U
be any neighbourhood of x.
Let F ∈ F .
Then we have to show U ∩ F ̸= ϕ.
F ∈ F ⇒ ∃ m ∈ D ∋ Bm ⊂ F.
U is a neighbourhood of x and m ∈ D
⇒ n ∈ D ∋ n ≥ m and S(n) ∈ U (by the definition of cluster point
of a net)
⇒ S(n) ∈ U ∩ F.
Since U and F are arbitrary, we get x is a cluster point of F .
Conversely, suppose x is a cluster point of F .
Let U be any neighbourhood of x and m0 ∈ D.
We have Bm m∈D where Bm = S(n) : n ≥ m is a base for F .
⇒ Bm ∈ F .

31
There fore Bm0 ∈ F .
Hence Bm0 U ̸= = ϕ,
x is a cluster point of F .
There fore ∃ n ≥ m0 ∋ S(n) ∈ U.
Since neighbourhood N of x and m0 ∈ D are arbitrary, we get x is a
cluster point of S.

Proposition 3.1.4. Let F be a filter in a space X and S be the

associated net in X. Let x ∈ X. Let F converges to x as a filter if
and only if S converges to x as a net. Moreover, x is a cluster point
of the filter F if and only if it is a cluster point of the net S

3.2 Characterisation of filters

Hausdorff property

Theorem 3.2.1: A topological space is a Hausdorff if and only if

no filter can converge to more than one point in it.

proof

Suppose X is a Hausdorff space and F is a filter on X converging to

x and y.
This means ηx ⊂ F and ηy ⊂ F , where ηx and ηy are the neigh-
bourhood system of x and y respectively.
Now if x ̸= y ∃ U ∈ ηx and V ∈ ηy ∋ U ∩ V = ϕ, which contradict

32
that F has finite intersection property.
So x = y.
Conversely assume that no filter in X converges to more than one
limit in X.
If X is not Hausdorff , there exists x, y ∈ X, x ̸= y such that every
neighbourhood of x intersects every neighbourhood of y.
∴ ηx ∪ ηy has finite intersection property,so there exists a filter F
on X containing ηx ∪ ηx .
Then clearly F converges to both x and y which contradict our hy-
pothesis.
So X is a hausdorff space.

Theorem 3.2.2: For a topological space X, the following statements

are equivalent:
(1) X is compact.
(2) Every filter on X has a cluster point.
(3) Every filter on X has a convergent subfilter.

Proof is followed from the equivalence condition of compactness us-

ing nets by Theorem 2.2.4.

Definition [Link] Filter

Suppose X,Y are topological spaces and f : X → Y is a function.
Then the image filter on Y is defined as the filter generated by the

33
base {f(A) : A ∈ F } . The image filter is denoted by f( F ).

Claim: B = {f(A) : A ∈ F } is a base for a filter.

/ f( F ).
Evidently ϕ ∈
Also for B1 , B2 ∈ B there exists A1 , A2 ∈ F ∋ f(A1 ) =B1 and f(A2 )
= B2 .
We know that A1 ∩ A2 ∈ F and B1 ∩ B2 ⊃ f(A1 ∩ A2 ).
Also f(A1 ∩ A2 ) ∈ B.
i.e, for any B1 , B2 ∈ B we are able to find B3 = f(A1 ∩ A2 ) ∈ B ∋
B3 ⊂ B1 ∩ B2 .
Therefore B is a base for a filter.

Continuity of a function

Proposition 3.2.1: Let X, Y be a topological space, x X, and f :

X → Y a function. Then f is continuous at x if and only if whenever
a filter F converges to x in X, the image filter f(F ) of F converges
to f(x) in Y.

proof

Assume first that f is continuous at x and F is a filter which con-

verges to x in X.
We have to show f(F ) converges to f(x) in Y.
Let N be any neighbourhood of f(x) in Y.
Then by continuity f 1 (N) is a neighbourhood of x in X.

34
By convergence of F to x f 1 (N) ∈ F .
So N ⊇ f 1 (N) ∈ f(F ) .
N is an arbitrary neighbourhood of f(x), implies f(F ) converges to
f(x).
Conversely suppose the condition about filter is satisfied.
We have to show that f is continuous at x. If not there exists a
neighbourhood N of f(x) such that f 1 (N)is not a neighbourhood of
x.
So every neighbourhood of x intersects X / f 1 (N) and hence the fam-
ily, S = ηx ∪ X / f 1 (N)has finite intersection property.
Therefore S generates a filter F and converges to x.
However f(F ) is not converging to f(x).
Since X / f 1 (N) ∈ F we get f(X /f 1 (N) ) ∈ f(F ) .
But Y / N contains f(X / f 1 (N)) Y / N ∈ f(F ).
Therefore N ∈
/ f(F ) [ no filter can contain both a set and its com-
pliment].
Thus f(F ) does not converge to f(x). This contradiction proves that
f is continuous at x.

Theorem 3.2.3: Let X be a topological product of an indexed fam-

ily of spaces {Xi : i ∈ I}. Let F be a filter on X and let x ∈ X. Then
F converges to a point x in X if and only if for each i ∈ I, the filter
πi (F) converges to πi (x) in Xi .

35
proof

The necessity condition holds using the continuity of πi and propo-

sition 3.2.1.
Conversely suppose that πi (F) converges to πi (x) in Xi for all i.
Claim: F → x in X.
Let N be any neighbourhood of x.
Then N contains basic open set V of x.
Let V = Πi∈I where Vi = Xi for all i except for some finite i1 , i2 , ...,
in .
Without loss of generality let it be 1,2,...,n.
Now πi (F) → πi (x) ∀ i = 1, 2, ..., n.
So ∃ Fi ∈ F ∋ Vi ⊃ πi (Fi ) for i = 1, 2, ..., n .
But πi−1 (Vi )⊃ πi−1 (πi (Fi ))⊃ i for i=1,2,.......n.
So N ⊃ V = ni=1 πi−1 (Vi )⊃ ni=1 Fi
T T

But ni=1 Fi is in F since F is closed under finite intersection.

T

N ∈ F ⇒ ηx ⊃ F .
∴ F converges to x.

3.3 Ultrafilters

Definition 3.3.1. Ultrafilter

An ultrafilter is a maximal filter on X. i.e, a filter F is an ultrafilter

on X if G is any other filter on X with F ⊂ G , then F = G .

36
Theorem 3.3.1. Let F be a filter on a set X,then there is an ul-
trafilter F on X with F ⊃ F .

proof

(a) Let {Fi }i∈I be a chain (i.e, totally ordered set of filters on X) of
collection of filters containing F with the inclusion ordering.
For the chain, take ∪i∈I Fi = G and F ⊂ Fi ∀i ⇒ F ⊆ G is a filter
which is an upper bound for the chain.
So there exist a maximal filterF which contains F by Zorn’s lemma,
as we required.

Proposition 3.3.2. For a filter F on X, the following statements

are equivalent:
(i) For every subset Y ⊂ X we have Y ∈ F or X / Y ∈ F .
(ii) F is an ultrafilter.

proof

(i) ⇒ (ii)
For all Y ⊂ X either Y ∈ F or X / Y ∈ F .
If there exist a larger filter then it contains both Y and X/Y which
is not possible as Y ∩ (X / Y ) = ϕ.
(ii) ⇒ (i)
Let Y ⊂ X. If there exist A1 , A2 ∈ F such that Y ∩ A1 = (X / Y )
∩ A2 = ϕ ⇒ A1 ∩ A2 = ϕ which is a contradiction.

37
So we must have
Y ∩ A = ϕ A ∈ F or (X / Y ) ∩ A ̸= ϕ ∀ A ∈ F .
If Y ∩ A ̸= ϕ ∀ A ∈ F , then by (part b) there exist a filter G con-
taining F ∪ {Y ,since F is an ultrafilter.
So we must have G = F .
If (X / Y ) ∩ A ̸= ϕ A ∈ F , similarly we will get G ⊇ F ∪ {X / Y
} ⇒ G = F.
G = F implies either Y ∈ G or X / Y ∈ G .

Proposition 3.3.3. Let F be an ultrafilter on X, let f : X → Y be

a function and let f(F ) = {Y ⊃ f(A): A ∈ F } is an ultrafilter of Y.

proof

We will use the characterisation of Proposition 3.3.2.

Let W ⊂ Y then f −1 (Y / W) = X / f −1 (W).
Since F is an ultrafilter either f −1 (W) ∈ F or f −1 (Y /W) ∈ F .

W ⊃ f( f −1 (W)) ∈ f(F ).

where as,

Y / W ⊃ f( f −1 (Y W)) ∈ f(F ).

So f(F ) is an ultrafilter on Y .

Proposition 3.3.4. An ultrafilter F converges to a point iff that

point is a limit point of it.

38
proof Suppose x is a limit point.

Then if U is a neighbourhood of x, U intersects A ∀ A ∈ F .

By part (b) of Proposition 3.1.2. there exist G containing F ∪ {U}
but F is maximal which implies that U ∈ F .
i.e, U ∈ F for every neighbourhood U of x implies ηx ⊆ F ⇒ F
converges to x.
Conversely suppose F converges to x.
Then any neighbourhood U ∈ F ⇒ U ∩ A ̸= ϕ ∀ A ∈ F ⇒ x is a
limit point of F .

Proposition 3.3.5. A topological space is compact iff every ultra-

filter in it is convergent.

proof Suppose X is compact.

Suppose F is an ultrafilter and it has no limit point.

Let x ∈ X and Ux be a neighbourhood of x.
We know either Ux or (Ux )c ∈ F (by the characterisation of ultrafil-
ter).
Since there exist A ∈ F such that A ∩ Ux = ϕ ⇒ (Ux )c ∈ F
{Ux }x∈X is a cover for X.
Hence it has finite subcover.
Sn
Let it be Ux1 , Ux2 , ..., Uxn such that i=1 Uxi = X.
Correspondingly there exist A1 , A2 , ..., An such that Ai ∩ Uxi = ϕ.

39
Clearly A = A1 ∩ A2 ∩ ... ∩ An ∈ F .
(Ux1 )c ∈ F such that (Ux1 )c = ni=2 Uxi ∈ F F .
S

Then,

A ∩ (Ux1 )c = A ∩ ( ni=2 Uxi ) = ϕ.

S

which is a contradiction.
Hence there exist a limit point for F ⇒ every ultrafilter is conver-
gent.
Conversely suppose every ultrafilter in it is convergent by proposi-
tion 3.3.3.
Suppose X is not compact.
Then there exist collection C of closed sets of X having finite inter-
section property and C∈C C = ϕ (⇒ {C c : C ∈ C } is an open cover
T

for X which has no finite sub-cover. )

Any collection of sets having finite intersection property is a sub-base
for a filter.
Hence ∃ a ultrafilter G containing C .
Let x be a limit point of G then there exists X / C ∋ x ∈ X / C (∴
{C c : C ∈ C } is an open cover).
As x is a limit point X/C intersect all elements of G but C ∈ C ⊆
G . So C ∩ X / C = ϕ which is not possible.
Therefore ∄ any limit point for G , which is a contradiction.
Hence X is Compact.

40
 CONCLUSION
The importance of nets and filters in general topology, especially
where compactness is involved, is established here. The nets were
introduced by E H Moore in the early 20th century while the idea
of filters and its applications are primarily developed by European
mathematicians, beginning with the work of H Cartan which is fol-
lowed by Bourbaki. Some topological properties which are not easy
to characterise using sequences are also characterised with the help
of nets and filters. However, the importance of convergence theory
of nets and filters extends beyond this project.

41
 BIBLIOGRAPHY
1. Paul R. Chernoff, A simple proof of Tychonoff’s theorem via nets,
The American Mathematical Monthly, Vol.99, 932-934 (Dec 1992)
2. Abdellatif Dasser, The use of filters in topology, University of
central Florida(2004)
3. K. D. Joshi, Introduction to General Topology, New age interna-
tional(P) limited, publishers,India(2006).
4. J. L. Kelley, General topology, Springer-Verlag, New York,1975.
5. Tommaso Russo, Introduction to Nets, [Link].
6. Rui Xiong, Nets and Filters in Topology, [Link], Feb(2017).

42
DON BOSCO ARTS & SCIENCE
COLLEGE ANGADIKADAVU

DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

PECULIAR FORMS OF NUMBERS AND

RELATED THEOREMS
 DON BOSCO ARTS & SCIENCE COLLEGE
ANGADIKADAVU
DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

PECULIAR FORMS OF NUMBERS AND

RELATED THEOREMS

Dissertation submitted in the partial

Fulfillment of the requirement for the award of
MSc Degree in Mathematics of
Kannur University
Name : ANAGHA C
Roll No. : C1PSMM1903

Examiners:
1.
2.
 KANNUR UNIVERSITY

BONAFIDE CERTIFICATE

Certified that this project report “PECULIAR FORMS OF NUM-

BERS AND RELATED THEOREMS” is the bonafide work of ANAGHA.
C who carried out the project work under my supervision.

Signature of Head Signature of Supervisor

 DECLARATION

I, ANAGHA.C hereby declare that the Project work entitled PE-

CULIAR FORMS OF NUMBERS AND RELATED THEOREMS
has been prepared by me and submitted to Kannur University in
partial fulfillment of requirement for the award of Master of Science
is a record of original work done by me under the supervision of Mrs.
NAJUMUNNISA K, Assistant Professor, Department of Mathemat-
ics, Don Bosco Arts & Science College, Angadikadavu. I, also declare
that this Project work has not been submitted by me fully for the
award of any Degree, Diploma, Title or recognition before any au-
thority.

Place: Angadikadavu ANAGHA.C

Date: (C1PSMM1903)
 ACKNOWLEDGEMENT

Introduction is the proper place to begin. But first I bow my head

before the Almighty who is always with me. Also I must express
my deepest gratitude to people along the way. No words can ad-
equately express the sense of gratitude, still I try to express my
heartfelt thanks through words. The outset, I am deeply indebted
to my project supervisor Mrs. NAJUMUNNISA K Assistant Pro-
fessor, Department of Mathematics, Don Bosco Arts and Science
College, Angadikadavu, for the invaluable guidance, loving encour-
agement and meticulous care towards me throughout my career. I
express my deep sense of gratitude to all the faculty members of the
Department of Mathematics, Don Bosco Arts and Science College,
Angadikadavu. I can never forget the support and encouragement
rendered by the Principal and the Staff of Don Bosco Arts & Sci-
ence College, Angadikadavu. I could not name many who sincerely
supported and helped for the successful completion of this Project.
It is my pleasure and duty to thank each and everyone of them who
walked with me.

ANAGHA.C
 CONTENTS

1. INTRODUCTION ........................................................... 01

2. PRELIMINARY RESULTS ....................................... 03

3. PERFECT NUMBERS ................................................ 08

4. MERSENNE PRIME ................................................... 18

5. FERMAT NUMBERS .................................................. 28

6. FIBONACCI .................................................................... 35

7. CONCLUSION ................................................................ 42

8. BIBLIOGRAPH ............................................................... 43
 INTRODUCTION
Number theory an old branch of mathematics is the science of
studying the properties of numbers. In accordance with research
methods and objective, we briefly divided Number theory into four
classes:Elementary Number theory, Analytic Number theory,Geometric
Number theory.

The theory of numbers has always occupied a unique position in the

World of mathematics. Number theory, by natural a discipline that
demanded a high standard of rigor. The aim of the elementary the-
ory of numbers is to investigate the properties of integers.

There are plenty of numbers with some interesting and magical prop-
erties in the history of the numbers. The numbers with some unique
and fascinating properties have always received a great attention in
the world of mathematics. There are so many kinds of numbers like
P erf ect number, M ersenne numbers , F ermat numbers ,f ibanocci
etc which shows some special characteristics. it was well said by
great mathematician Kronecker that ”God created the natural num-
bers,and all rest is the work of man”.

The second chapter deals with perfect numbers and related theo-
rems of perfect numbers. The third chapter consists of study on
M ersenne numbers with some theorem relating perf ect and mersenne
numbers. In fourth chapter we discus about F ermat numbers with
some [Link] in the last chapter we discus about a sequence

1
called f ibanocci and its terms are called f ibanocci numbers

The main aim of this project is to provide introduction to some

”Numbers of peculiar form” and some of its basic properties and
characteristics with some theorems.

2
 CHAPTER - 1
PRELIMINARY RESULTS

Definition 1.1

Let n be a fixed integer. Two integers a and b are said to be congruent

module n,

a ≡ b(modn)

If n divides the difference a − b; that is, provided that a − b = kn for

some integer k.

Example

(i) 15 ≡ 3(mod4)
(ii) 32 ≡ −1(mod5)

Theorem 1.2

Assume a ≡ b(modm) if d/m and d/a then d/b.

Definition 1.3

Given a positive integer n, let τ (n) denote the number of positive

divisors of n and σ(n) denoted the sum of these divisors.

Theorem 1.4

If n = pk11 · pk22 · · · pkr r is the prime factorization of n > 1, then

τ (a)=(k1 + 1)(k2 + 1) · · · (kr + 1)

3
 k +1 k +1
P1 1 −1 P2 2 −1 Prkr +1 −1
σ(n)= p1 −1 p2 −1 · · · pr −1

Definition 1.5

For a real or complex α and any integer n ≥ 1 we define

P
σα (n) = d/n

the sum of the divisors of n.

Definition 1.6

An arithmetic function f is called multiplicative if f is not identically

zero and if f (mn) = f (m)f (n) whenever (m, n) = 1

Theorem 1.7

F ERM AT ′ S theorem

Let p be a prime and suppose that p ̸ a then ap−1 ≡ 1(modp)

Theorem 1.8

If p is a prime and p ab then p a or p b.

Theorem 1.9
If p is a prime ,then

1 if p ≡ 1(mod8)orp ≡ 7(mod8)

(2 p) =
−1

if p ≡ 3(mod8)orp ≡ 5(mod8).

Theorem 1.10

Let the integer a have order k modulo n. Then ah ≡ 1(modn) if and

4
only if k h; in particular, k ϕ(n).

Theorem 1.11

Euler’s criterion : Let p be an odd prime and gcd(a, p) = 1. Then

a is a quadratic residue of p if and only if a(p−l)/2 ≡ 1(modp).

Definition 1.12

For n ≥ 1, let ϕ(n) denote the number of positive integers not ex-
ceeding n that are relatively prime to n.

Theorem 1.13

Let the integer a have order k modulo n. Then ah ≡ 1(modn) if and

only if k/h; in particular, k/ϕ(n).

Definition 1.14

An integer q is called a quadratic residue modulo n if it is congruent

to a perfect square modulo n.
ie, x2 ≡ q(modn)
Otherwise, q is called a quadratic non-residue modulo n.

Definition 1.15

Legendresymbol is a function of
1 if a is a quadratic residue modulo p and a ̸≡ 0 (mod p),


   
a 
= −1 if a is a quadratic non-residue modulo p,
p 



0 if a ≡ 0 (mod p).


Legendre’s original definition was by means of the explicit formula

5
   
a p−1 a
≡ a 2 (mod p) and ∈ {−1, 0, 1}.
p p

Definition 1.17
The Euclidean Algorithm for finding gcd(A, B) is as follows:

(1) If A = 0 then gcd(A, B) = B, since the gcd(0, B) = B, and we

can stop.

(2) If B = 0 then gcd(A, B) = A, since the gcd(A, 0) = A, and we

can stop.

(3) Write A in quotient remainder form (A = B · Q + R)

(4) Find gcd(B, R) using the Euclidean Algorithm since gcd(A, B) =

gcd(B, R)

Lemma :18

Suppose a and b are not both zero.

a) (a,b)=(b,a)

b) if a > 0 and a/b then (a, b) = a

c) if a ≡ c(mod b), then (a, b) = (c, b)

Example :

A = 78, B = 36

(a) A ̸= 0

(b) B ̸= 0

6
(c) Use long division to find that 78/36 = 2 with a remainder of 6.
We can write this as:

(d) 78 = 36 · 2 + 6

(e) Find gcd(36, 6), since gcd(78, 36) = gcd(36, 6)

Theorem 1.19

ar = (a − 1)(ar−1 − ar−2 · · · a + 1)

Theorem 1.20

if d/a and d/b then d/a − b

Theorem 1.21
ax −1
a+1 = a2x−1 − a2x−2 + · · · + a − 1

NOTATION

Pn : Perfect Numbers

Mn : Mersenne Numbers

Mp : Mersenne Prime Numbers

un : Fibonacci Numbers

Fn : Fermat Numbers

7
 CHAPTER - 2
PERFECT NUMBERS
The history of the theory of numbers abounds with famous conjec-
tures and open question. The present chapter focuses on some of
the intriguing conjectures associated with perfect numbers. A few of
these have been satisfactorily answered, but most remain unresolved;
all have stimulated the development of the subject as a whole.

The P ythagorean′ s considered it rather remarkable that the number

6is equal to the sum of its positive divisor, other than it self:

6=1+2+3

The next number after 6 having this feature is 28; for the positive
divisors of 28 are found to be1,2,4,7,14,and 28, and

28 =1+2+4+7+14

In line with their philosophy of attributing mystical qualities to num-

bers, the P ythagorean′ s called such numbers ”perfect”. We state
this precisely in definitions.

Definition 2.1

A positive integer nis said to be perfect if n is equal to the sum of

all its positive divisors, excluding n itself.
The sum of the positive divisors of an integer n ,each of them less

8
than n, is given by σ(n) - [Link], the condition ”n is perfect”
amounts to asking that σ(n) - n = n or equivalently, that

σ((n) = 2n

For example, we have

σ(6) = 1 + 2 + 3 + 6 = 2 · 6

and

σ(28) = 1 + 2 + 4 + 7 + 14 + 28 = 2 · 28

so that 6 an 28 are both perfect numbers.

For many centuries , philosophers were more concerned with the

mystical or religious significance of perfect numbers than with their
mathematical properties.

Only four perfect numbers were known to the ancient Greeks.

N icomachus in his Introductio Arithmeticae (circa 100 A.D )
lists

P1 = 6 P2 = 28 P3 = 496 P4 = 8128

He says that they are formed in an ”orderly” fashion,one among the

units, one among the tens, one among the hundreds, and one among
the thousands (this is, less than 10,000). Based on this meager
evidence, it was conjectured that

1. The nth perfect number Pn contains exactly n digits; and

9
 2. The even perfect numbers end , alternately, in 6 and 8.

Both assertions are wrong. There is no perfect number with 5 digits;

the next perfect number (first given correctly in an anonymous 15th
century manuscript) is

P5 = 33550336

While the final digit of P5 is 6, the succeeding perfect number

namely,

P6 = 8589869056

Also ends in 6 ,not 8 as conjectured. To salvage something positive

direction, we shall show later that the even perfect number do always
end in 6 or 8-but not necessarily alternately.
If nothing else, the magnitude of P6 should convince the reader of
the rarity of perfect number. It is not yet known whether there are
finitely many or infinitely many of them.
The problem of determining the general form of all perfect numbers
dates back almost t the beginning of mathematical time. It was
partially solved by Euclid of the elements he proved that if the sum

1
1+2+22 +23 +....+ 2k− =p

is a prime number , then 2k− 1p is a perfect number (of necessity

even ). For instance, 1+2+4=7 is a prime; hence, 4 · 7 = 28 is a
perfect number. Euclid’s argument makes use of the formula for the
sum of a geometric progression.

10
 1 + 2 + 22 + 23 + .... + 2k−1 = 2k − 1

which is found in various P ythagoreans texts. In this notation, the

result read as follow : If 2k − 1, is prime (k>1) , then
n=(2k−1 )(2k -1) is perfect number. About 2000 years after Euclid,
Euler took a decisive step in proving that all even numbers must
be of this type. We incorporate both these statements in our first
theorem.

Theorem - 2.2

If 2k -1 is prime (k>1), then n=(2k−1 )(2k -1) is perfect and every

even perfect number is of this form.

Proof :
Let 2k -1 =P , a prime, and consider the integer n=(2k−1 )P . Inas-
much as gcd(2k − 1, P ) = 1, the multiplicativity of σ ( as well as
Theorem 1.6) entails that

σ (n) = σ(2k−1 )P )

=(2k -1)(P+1)

=(2k -1)2k

=2n

making n a perfect number. For the converse, assume that n is an

even perfect number. We may write n as n=(2k−1 )m, where m is an
odd integer and k≥ 2 . lt follows from gcd(2k − 1, m) = 1 that

11
 σ (n) = σ ((2k−1 )m)

= σ (2k−1 ) σ (m)

= (2k − 1) σ (m)

whereas the requirement for a number to be perfect gives

σ (n) = 2n = 2k m

Together, these relations yield

2k m=2k − 1)σ(m)

which is simply to say that (2k − 1)/2k m. But 2k − 1 and 2k are

relatively prime , whence 2k − 1)/m ; say, m = 2k − 1M . Now the
result of substituting this value of m into the last-displayed equation
and canceling 2k − 1 is that σ(m) = 2k m . Because m and M are
both divisors of m (with M < m), we have

2k m = σ(m) ≥ m + M = 2k M

leading to σ(m) = m + M . The implication of this equality is that

m has only two positive divisors, to wit, M and m itself. It must be
that m is prime and M = 1; in other words, m = (2k −1)M = (2k −1)
is a prime number,completing the present proof.
Because the problem of finding even perfect numbers is reduced to
the search for primes of the form (2k − 1), a closer look at these
integers might be fruitful. One thing that can be proved is that if

12
2k − 1 is a prime number, then the exponent k must itself be prime.
More generally, we have the following lemma.

Lemma - 2.3
If ak − 1 is prime (a > 0, k ≥ 2), then a = 2 and k is also prime

Proof :
It can be verified without difficulty that

ak − 1 = (a − 1)(ak−1 + ak−2 + + a + 1)

where, in the present setting,

ak−1 + ak−2 + ... + a + 1 ≥ a+ 1 > 1

Because by hypothesis ak − 1 is prime, the other factor must be 1;

that is,a - 1 = 1 so that a= 2.
If k were composite, then we could write k = rs, with 1 < r and
1 < s. Thus,

ak − 1 = (ar )s − 1
(s−2)
=(ar − l)(ar(s−1) + ar + ... + ar + 1)

and each factor on the right is plainly greater than 1. But this
violates the primality of ak − 1 , so that by contradiction k must be
prime

For p = 2, 3, 5, 7, the values 3, 7, 31, 127 of 2P − 1 are primes, so

that are all perfect numbers.

13
 2(22 − 1) = 6

22 (23 − 1) = 28

24 (25 − 1) = 496

26 (27 − 1) = 8128

Many early writers erroneously believed that 2P − 1

is prime for every choice of the prime number p. But in 1536, Hudal-
richus Regius in a work entitled U triusque Arithmetices exhibits
the correct factorization

211 − 1 = 2047 = 23.89

If this seems a small accomplishment, it should be realized that his

calculations were in all likelihood carried out in Roman numerals,
with the aid of an abacus (not until the late 16th century did the
Arabic numeral system win complete ascendancy over the Roman
one). Regius also gave p = 13 as the next value of p for which the
expression 2P − 1 is a prime. From this, we obtain the fifth perfect
number.

212 (213 − 1) = 33550336

One of the difficulties in finding further perfect numbers was the

unavailability of tables of primes. In 1603, Pietro Cataldi, who is
remembered chiefly for his invention of the notation for continued
fractions, published a list of all primes less than 5150. By the direct
procedure of dividing by all primes not exceeding the square root

14
of a number, Cataldi determined that 217 − 1 was prime and, in
consequence, that

216 (217 − 1) = 8589869056

is the sixth perfect number. A question that immediately springs to

mind is whether there are infinitely many primes of the type 2P − 1,
with p a prime. If the answer were in the affirmative, then there
would exist an infinitude of (even) perfect numbers. Unfortunately,
this is another famous unresolved problem.
This appears to be as good a place as any at which to prove our
theorem on the final digits of even perfect numbers.

Theorem - 2.4
An even perfect number n ends in the digit 6 or 8; equivalently,
either n ≡ 6(mod10) or n ≡ 8(mod10).

Proof :
Being an even perfect number, n may be represented as
n = 2k−l (2k − 1), where 2k − 1 is a prime. According to the last
lemma, the exponent k must also be prime. Ifk = 2, thenn = 6, and
the asserted result holds. We may therefore confine our attention to
the case k > 2. The proof falls into two parts, according as k takes
the form 4m + 1 or 4m + 3
If k is of the form 4m + 1, then

n = 24m (24m+1 − 1)
= 28m+1 − 24m

15
 = 2.162m − 16m

A straightforward induction argument will make it clear that

16t ≡ 6(mod10) for any positive integer t. Utilizing this congruence,
we get

n ≡ 26 − 6 ≡ 6(mod10)

Now, in the case in which k = 4m + 3,

n =24m+2 (24m+3 − 1)
= 28m+5 - 24m+2
=28m .25 - 4.16m
=2.162m .16 - 4.16m
= 2.162m+1 − 4.16m

Falling back on the fact that 16t ≡ 6(mod10), we see that

n ≡ 26 − 46 ≡ −12 ≡ 8(mod10)

Consequently, every even perfect number has a last digit equal to 6

or to 8.
Hence the proof.

A little more argument establishes a sharper result, namely, that

any even perfect number n = 2k−1 (2k − 1) always ends in the digits
6 or 28. Because an integer is congruent modulo 100 to its last two
digits, it suffices to prove that, if k is of the form 4m + 3, then
n ≡ 28(mod100) . To see this, note that

16
 2k−1 = 24m+2 = 16m 4 ≡ 64 ≡ 4(mod10)

Moreover, for k > 2, we have 4/2k−1 , and therefore the number

formed by the last two digits of 2k−1 is divisible by 4. The situation
is this: The last digit of 2k−1 is 4, and 4 divides the last two digits.
Modulo 100, the various possibilities are

2k−1 ≡ 4, 24, 44, 64, or 84

But this implies that

2k − 1 = 2.2k−1 − 1 ≡ 7, 47, 87, 27, or 67 (mod100)

whence

n = 2k−1 (2k − 1)

≡ 47, 2447, 4487, 6427, or 84.67(mod100)

It is a modest exercise, which we bequeath to the reader, to verify

that each of the products on the right-hand side of the last congru-
ence is congruent to 28 modulo 100.

17
 CHAPTER - 3
MERSENNE PRIMES
It has become traditional to call numbers of the form

Mn = 2n − 1 n >1

M ersenne numbers after Father Marin Mersenne who made an in-

correct but provocative assertion concerning their primality.
Now let us consider some Mersenne numbers

M1 M2 M3 M4 M5 M6 M7 M8
1 3 7 15 31 63 127 255

Those Mersenne numbers that happen to be prime are said to be

M ersenne primes. By what we proved in above, the determination
of Mersenne primes Mn—and, in turn, of even perfect numbers—is
narrowed down to the case in which n is itself prime.
Now let us see the first 5 Mersenne primes numbers.

M2 M3 M5 M7 M13
3 7 31 127 8191

In the preface of his Cogitata Physica-Mathematica (1644), Mersenne

stated that Mp is prime for p = 2, 3, 5, 7, 13, 17, 19, 31, 67, 127,
257 and composite for all other primes
p < 257. It was obvious to other mathematicians that Mersenne
could not have tested for primality all the numbers he had an-

18
nounced; but neither could they. Euler verified (1772) that M31
was prime by examining all primes up to 46339 as possible divisors,
but M67 , M127 , and M257 were beyond his technique; in any event,
this yielded the eighth perfect number

230 (231 − 1) = 2305843008139952128

It was not until 1947, after tremendous labor caused by unreliable

desk calculators, that the examination of the prime or composite
character of MP for the 55 primes in the range p < 257 was com-
pleted. We know now that M ersenne made five mistakes. He erro-
neously concluded that M67 and M257 are prime and excluded M61 ,
M89 , and M107 from his predicted list of primes. It is rather astonish-
ing that over 300 years were required to set the good friar straight.
All the composite numbers Mn with n ≤ 257 have now been com-
pletely factored.

In the study of Mersenne numbers, we come upon a strange fact:

when each of the first four Mersenne primes (namely, 3, 7, 31, and
127) is substituted for n in the formula 2n − 1, a higher M ersenne
prime is obtained. Mathematicians had hoped that this procedure
would give rise to an infinite set of M ersenne primes; in other words,
the conjecture was that if the number Mn is prime, then MMn is also
a prime.
Alas, in 1953 a high-speed computer found the next possibility

MM13 = 2M13 − 1 = 28191 − 1

19
to be composite.

There are various methods for determining whether certain special

types of M ersenne numbers are prime or composite. One such test
is presented next.

Theorem - 3.1
If p and q = 2p + 1 are primes, then either q Mp or q Mp + 2, but
not both.

Proof :

With reference to Fermat’s theorem, we know that

2q−1 − 1 ≡ 0(modq)

and, factoring the left-hand side, that

(2(q−1)/2 − 1)(2(q−1)/2) + 1) = (2P − 1)(2P + 1)

≡ 0(modq)

What amounts to the same thing:

Mp (Mp + 2) ≡ 0(modq)

The stated conclusion now follows directly from Theorem 1.8. We

cannot have both q Mp and q Mp + 2, for then q 2, which is impos-
sible.
EXAMPLE 1

20
Let us consider p = 23 ,then q = 2p + 1 =47 is also a prime. so
that we may consider the case of M23 . The question reduces to one
of whether 47 M23 or, to put it differently, whether 223 ≡ 1(mod47).
Now, we have

223 = 23 220
= 23 (25 )4
≡ 23 (−15)4 (mod47)

Also we get

(−15)4 = (225)2 ≡ (−10)2 ≡ 6(mod47)

From these two congruence’s we get

223 ≡ 23 · 6 ≡ 1(mod47)

that is M23 .

EXAMPLE 2

Let p = 29 then q = 2 · 29 + 1 = 59

Mp = M29 = 229 − 1 = 53, 68, 70, 912 − 1

= 53, 68, 70, 911

here neither q Mp nor q Mp + 2 . It reasonable to ask: What

conditions on q will ensure that q MP ? The answer is to be found
in Theorem 3.2.

Theorem 3.2

21
If q = 2n + 1 is prime, then we have the following:
(a) q Mn , provided that q ≡ 1(mod8) or q ≡ 7(mod8).
(b) q Mn + 2, provided that q ≡ 3(mod8) or q ≡ 5(mod8).

Proof :

(a) To say that q Mn is equivalent to asserting that

2(q−l)/2 = 2n ≡ 1(modq)

In terms of the Legendre symbol, the latter condition becomes

the requirement that, (2/q) = 1. But according to Theorem 1.9
(2/q) = 1 when we have q = 1(mod8) or q = 7(mod8).
(b) Similerly we can say that q/Mn + 2 by

2(p−1)/2 = 2n ≡ −1(modq)

In terms of Legende symbols the requeired condition becomes

(2/q) = −1
Also we have (2/q) = −1 when we have q ≡ 3(mod8) or q ≡ 5(mod8)
from theorem 1.9.

Corollary :3.3

If p and q = 2p + 1 are both odd primes, with p ≡ 3(mod4), then

q/Mp .

Proof :

We know that an odd prime is either of the form 4k + 1 or 4k + 3.

If p = 4k+3 then q = 2p+1 = 2(4k+3)+1 = 8k+7 then buy above
theorem

22
q ≡ 7(mod8).
⇒ q/mp
If p = 4k + 1 then q = 2p + 1 = 2(4k + 1) + 1 = 8k + 3 then buy
above theorem q ≡ 3(mod8). ⇒ q/Mn + 2 then by using theorem
3.1 q ̸ |Mn

Theorem 3.4

If p is an odd prime, then any prime divisor of Mp is of the form

2kp + 1.

Proof :

We have,

Mp = 2p − 1

Let q be any prime devisor of Mp then,

2p ≡ 1(modq)

If k is the smallest positive integer that satisfies 2k ≡ 1(modq) then

by theorem 1.10 we get k/p.
The case k = 1 cannot arise; for this would imply that q/1, an
impossible situation. Therefore, because both k/p and k > 1, the
primality of p forces k = p.

In compliance with Fermat’s theorem, we have 2q − 1 ≡ 1(modq),

and therefore, thanks to Theorem 1.10 again, k/q − 1. Knowing that
k = p, the net result is p/q − 1. To be definite, let us put q − 1 = pt;
then q = pt + 1. The proof is completed by noting that if t were

23
an odd integer, then q would be even and a contradiction occurs.
Hence, we must have q = 2kp + 1 for some choice of k, which gives
q the required form.

Theorem 3.5

If p is an odd prime, then any prime divisor q of Mp is of the form

q ≡ ±1(mod8) .

Proof:

Suppose that q is a prime divisor of Mp , so that 2P ≡ 1(modq).

According to Theorem 3.4, q is of the form q = 2kp + 1 for some in-
teger k. Thus, using Euler′ s criterion, (2/q) ≡ 2(q−l)/2 ≡ 1(modq),
whence (2/q) = 1. Theorem 1.9 can now be brought into play again
to conclude that q ≡ ±1(mod8)

An algorithm frequently used for testing the primality of Mp is the

Lucas − Lehmer test. It relies on the inductively defined sequence

S1 = 4 Sk+1 = Sk2 − 2 k≥1

Thus , the sequence begins with the values 4 , 14 , 194 , 37634 , ·

The basic theorem, as perfected by Derrick Lehmer in 1930 from
the pioneering results of Lucas, is this : For p > 2 , Mp is prime if
and only if Sp−1 ≡ 0(modMp ). An equivalent formulation is that MP
is prime if and only if Sp−2 ≡ ±2(p+l)/2 (modMp ). A simple example
is provided by the M ersenne number M7 = 27 − 1 = 127.

Working modulo 127, the computation runs as follows:

24
 S1 S2 S3 S4 S5 S6
4 14 67 42 -16 0

This establishes that M7 is prime.

Theorem 3.6 : Euler

If n is an odd perfect number, then

2 2 2
n = pk11 · p2j2 · p3j3 · · · pr jr

where the Pi ’s are distinct odd primes and P 1 ≡ k1 ≡ 1(mod4).

Proof:

Let n = pk11 · pk22 · · · pkr r be the prime factorization of n. Because n is

perfect, we can write

2n = σ(n) = σ(pk11 ) · σ(pk22 ) . . . σ(pkr r )

Being an odd integer, either n ≡ 1(mod4) or n ≡ 3(mod4); in any

event, 2n ≡ 2(mod4). Thus, σ(n) = 2n is divisible by 2, but not by
4. The implication is that one of the σ(pki i ) , say σ(pk11 ), must be an
even integer (but not divisible by 4 ), and all the remaining σ(pki i )’s
are odd integers.

For a given Pi , there are two cases to be considered: Pi ≡ 1(mod4)

and Pi ≡ 3(mod4). If Pi ≡ 3equiv − 1(mod4), we would have

σ(pki i ) = 1 + Pi + p2i + · · · + pki i

≡ 1 + (−1) + (−1)2 + · · · + (−1)ki i (mod4)

25
 
0(mod4)

if ki is odd
≡

1(mod4) if ki is even

Because σ(pk11 ) ≡ 2(mod4), this tells us that P1 ̸≡ .3(mod4) or, to

put it affirmatively, p1 ≡ 1(mod4). Furthermore, the congruence
σ(piki ) ≡ 0(mod4) signifies that 4 divides σ(pki i ), which is not possi-
ble. The conclusion: if Pi ≡ 3(mod4), where i = 2, · · · , r, then its
exponent ki is an even integer.

Should it happen that Pi ≡ 1(mod4)—which is certainly true for

i = 1—then

σ(pki i ) = 1 + Pi + p2i + · · · + pki i

≡ 1 + 11 + 12 + + 1ki ; (mod4)

≡ ki + 1(mod4)

The condition σ(pk11 ) ≡ 2(mod4) forces k1 ≡ 1(mod4). For the

other values of i, we know that σ(pki i ) ≡ 1or3(mod4), and therefore
ki ≡ 0or2(mod4); in any case, ki is an even integer. The crucial
point is that, regardless of whether Pi ≡ 1(mod4) or Pi ≡ 3(mod4),
ki is always even for i ̸= 1. Our proof is now complete.
In view of the preceding theorem, any odd perfect number n can be
expressed as
2 2 2
n = pk11 · p2j2 · p3j3 · · · pr jr
2
= p1k1 (pj22 · pj33 · · · pjrr )

26
 = pk11 m2

This leads directly to the following corollary

Corollary

If n is an odd perfect number, then n is of the form

n = pk m2

where p is a prime, p ∤ m, and p ≡ k ≡ 1(mod4); in particular,

n ≡ 1(mod4).

Proof.

The last assertion is the only non-obvious one. Because p ≡ 1(mod4),

we have pk ≡ 1(mod4). Notice that m must be odd; hence,
m ≡ 1 or 3(mod4), and therefore upon squaring, m2 ≡ 1(mod4). It
follows that

n = pk m2 = 11 ≡ 1(mod4)

establishing our corollary.

27
 CHAPTER - 4
FERMAT NUMBERS
Let us mention another class of numbers that provides a rich source
of conjectures, the Fermat numbers. These may be considered as a
special case of the integers of the form 2m + 1. We observe that if
2m + 1 is an odd prime, then m = 2n for some n ≥ 0. Assume to the
contrary that m had an odd divisor 2k + 1 > 1, say m = (2k + 1)r
; then 2m + 1 would admit the nontrivial factorization

2m + 1 = 2(2k+1)r + 1 = (2r )2k+1 + 1

= (2r + 1)(22kr − 2(2k−1)r + ... + 22r − 2r + 1)

which is impossible. In brief, 2m + 1 can be prime only if m is a

power of 2.

Definition 4.1.
A Fermat number is an integer of the form
n
Fn = 22 + 1, n≥0

If Fn is prime, it is said to be a Fermat prime.

Let us see first 5 prime Fermat numbers,

F0 = 3 F1 = 5 F2 = 17 F3 = 257 F4 = 65537

Fermat announced that: ”I have found that numbers of the form

n
22 + 1 are always prime numbers and has long since signified to
analysts the truth of this theorem”.But he fails to prove [Link]
Euler resolved that Fermat’s assumption was wrong,by founding

28
 5
F5 = 22 + 1 = 424967297

to be divisible by [Link] consider the theorem that shows 641/F5

,

Theorem 4.2
The Fermat number F5 is divisible by 641.

Proof
We begin by putting a = 27 and b = 5, so that

1 + ab = 1 + 27 · 5 = 641

It is easily seen that

1 + ab − b4 = 1 + (a − b3 )b
= 1 + 640 − 625
= 1 + 15
= 1 + 3b = 24

But this implies that

5
F5 = 22 + 1 = 232 + 1

= 24 a4 + 1

= (1 + ab − b4 )a4 + 1

= (1 + ab)a4 + (1 − a4 b4 )

= (1 + ab)[a4 + (1 − ab)(1 + a2 b2 )]

which gives 641/Fn .

29
To this day it is not known whether there are infinitely many Fermat
primes or, for that matter, whether there is at least one Fermat prime
beyond F4 . The best guess is that all Fermat numbers Fn > F4
are composite.
A useful property of Fermat numbers is that they are relatively prime
to each other. Or ”No two Fermat numbers have a common divisor
grater than 1”.

Theorem 4.3
No two Fermat numbers have a common divisor grater than 1

Proof
suppose that Fn and Fn+k ,where k>0, are two Fermat numbers , and
that m/Fn , m/Fn+k
n
If x = 22 , we have

n+k
Fn+k − 2 22 + 1 − 2
=
Fn 22n + 1

xk −1
= x+1

= x2k−1 − x2k−2 + · · · − 1

and so Fn /Fn+k -2 . Hence m/Fn+k , it follows that m/(Fn+k − 2).

Therefore m/2 .Since Fn is odd, m=1 which proves the theorem.

It follows that each of the number F1 ,F2 ,F3 ,F4 · · · Fn is divisible

by an odd prime which does not divided any of the other. It shows

30
that there are infinitely many Fermat numbers, which leads to the
number of Fermat prime is also infinite.

Later Jesuit priest [Link] devised the practical test (Pepin’s test)
for determining the primality of Fn that is embodied in the following
theorem.

Theorem 4.4 Pepin’s test

n
For n 2 ≥ 1, the Fermat number Fn = 22 + 1 is prime if and only if

3(Fn −1)/2 ≡ −1(modFn )

Proof First let us assume that

3(Fn −1)/2 ≡ −1(modFn )

Upon squaring both sides, we get

3Fn −1 ≡ 1(modFn )

The same congruence holds for any prime p that divides Fn :

3Fn −1 ≡ 1(modp)

Now let k be the order of 3 modulo p. Theorem 1.13 indicates that

k Fn − 1, or in other words, that k 22” ; therefore k must be a
power of 2.

It is not possible that k = 2r for any r ≤ 2n − 1. If this were so,

repeated squaring of the congruence 3k ≡ 1(modp) would yield
2n −1
32 ≡ 1(modp)

31
or

3(Fn −1)/2 ≡ 1(modp)

We would then arrive at 1 ≡ −1(modp), resulting in p = 2, which is

a contradiction. Thus the only possibility open to us is that
n
k = 22 = Fn − 1

Fermat’s theorem tells us that k ≤ −1, which means, in turn, that

Fn = k + 1 ≤ p. Because p Fn , we also have p ≤ Fn .Together these
inequalities mean that Fn = p, so that Fn is a prime.

On the other hand, suppose that Fn , n ≥ 1, is prime. The Quadratic

Reciprocity Law gives

(3/Fn ) = (Fn /3) = (2/3) = −1

when we use the fact that Fn = (−1)2n + 1 = 2(mod3). Applying

Euler’s Criterion,we end up with

3(Fn −1)/2 ≡ −1(modFn )

Let us demonstrate the primality of F3 = 257 using Pepin’s test.

Working modulo 257, we have
n
Fn = 22 + 1
3
F3 = 22 + 1
F3 = 256 + 1
F3 = 257
3(F3 −l)/2 = 3256/2

32
 = 3128
= 33 (35 )25
≡ 27(−14)25
≡ 27.1424 (−14)
≡ 27(17)(−14)
≡ 2719 ≡ 513 ≡ −1(mod257)

so that F3 is prime.

Theorem 4.5
n
Any prime divisor p of the Fermat number Fn = 22 +1, where n ≥ 2,
is of the form p = k · 2n+2 + 1.
Proof

For a prime divisor p of Fn ,

4
22 ≡ −1(modp)

which is to say, upon squaring, that

n+1
22 ≡ 1(modp)

If h is the order of 2 modulo p, this congruence tells us that

h 2n+1

We cannot have h = 2r where 1 ≤ r ≤ n, for this would lead to

n
22 ≡ 1(modp)

and, in tum, to the contradiction that p = 2. This lets us conclude

that h = 2n+1 .Because the order of 2 modulo p divides ϕ(p) = p − 1,

33
we may further conclude that 2n+1 p − 1. The point is that for
n ≥ 2, p ≡ 1(mod8), and therefore, by Theorem 1.9 , the Legendre
symbol (2/p) = 1. Using Euler’s criterion, we immediately pass to

2(p−1)/2 ≡ (2/p) = 1(modp)

An appeal to Theorem 1.10 finishes the proof. It asserts that h (p −

1)/2, or equivalently, 2n + 1 (p − 1)/2. This forces 2n+2 p − 1, and
we obtain p = k · 2n+2 + 1 for some integer k.

A resume of the current primality status for the Fermat numbers Fn ,

where 0 ≤ n ≤ 35, is given below.

n Character of Fn
0, 1, 2, 3, 4 prime
5, 6, 7, 8, 9, 10, 11 completely factored
12, 13, 15, 16, 18, 19,25,27,30 two or more prime factors known
17,21,23,26,28,29,31,32 only one prime factor known
14,20,22,24 composite, but no factor known
33,34 ,35 character unknown

The case for F16 was settled in 1953 and lays to rest the tantalizing
conjecture that all the terms of the sequence
2 22 222
2 + 1, 22 + 1, 22 + 1, 22 + 1 22 + 1, · · ·

are prime numbers.

34
 CHAPTER - 5
FIBONACCI
The Fibonacci numbers may be defined by the recurrence relation
u0 = 0, u1 = 1,
and
un = un−1 + un−2 for n > 1.
1, 1,2,3,5,8, 13,21,34,55,89, 144,233,377, ...
is called the Fibonacci sequence and its terms the Fibonacci numbers.

u0 u1 u2 u3 u4 u5 u6 u7 u8 u9 u10 u11 u12 u13

0 1 1 2 3 5 8 13 21 34 55 89 144 233

Theorem:5.1

For the Fibonacci sequence, gcd(un , un+1 ) = 1 for every n ≥ 1.

Proof

Let us suppose that the integer d > 1 divides both un and un+1 ·
Then their difference un+1 − un = un−1 is also divisible by d. From
this and from the relation un − un−1 = un − 2, it may be concluded
that d un−2 · Working backward continuously like this we can say
that d un−3 ,d un−4 , ... , and finally that d u1 . But u1 = 1, which
is certainly not divisible by any d > 1. This contradiction ends our
proof.

We can see that u3 = 2, u5 = 5, u7 = 13, and u11 = 89 are all prime

numbers,which might lead us to guess that un is prime whenever the

35
subscript n > 2 is a prime.

But the term u19 shows that the assumption is false. Since the term

u19 = 4181 = 37 · 113

As we know, the greatest common divisor of two positive integers can

be found from the Euclidean Algorithm after finitely many divisions.
we know ”Given n > 0, there exist positive integers a and b such that
to calculate gcd(a, b) by means of the Euclidean Algorithm exactly
n divisions are needed”.

To verify this ,let a = un+2 and b = un+l · The Euclidean Algorithm

for obtaining gcd(un+2 , un+1 ) leads to the system of equations;

un+2 = 1 · un+l + un
un+l = 1 · un + un−1
·
·
·
u4 = 1 · u3 + u2
u3 = 2 · u2 + 0

the number of divisions necessary here is n. Also we get;

gcd(un+2 , un+1 ) = u2 = 1

which confirms anew that successive Fibonacci numbers are rela-

tively prime.
For example consider; n = 6. The following calculations show that

36
we need 6 divisions to find the greatest common divisor of the inte-
gers u8 = 21 and u7 = 13:

21 = 1 · 13 + 8
13 = 1.8 + 5
8=1·5+3
5=1·3+2
3=1·2+1
2=2·1+0

Gabriel Lame observed in 1844 that if n division steps are required

in the Euclidean Algorithm to compute gcd(a, b), wherea > b > 0,
then a ≥ un+2 , b ≥ un+l · Consequently, it was common at one time
to call the sequence un the Lame sequence.

Also we can see that one of the features of the Fibonacci sequence is
that the greatest common divisor of two Fibonacci numbers is itself
a Fibonacci [Link] equation

um+n = um−1 un + um un+1

bring out this fact.

Now let us prove the equation holds for all m ≥ 2 and n ≥ 1.

For this fix m ≥ 2 , this identity (equation ) is established by induc-

tion on n.
Let n = 1, then the equation becomes

um+1 = um−1 u1 + um u2 = um−1 + um

37
it is obviously true. Let us assume that it is true for n = 1, 2, ..., k.
Now we have to show that the equation holds, when n = k + 1.
The equation holds for n=k, implies

um+k = um−1 uk + um uk+1

The equation holds for n=k-1, implies

um+(k−1) = um−1 uk−1 + um uk

Now add this two equation we get;

um+(k−1) + um+k = um−1 uk−1 + um uk + um−1 uk + um uk+1

⇒ um+(k−1) + um+k = um−1 (uk + uk−1 ) + um (uk+1 + uk )

By the definition of e Fibonacci numbers,we get;

um+(k+1) = um−1 uk+1 + um uk+2

which is of the form of the required equation ,that is the equation

holds for n = k + [Link] the induction step is [Link] the
equation holds for all m ≥ 2 and n ≥ 1.

Theorem 5.2 :

For m ≥ 1, n ≥ 1umn is divisible by um .

Proof :

We argue by induction on n, the result being certainly true when

n = 1. For our induction hypothesis, let us assume that umn is
divisible by um for n = 1, 2, ..., k. The transition to the case um(k+l) =
umk+m is realized using the equation;

38
 um+n = um−1 un + um un+1

Because um divides umk by supposition, the right-hand side of this

expression must be divisible by um · Accordingly, um /um(k+l) . That
is umn is divisible by um This is what we have to prove.

Lemma:5.3

If m = qn + r, then gcd(um , un ) = gcd(ur , un ).

Proof:

From the equation

um+n = um−1 un + um un+1

we can wright

gcd(um , un ) = gcd(uqn+r , un )
= gcd(uqn−1 ur + uqn ur+1 , un )

From theorem 5.2 we can have umn is divisible by um and also we

know the fact that
gcd(a + c, b) = gcd(a, b), whenever b/c,
From theorem 5.2 we can say uqn is divisible by un that is uqn ur+1
is divisible by un and now use the fact gcd(a + c, b) = gcd(a, b),
whenever b/c and write

gcd(uqn−1 ur + uqn ur+1 , un ) = gcd(uqn−1 ur , un )

Claim : gcd(uqn−1 , un ) = 1
Let d = gcd(uqn−1 , un ).The relations d/un and un /uqn imply that

39
d/uqn ,and therefore d is a positive common divisor of the succes-
sive Fibonacci numbers uqn−1 and uqn · Because successive Fibonacci
numbers are relatively prime, the effect of this is that d = 1.

We have gcd(a, be) = gcd(a, b) when gcd(a, c) = 1

gcd(um , un ) = gcd(uqn−1 ur , un ) = gcd(ur , un )

hence the lemma.

Theorem:5.4

The greatest common divisor of two Fibonacci numbers is again a

Fibonacci number; specifically,

gcd(um , un ) = ud where d = gcd(m, n)

Proof:

Assume that m ≥ n Applying the Euclidean Algorithm torn and n,

we get the following system of equations:

m = q1 · n + r1 0 < r1 < n
n = q1 · r1 + r2 0 < r2 < r1
r1 = q3 · r2 + r3 0 < r3 < r2
·
·
·
rn−2 = qn · rn−1 + rn 0 < rn < rn−1
rn−1 = qn+1 · rn + 0

From the previous lemma,

40
 gcd(um , un ) = gcd(ur1 , un ) = gcd(ur1 , ur2 ) = · · · = gcd(urn−1 , urn )

Because rn /rn−1 , Theorem 14.2 tells us that urn /urn−1 whence gcd(urn−1 , urn ) =
urn . But rn , being the last nonzero remainder in the Euclidean Al-
gorithm for rn and n, is equal to gcd(m, n).We get,

gcd(um , un ) = ugcd(m,n)
= ud

Hence the theorem.

Corollary:5.5

In the Fibonacci sequence, um /un if and only if m/n for n ≥ m ≥ 3

Proof:

A good illustration of Theorem 5.4 is provided by calculating gcd(u16 , u12 ) =

gcd(987, 144). From the Euclidean Algorithm,

987 = 6. 144 + 123

144 = 1 . 123 + 21
123 = 5 . 21 + 18
2 1= 1 . 18 + 3
18 = 6. 3 + 0

and therefore gcd(987, 144) = 3. The net result is that

gcd(u16 , u12 ) = 3 = u4 = ugcd(l6,12)

as asserted by Theorem 5.4

41
 CONCLUSION

The purpose of the project is to give a simple account of some spacial

form of numbers and their [Link] project gives the more de-
tails about perf ect numbers, mersenne prime, f ermat, nummber,
and f ibanocci numbers

In first section we consider a number form called P erf ect number,

and its [Link] second section we consider a number form called
mersenne number, and its properties. In third section we consider
a number form called f ermat number, and its [Link] fourth
section we consider a series called f ibonacci series, and its proper-
ties.
It is a section in the subject called NUMBER THEORY also some
time known as “HIGHER MATHEMATICS”.

42
 BIBLIOGRAPHY

1. David .[Link] Elementary Number theory (6th edition )

Tata mcgraw-HILLEDITION 2010

2. David .[Link] Elementary Number theory (7th edition )

Tata mcgraw-HILLEDITION 2012

3. George .E Andrews, Number theory.

4. [Link] and E.M. Wright An Introduction to the the-

ory of Numbers (6th edition)

43
DON BOSCO ARTS & SCIENCE
COLLEGE ANGADIKADAVU

DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

TOPOLOGICAL VECTOR SPACE

 DON BOSCO ARTS & SCIENCE COLLEGE
ANGADIKADAVU
DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

TOPOLOGICAL VECTOR SPACE

Dissertation submitted in the partial

Fulfillment of the requirement for the award of
MSc Degree in Mathematics of
Kannur University
Name : PRANAV K P
Roll No. : C1PSMM1901

Examiners:
1.
2.
 KANNUR UNIVERSITY

BONAFIDE CERTIFICATE

Certified that this project report “ TOPOLOGICAL VECTOR

SPACE” is the bonafide work of PRANAV K P who carried out the
project work under my supervision.

Signature of Head Signature of Supervisor

 DECLARATION

I, PRANAV K P hereby declare that the Project work entitled

TOPOLOGICAL VECTOR SPACE has been prepared by me and
submitted to Kannur University in partial fulfillment of requirement
for the award of Master of Science is a record of original work done
by me under the supervision of Mrs. AJEENA JOSEPH, Assistant
Professor, Department of Mathematics, Don Bosco Arts & Science
College, Angadikadavu. I, also declare that this Project work has not
been submitted by me fully for the award of any Degree, Diploma,
Title or recognition before any authority.

Place: Angadikadavu PRANAV K P

Date: (C1PSMM1901)
 ACKNOWLEDGEMENT

Introduction is the proper place to begin. But first I bow my head

before the Almighty who is always with me. Also I must express
my deepest gratitude to people along the way. No words can ad-
equately express the sense of gratitude, still I try to express my
heartfelt thanks through words. The outset, I am deeply indebted
to my project supervisor Mrs. AJEENA JOSEPH Assistant Pro-
fessor, Department of Mathematics, Don Bosco Arts and Science
College, Angadikadavu, for the invaluable guidance, loving encour-
agement and meticulous care towards me throughout my career. I
express my deep sense of gratitude to all the faculty members of the
Department of Mathematics, Don Bosco Arts and Science College,
Angadikadavu. I can never forget the support and encouragement
rendered by the Principal and the Staff of Don Bosco Arts & Sci-
ence College, Angadikadavu. I could not name many who sincerely
supported and helped for the successful completion of this Project.
It is my pleasure and duty to thank each and everyone of them who
walked with me.

PRANAV K P
 CONTENTS

1. INTRODUCTION ................................................................ 01

2. PRELIMINARIES .................................................................03

3. VECTOR TOPOLOGY AND LOCAL BASE ...................... 08

4. NORMABLE SPACES ........................................................ 23

5. APPLICATIONS .................................................................. 38

6. BIBLIOGRAPHY................................................................ 40
 INTRODUCTION

Functional Analysis plays an increasing role in applied science as

well as in Mathematics itself. Consequently, it becomes more and
more desirable to introduce the student to the field at an early stage
of study. Functional Analysis is the study of certain Topological-
Algebraic structures and of these structures and of the methods by
which knowledge of these structures can be applied to analytic prob-
lems. A good introduction on this subject should include a presen-
tation of its axiomatic. That is of the general theory of Topological
vector spaces. Also, it contains some interesting applications to the
branches of Mathematics.

In Mathematics, if a set is endowed with Algebraic and Topological

structures, then it always fascinating to prove relationship between
these to structures. A Topological Vector Space (also called Linear
Topological Space) is a basic structure on Topology on which a vec-
tor space X over a field F. As the name suggests the space blends a
topological structures with the algebraic concept of a vector space.
The elements of topological vector spaces are typically functions or
linear operators acting on topological vector spaces, and the topology
is often defined so as to capture a particular notion of convergence of
sequences of functions. We know that Normed space (linear space)
means a vector space together with a norm on it. Also we know that
every normed space is a metric space. Hilbert space and Banach
spaces are well known examples for this. Also, there is no discus-
sion of uniform spaces, of completeness occurs only in the context of
metric spaces. One can study vector spaces over the field of real or
complex numbers along with a suitable topology. They are known

1
as Topological vector spaces.

This topic Topological vector spaces give a minimum introduction to

topological vector spaces and it deal with some of the basic theories
of topological vector spaces. Notion of vector topology, bounded-
ness, compactness, local convexity are introduced. In this present
dissertation our aim is to generalize the concept of normed space to
topological vector space so that normed space is a special case of
topological vector spaces.

2
 PRELIMINARIES
Definition 0.1(Vector spaces): A vector space over K (where
K = R or C) is a set X, whose elements are called vectors, and in
which two operations, vector addition and scalar multiplication, are
defined with following algebraic properties:

1. For every pair of vectors x and y corresponds a vector x + y,in

such a way that x + y = y + x and x + (y + z) = (x + y) + z.

2. X contains the unique vector zero such that x+0 = x for every
x ∈ X.

3. For x ∈ X, there corresponds to a unique vector −x in X such

that x + (−x) = 0.

4. To every pair (α, x) with α ∈ K and x ∈ X corresponds a

vector αx, in such a way that

1 × x = x,

α(βx) = (αβ)x,

and such that the two distributive laws

α(x + y) = αx + αy

(α + β)x = αx + βx

hold.

3
Definition 0.2.(Topological space): A Topological space is a pair
(X, τ ) where X is a set and τ is a family of subsets of X satisfying:

1.ϕ ∈ τ and X ∈ τ .

2.τ is closed under arbitrary unions.

3.τ is closed under finite intersections.

The family τ is said to be a topology on the set. Members of τ are

said to be open sets in X.

Definition 0.3. (Normed spaces): Let X be a linear space over

K. A norm on X is a function || · || from X × X to R such that for
all x, y ∈ X and k ∈ K:

1. ||x|| ≥ 0 and ||x|| = 0 if and only if x = 0

2. ||x + y|| ≤ ||x|| + ||y||

3. ||kx|| = |k| · ||x||

A normed space is a linear space with a norm on it.

Definition 0.4. (Discrete and indiscrete topological space):

Let X be a set and let τ be the collection of all subsets of X. Then
τ is the topology on X and is known as the discrete topology.
Also, the set U = {ϕ, X} is a topology on X called the indiscrete
topology.
Definition 0.5. (Banach Spaces): A complete normed space is
called a Banach space.

Definition 0.6. (Homeomorphism): A Homeomorphism from a

space X to a space Y is a bijective function f from X to Y such

4
that both f and f −1 are continuous.
Definition 0.7. (Open set): A subset A ⊆ X is said to be open
if for every x ∈ A, there exists some open ball around x which is
contained in A.

Definition 0.8. (Local base): Let X be a space, x ∈ X. Then

a local base at x is a collection L of neighbourhoods of x such that
given any neighbourhood N of x, there exists L ∈ L satisfying L ⊂
N.

Definition 0.9. (Continuous function): Let (X, τ ) and (Y, U)

be topological spaces. A function f : X → Y is continuous at a point
x in X provided that for each neighbourhood V of f (x) there is a
neighbourhood U of x such that f (U ) ⊂ V . A function f : X → Y
is continuous provided it is continuous at each point of x.

Definition 0.10. (Compact subset): A subset A of a space X

is said to be a compact subset of X if for every cover of A by open
subsets of X has finite sub cover. A space X is said to be compact
if X is a compact subset of itself.

Definition 0.11. (Directed set): A directed set is a pair (D, ≥)

where D is a non-empty set and ≥ a binary relation on D satisfying:

1. For all m, n, p ∈ D, if m ≥ n and n ≥ p, then m ≥ p

2. For all n ∈ D, n ≥ n

3. For all m, n, p ∈ D, there exists p ∈ D such that p ≥ m and

p ≥ n.

Definition 0.12. (Net): A net in a set X is a function f : D → X

where D is a directed set.

5
Definition 0.13. (Filter): A filter on a set X is a non-empty
family F of subsets of X such that:

1. ϕ ∈
/F

2. F is closed under finite intersections

3. If B ∈ F and B ⊂ A, then A ∈ F for all subsets A, B of X.

Definition 0.14. (Ultra filter): A filter F on a set X is said to be

an ultra filter if it is a maximal element in the collection of all filters
on X, partially ordered by inclusion, that is, F is an ultra-filter if it
is not properly contained in any filter on X.

Definition 0.15. (Metrizable): A topological space X is said to

be metrizable if its topology is generated by a metric.

Definition 0.16. (Filter base): Let F be a filter on a set. Then

a sub-family B of F is said to be a base for F (filter base) if for any
A ∈ F there exists B ∈ B such that B ∈ A.

Definition 0.17. (Locally compact): A topological space X is

said to be locally compact at a point x ∈ X if x has a compact neigh-
bourhood in X. X is called locally compact if it is locally compact
at every point.

Definition 0.18. (Hausdorff space): A space X is said to be

a Hausdorff space if for every distinct points x, y ∈ X there exist
disjoint open sets U, V in X such that x ∈ U and y ∈ V .

Theorem 0.19. (The Hahn-Banach Theorem): If M is a linear

subspace of a normed linear space X and if f is a bounded linear
functional on M , then f0 can be extended to a bounded linear func-
tional f defined on the whole space such that ||f || = ||f0 ||.

6
Definition 0.20. (Tottaly bounded): Let A be a subset of a
metric space (X, d). Let ϵ > 0. A finite set {x1 , x2 , ..., xn } of points
of X is said to form a finite ϵ-net if A ⊆ Bϵ (x1 ) ∪ ... ∪ Bϵ (xn ). The
set A is totally bounded if A has a finite ϵ-net for every ϵ > 0.

7
 CHAPTER 1

VECTOR TOPOLOGY AND

LOCAL BASE
Definition 1.1 A topological vector space X is a vector space with
a topology τ under which the mappings

• (x, y) → x + y of X × X into X

• (α, x) → αx of K × X into X, K = C or R

are continuous. This topology τ is known as linear or vector topology.

Example 1.2. Let X ̸= {0} be a vector space and let τ be the

indiscrete topology in X. That is, τ = {ϕ, X}. Then with this
topology, X is a topological vector space.

Proof. To prove that ‘+’ is continuous at (x, y), recall that f is

continuous at x if and only if for every neighbourhood V of f (x)
there exists a neighbourhood U of x such that f (U ) ⊆ V .
The only neighbourhood of +(x, y) = x + y is X. Also, X × X
is a neighbourhood of (x, y) and +(X × X) = X and ·(α, x) =
αx, ·(F × X) = X. We get addition and scalar multiplication as
continuous at (x,y).

Example 1.3. A normed linear space is a topological vector space.

Proof. To prove this, we have a normed linear space is a vector space

with a norm. Every normed space may be regarded as a metric space
in which the distance d(x, y) between x and y is ∥x − y∥.

8
 Next, to verify that the algebraic operations addition and scalar
multiplications are continuous, we have to show that given ϵ > 0,
there is a δ > 0 such that

∥(x + y) − (a + b)∥ < ϵ, ∀x, y ∈ X

with ∥x − a∥ < δ and ∥y − b∥ < δ.

Consider

∥x + y − (a + b)∥ = ∥(x − a) + (y − b)∥

≤ ∥x − a∥ + ∥y − b∥

≤δ+δ

= 2δ
Take δ = ϵ/2, we get ∥x + y − (a + b)∥ < ϵ. Therefore, vector
addition is continuous.
Now, to prove scalar multiplication is continuous, we have the
identity

λx − αa = λx − αa + λa − λa
= λ(x − a) + a(λ − α).

∥λx − αa∥ = ∥λ(x − a) + a(λ − α)∥

≤ ∥λ(x − a)∥ + ∥a(λ − α)∥

≤ |λ|∥x − a∥ + |λ − α|∥a∥
ϵ ϵ
Hence, ∥λx − αa∥ < ϵ, whenever ∥x − a∥ < 2|λ| and |λ − α| < 2∥a∥ .
Hence, the result.
Here, a normed linear space satisfies the conditions of a topolog-
ical vector space.

9
Example 1.4 The real line with discrete topology is not a topolog-
ical vector space.

Proof. To show that the real line R with discrete topology is

not a topological vector space, we will show that there is a point
(λ, x) ∈ F × X at which the function ’·’ is not continuous.
Let λx = z ̸= [Link] Uz is a neighborhood of z.
Since Uλ = {(λ − n1 , λ + n1 ) : n = 1, 2, . . .} is a basis basis of the
neighborhoods of λ, there exists n ∈ N such that

1 1
(λ − , λ + ){x} ⊂ Uλ Ux ⊂ Uz = {z} = {λx}
n n
1 1
⇒ (λ − , λ + ){x} ⊆ {λx}
n n
This is not possible. Since interval ̸⊂ singleton set

Definition 1.5. A topological vector space is said to be Hausdorff

or separated if x ̸= 0 =⇒ 0 ∈ U and x ∈
/ U for some U .

Example 1.6 The empty space is a Hausdorff space.

Example 1.7 Any discrete space is a Hausdorff space. That is, a

topological space with the discrete topology.

Example 1.8. Any metric space is Hausdorff in the induced topol-

ogy. That is, any metrizable space is Hausdorff.

Proposition 1.9. Let X be a topological vector space. Then for

any x ∈ X, the mapping fx : X → X given by fx (y) = x + y is a
homeomorphism.

Proof Suppose that X is a topological vector space and for any

10
x ∈ X, fx : X → X is defined by fx (y) = x + y.
To prove that fx is a homeomorphism
suppose fx (y) = fx (z)
=⇒ x + y = x + z
=⇒ y = z

(1.9.1) Therefore,fx is injective.

Now to show that fx is surjective.

Gen y ∈ X, there is a y − x ∈ X, with
fx (y − x) = x + y − x = y.

(1.9.2) Therefore, fx is surjective.

It remains to show that f (x) is continuous.
Consider the function g : X → X × X defined by g(x) = (x, y).
Given U × {y} = W , a neighborhood of g(x), we have a neighbor-
hood of U of x with

g(U ) = U × {y} ⊂ W

Hence, g is continuous.
By the definition of a topological vector space, the function h :
X × X → X defined by h[(x, y)] = x + y.
Then h is continuous.
Therefore, the composition h ◦ g is continuous.

But, h ◦ g(x) = h(g(x)) = h[(x, y)] = x + y = fx (y), i.e., h ◦ g(x) =

fx (y).

(1.9.3) Therefore, fx is continuous.

11
Similarly, we can show that fx−1 : X → X defined by

(1.9.4) fx−1 (y) = y − x is continuous .

From equations (1.9.1), (1.9.2), (1.9.3), and (1.9.4), it follows that

fx is a homeomorphism.

Definition 1.10 Let X be a vector space over the field K, where

K = R or C. Then a non-empty set A in X is said to be convex if

λx + (1 − λ)y ∈ A
whenever x, y ∈ A and 0 ≤ λ ≤ 1, or equivalently A is convex if
x, y ∈ A, λ ≥ 0, µ ≥ 0, and λ + µ = 1. Then λx + µy ∈ A.

Example 1.11. In a normed linear space X, the closed unit ball

B = {x ∈ X : |x| ≤ 1} is convex.
Proof. For if x, y ∈ B and 0 ≤ λ ≤ 1, then ∥x∥ ≤ 1 and ∥y∥ ≤ 1.
Hence,

∥λx + (1 − λ)y∥ ≤ λ∥x∥ + (1 − λ)∥y∥ ≤ λ + (1 − λ)

Therefore, λx + (1 − λ)y ∈ B and B is convex.

Definition 1.12. A non-empty set A of a vector space X is said

to be balanced or critical if x ∈ A, |λ| ≤ 1 imply λx ∈ A. That is,
λA ⊂ A for all λ with |λ| ≤ 1.

Example 1.13. In Rn , the closed unit ball B(0, 1) = {x ∈ Rn :

|x| ≤ 1} is a balanced set

Proof Let x ∈ B(0, 1), and |λ| ≤ 1.

Now x ∈ B(0, 1) =⇒ ∥x∥ ≤ 1,then

12
∥λx∥ = |λ|∥x∥ ≤ 1

=⇒ λx ∈ B(0, 1).
This show that B(0,1) is balanced.

Definition 1.14. A nonemty set A of a vector space X is said to

be absorbing or absorbent if for each x ∈ X,there exist some ρ > 0
such that λx ∈ A for all |λ| ≤ ρ

Examble 1.15. In Rn ,the closed unit ball B(0,1) = {x ∈ Rn : ∥x∥ ≤

1} is an absorbent set

Proof. Let x ∈ B(0, 1) =⇒ ∥x∥ ≤ 1 and |λ| ≤ 1.

Choose ρ = 1 >0 such that |λ| ≤ ρ.
That is,λx ∈ B(0, 1) for all |λ| ≤ ρ.
i.e., for each x ∈ X, there exist some ρ > 0 such that λx ∈ B(0, 1)
∀|λ| ≤ ρ.

Hence B(0,1) is absorbent.

Remark 1.16. An absorbing set contains 0.

Defintion 1.17. A non-empty set A of a vector space X is said to

be absolutely convex if

λA + µA ⊂ A

whenever |λ| + |µ| ≤ 1.

Definition 1.18. Let V be an absorbing subset of a vector space

X. For each x ∈ X let

pV (x) = inf {λ > 0 : x ∈ λV }.

Then pV is called guage or Minkowski functional of V.

13
Theorem 1.19. A set A is absolutely convex if and only if A is
both balanced and convex.

Proof. Suppose that A is absolutely convex.

i.e., if λA + µA ⊂ A whenever |λ| + |µ| ≤ 1.
If λ ≥ 0, µ ≥ 0 and λ + µ = 1 and x, y ∈ A,then we have

(1.19.1) (λx + µy) ∈ A.

Since A is absolutely convex, clearly A is convex.

Next we want to show that A is balanced.
For this take µ = 0 in (1.19.1), we get

x ∈ A, |λ| ≤ 1 =⇒ λx ∈ A.

Hence A is balanced.
Conversely suppose that A is convex and balanced.
We want to show that A is absolutely convex.
Let x, y ∈ A and |λ| + |µ| ≤ 1.

• Case 1. If λ = 0 or µ = 0
then λx + µy ∈ A, because A is balanced.

• Case 2. If λ ̸= 0 and µ ̸= 0

λ λ
then |λ| x ∈ A, because |λ| = 1 and A is balanced.
 
µ
similarly we get |µ| y ∈ A.
Also,

|λ| |µ| |λ|+|µ|

|λ|+|µ| + |λ|+|µ| = |λ|+|µ| =1

But A is convex. Hence

   
|λ| λ |µ| µ
(1.19.2) |λ|+|µ| |λ| x + |λ|+|µ| |µ| y ∈ A

14
    
λ µ |λ| |µ|
[Since |λ| x ∈ A, |µ| y ∈ A,and |λ|+|µ| + |λ|+|µ| = 1,A is con-
vex.]
λx µy
Therefore by equation (1.19.2) we get, |λ|+|µ| + |λ|+|µ| ∈A
Now A is balanced and |λ| + |µ| ≤ 1 we have
 
λx µy
(|λ| + |µ|) |λ|+|µ| + |λ|+|µ| ∈ A

=⇒ λx + µy ∈ A

Therefore A is absolutely convex.

Theorem 1.20. Let X be topological vector space and let U be

the neighbourhood of 0. Then λU is a neighbourhood of 0 for every
non-zero real number λ.

Proof. Given that λ ̸= 0

Consider the mapping f : X → X defined by f (x) = λx.
Then f is continuous.
[since by the definition of a topological vector space scalar multipli-
cation is continuous.]
Also, inverse of f is, f −1 : X → X which is given by f −1 (x) = ( λ1 )x
is also continuous.
Furthermore, f is a bijection.
Hence f is a homeomorphism.
Thus f is an open map.
[since we have the result that f is open if and only if f is a homeo-
morphism and continuous.]
Since U is a neighbourhood of 0, there exists an open neighbourhood
G of 0 such that G ⊂ U .
Since f is an open map, f (G) is an open neighbourhood of 0, such
that
f (G) ⊂ λU [ Sincef (G) = λG ⊂ λU =⇒ f (G) ⊂ λU .]

15
So, λU is a neighbourhood of 0.

Theorem 1.21. Every neighbourhood of 0 in a topological vector

space X is absorbing.

Proof. Let U be any neighbourhood of 0. Fix x ∈ X. Consider the

map f on the scalar field into X defined by, f (λ) = λx.
This map f is continuous.
Hence there exist ρ > 0 such that |λ| < ρ =⇒ f (λ) ∈ U .
That is |λ| < ρ =⇒ λx ∈ U .
Hence U is absorbing.
That is, every neighbourhood of 0 in a topological vector space X is
absorbing.

Theorem 1.22. Every neighbourhood of 0 in a topological vector

space X includes a balanced neighbourhood of 0.

Proof. Let U be any neighbourhood of 0 in X. Fix x ∈ X. Consider

the function f : K × X → X defined by, f (λ, x) = λx.
Then f is continuous, by the definition of a topological vector space.
Hence there exists a neighbourhood V of (0, 0) such that

(1.22.1) f (V ) ⊂ U

Take V = V1 × V2 where V1 is a neighbourhood of ‘0 ’ in K and V2 is

a neighbourhood of ‘0 ’ in X.
But then for some r > 0, V1 contains the closed ball S̄(0, r) = {µ :
|µ| ≤ r}
Now, for this fixed r, let
W = ∪{µV2 : |µ| ≤ r}.
Then rV2 ⊂ W and rV2 is a neighbourhood of 0 in X, by the previous
theorem(1.20).Consequently, W is a neighbourhood of 0 in X.

16
Also, by (1.22.1), W ⊂ U .
Finally, we shall show that W is balanced.
Let x ∈ W and 0 < |λ| ≤ 1.

x ∈ W =⇒ x ∈ µV2 with |µ| ≤ r

=⇒ λx ∈ λµV2 with |λµ| ≤ r

=⇒ λx ∈ W .

Hence W is balanced.
i.e., every neighbourhood of 0 in a topological vector space X in-
cludes a balanced neighbourhood of 0.

Theorem 1.23. Let X be a topological vector space and A ⊂ X.

Then

Ā = ∩{A + U : U is a neighbourhood of 0}.

In particular,Ā ⊂ A + U for every neighbourhood U of 0.

Proof. First let x ∈ Ā.

Let U be any neighbourhood of 0. We have by the above theorem
“Every neighbourhood of 0 in a topological vector space X includes
a balanced neighbourhood of 0”, we may assume that U is balanced.
Then x + U is a neighbourhood of x.
Since x ∈ Ā, we have,

(x + U ) ∩ A ̸= ϕ

[we have the result that “Let A be a subset of a topological space,

and let x ∈ X. Then x ∈ Ā if and only if every neighbourhood of x
has a nonempty intersection with A” ].
And there exists x + u ∈ A for some u in U . Then, x ∈ A − U .
Since U is balanced, we have A − U = A + U .

17
Hence, x ∈ A + U .

Therefore x ∈ ∩{A + U : U is a neighborhood of 0}

Thus

(1.23.1) Ā ⊂ ∩{A + U }.

To prove, ∩{A + U : U is a neighbourhood of 0}∈ Ā.

For this, let x ∈ ∩{A + U : U is a neighbourhood of 0}.
Assume that x ∈
/ Ā.
Then there exists a balanced neighbourhood U of 0 such that (x +
U) ∩ A = ϕ
Hence, x ∈
/ A − U = A + U,
and thus x ∈
/ ∩{A + U : U is a neighbourhood of 0},
a contradiction to our assumption.
Thus x ∈ Ā. Arbitrariness of x gives

(1.23.2) ∩{A + U : U is a neighbourhood of 0}⊂ Ā.

Thus from equations (1.23.1) and (1.23.2) we get,

Ā = ∩{A + U : U is a neighbourhood of 0}.

Theorem 1.24. Let X be a topological vector space. Then the

1. Closure of a balanced set is balanced.

2. Closure of a convex set is convex.

3. Interior of a convex set is convex.

Proof. 1. Assume that X is a topological vector space and let A be

a balanced set.
We have to prove that Ā is balanced.

18
For this, let x ∈ Ā. Then for any neighbourhood U of 0, we have U
is balanced and x + U is a neighbourhood of x. As x ∈ Ā

(x + U ) ∩ A ̸= ϕ

Hence, there exists x0 ∈ (x + U ) ∩ A. Hence

λx0 ∈ λ[(x + U ) ∩ A], ∀|λ| ≤ 1.

Since U and A are balanced, we have

λ[(x + U ) ∩ A] ⊂ (λx + U ) ∩ A.

Hence,

λx0 ∈ (λx + U ) ∩ A,

for all λ with |λ| ≤ 1.

Hence (λx + U ) ∩ A ̸= ϕ and λx ∈ Ā for all λ with |λ| ≤ 1.
Therefore, Ā is balanced.
Hence (1) is proved.

2. Let A be a convex set. We have to prove that Ā is convex.

Let x, y ∈ Ā. Now given any neighbourhood U of 0, there exists a
balanced neighbourhood V of 0 such that

V + V ⊂ U.

Now

x ∈ Ā =⇒ (x + V ) ∩ A ̸= ϕ

y ∈ Ā =⇒ (y + V ) ∩ A ̸= ϕ

And hence there exists x0 ∈ (x + V ) ∩ A and there exists y0 ∈

(y + V ) ∩ A
Let λ be such that 0 ≤ λ ≤ 1.
Now, since A is convex, we have

19
λx0 + (1 − λ)y0 ∈ (λx + V ) ∩ λA + [(1 − λ)y + V ] ∩ (1 − λ)A

⊂ A ∩ [λx + (1 − λ)y + U ]
Therefore,

λx + (1 − λ)y + U ∩ A ̸= ϕ.

Hence,

λx + (1 − λ)y ∈ Ā

This proves that Ā is convex.

3. Let A be any convex set. We have to prove that interior of A is

convex. For this, let x, y ∈ int(A). Fix λ with 0 ≤ λ ≤ 1. Let

z = λx + (1 − λ)y

Let U and V be neighbourhoods of x and y respectively in A. Write

W = λU + (1 − λ)V

Then

W = ∪{λu + (1 − λ)V : u ∈ U }

is a neighbourhood of z. Also, W ∈ A. Hence z ∈ Int(A).

Therefore, Int(A) is also convex.

Theorem 1.25. Every neighbourhood of 0 in a topological vector

space X includes a closed balanced neighbourhood of 0.

Proof. We know that the function

f : X × X → X defined by, f [(x, y)] = x + y, is continuous

20
In particular, f is continuous at 0 = (0, 0) [since f is continuous at
each point]
Hence, given a neighbourhood U of 0, there exists a neighbourhood
N of (0, 0) with f (N ) ⊂ U .
[ since we have by definition of continuity, “Let X and Y be two
topological spaces, a function f from X to Y is said to be contin-
uous at x ∈ X if for each neighbourhood U of f (x) there exist a
neighbourhood V of x such that f (V ) ⊂ U ”. ]
Take W = V1 × V2 Where V1 and V2 are neighbourhoods of 0.
Let

V = V1 ∩ V2 .

Then

V + V ⊂ V1 + V2 = f (v1 , V2 ) ⊂ f (N ) ⊂ U .

Thus,

(1.25.1) V + V ⊂ U.

theorem 1.22, we can choose a balanced neighbourhood W of 0 such

that

(1.25.2) W ⊂V.

Then by theorem 1.23,

W ⊂V +W ⊂V +V

and hence, W ⊂ V + V , by (1.25.2).

That is, W ⊂ U by using equation (1.25.1).
That W is balanced follows from part (1) of theorem 1.24
Also we know that W is closed.
That is, W is a closed balanced neighbourhood of 0.

21
That is, Every neighbourhood of 0 in a topological vector space X
includes a closed and balanced neighbourhood of 0.

22
 CHAPTER 2

NORMABLE SPACES
Defintion 2.1. Let X be a vector space. A Semi-norm on X is a
function p : X → R satisfying

1. p(x) ≥ 0

2. p(λx) = |λ|p(x)

3. p(x + y) ≤ p(x) + p(y) ∀x, y ∈ X and λ ∈ F

A semi-norm is a norm if it satisfies the further condition:

4. p(x) = 0 ⇐⇒ x = 0

From (2) it follows that p(0) = 0.

Example 2.2. Let X be a vector space and let f be a linear func-

tional on X. Define

p(x) = |f (x)| ∀x ∈ X.

Then p is a semi-norm on X.

Proof. We have, p(x) = |f (x)| ∀x ∈ X.

Clearly, p(x) ≥ 0, because |f (x)| ≥ 0. Also,

p(λx) = |f (λx)|

= |λf (x)| , since f is linear

= |λ||f (x)|

23
 = |λ|p(x).

and

p(x + y) = |f (x + y)|

= |f (x) + f (y)|

≤ |f (x)| + |f (y)|

= p(x) + p(y).

Hence p is a semi-norm on X.

Example 2.3. A norm on a vector space X is always a semi-norm.

Proof. Let X be a vector space. We have norm on X is a real-valued

function which satisfying the following conditions

1. ∥x∥ ≥ 0∀x ∈ X

2. ∥x∥ = 0 ⇐⇒ x = 0, the zero element in X

3. ∥x + y∥ ≤ ∥x∥ + ∥y∥, ∀x, y ∈ X

4. ∥αx∥ = |α|∥x∥, ∀x ∈ X and for all scalars α

By the definition norm it is clear that it satisfies the conditions of

semi-norm. i.e., Norm on a vector space is always a semi-norm.

Proposition 2.4. If p is a semi-norm on X, then

|p(x) − p(y)| ≤ p(x − y) for all x, y in X

Proof. Assume that X be vector space and p is a semi-norm on X.

We want to show that |p(x) − p(y)| ≤ p(x − y) for all x, y ∈ X.
For that, we have, x = y + (x − y) and so

24
 p(x) = p[y + (x − y)]

≤ p(y) + p(x − y)

That is,

(2.4.1) p(x) ≤ p(y) + p(x − y)

[ by the definition of semi-norm p(x + y) ≤ p(x) + p(y) ]

By symmetry

p(y) ≤ p(x) + p(y − x)

= p(x) + p[(−1)(x − y)]

= p(x) + | − 1|p(x − y)

That is,

(2.4.2) p(y) ≤ p(x) + p(x − y)

From equations (2.4.1) and (2.4.2) we get

p(x) − p(y) ≤ p(x − y) and p(y) − p(x) ≤ p(x − y)

=⇒ p(x) − p(y) ≤ p(x − y) and −(p(x) − p(y)) ≤ p(x − y)

=⇒ p(x) − p(y) ≤ p(x − y) and −p(x − y) ≤ p(x) − p(y)

=⇒ −p(x − y) ≤ p(x) − p(y) ≤ p(x − y)

=⇒ |p(x) − p(y)| ≤ p(x − y)

Since x and y arbitrary, we get if p is a semi-norm on X, then

|p(x) − p(y)| ≤ p(x − y) for all x, y ∈ X.

25
Definition 2.5. A set in a topological vector space X is said to
be bounded if it is absorbed by every neighbourhood of 0. In other
words, a set A is bounded if and only if for every neighbourhood U
of 0 there exist ρ > 0 such that λA ⊂ U whenever |λ| < ρ.

Example 2.6. Every singleton set in a topological vector space is

a bounded set.

Theorem 2.7. Let A be a set in a topological vector space X. Then

the following assertions are equivalent:

1. A is bounded.

2. For every sequence (xn ) in A and for every sequence (ϵn ) of

scalars with ϵn → 0 it is true that ϵn xn → 0 as n → ∞.

3. For every sequence (xn ) in A, it is true that ( n1 )xn → 0 as

n → ∞.

Proof. Let X be a topological vector space and A be a set in X.

Step 1 : Suppose that A is bounded. We want to show that, for
every sequence (xn ) in A and for every sequence (ϵn ) of scalars with
ϵn → 0 it is true that ϵn xn → 0 as n → ∞.
Let (xn ) be a sequence in A. Let U be any neighbourhood of 0. Since
A is bounded there exist ρ > 0 such that λA ⊂ U whenever |λ| < ρ.
Let ϵn → 0. But then |ϵn | < ρ [Link]

ϵn xn ∈ ϵn A ⊂ U

(Since A is bounded) eventually. Hence

ϵn xn → 0 as n → ∞

(Since U is a neighbourhood of 0)
Since (xn ) is arbitrary, for every sequence (xn ) in A and for every

26
sequence (ϵn ) of scalars with ϵn → 0 it is true that ϵn xn → 0 as
n → ∞.

i.e., (1) =⇒ (2)

Step 2 : Suppose that for every sequence (xn ) in A and for every
sequence (ϵn ) of scalars with ϵn → 0 it is true that ϵn xn → 0 as
n → ∞.
We want to prove that for every sequence (xn ) in A, it is true that
( n1 )xn → 0 as n → ∞.
Let (xn ) be a sequence in A, and take ϵn = n1 , where n = 1, 2, ...
So that ϵn → 0.
Then from (2), we have ϵn xn → 0.
i.e.,

( n1 )xn → 0 as n → ∞

Thus (2) =⇒ (3)

Step 3 : Suppose that every sequence (xn ) in A, it is true that

( n1 )xn → 0 as n → ∞.We want to show that A is bounded.
Assume that A is not bounded. Then there exist a balanced neigh-
bourhood U of 0 such that λA ̸⊂ U for some λ with |λ| < ρ where
ρ > 0 is arbitrary. Now, U is balanced, and so

λA ̸⊂ U for all ρ > 0.

1
Take ρ = n so that we have
1
nA ̸⊂ U .

Choose

xn ∈ A − nU for n = 1, 2, ...

Then

27
 1
n xn ∈
/ U for n = 1, 2, ....

Hence n1 xn does not converges to zero.

Therefore (3) =⇒ (1)

Definition 2.8. Let X be a topological vector space. A subset A of

X is called totally bounded, if for each neighbourhood U of 0 there
is a finite subset F of X such that A ⊂ F + U .

Theorem 2.9. 1. Every compact set is totally bounded.

2. Every totally bounded set is bounded.

Proof. 1. Let A be any compact set. Let U be any open neighbour-

hood of 0.
Then

{a + U : a ∈ A}

is an open cover of A. Since A is compact we have every open cover

of A has a finite subcover. Therefore there exist a finite number of
points a1 , a2 , ..., an in A such that

A ⊂ ∪{ai + U }

This shows that A is totally bounded.

2. Let A be any totally bounded set. We want to show that A

is bounded. Let U be any balanced neighbourhood of 0. Pick a
balanced neighbourhood V of 0 such that

(2.9.1) V +V ⊂U

Since A is totally bounded, there is a finite subset F of X such that

28
(2.9.2) A⊂F +V

But F is bounded. Hence

(2.9 .3) F ⊂ nV for some n ≥ 1

From equations (2.9.2) and (2.9.3) we get

A ⊂ nV + V ⊂ nV + nV ⊂ n(V + V ),
That is A ⊂ n(V + V ) ⊂ nU by equation (2.9.1). Thus A ⊂ nU
which implies
1
tA ⊂ U for |t| < n

Hence A is bounded.

Theorem 2.10. A subset of a topological vector space X is compact

if and only if it is totally bounded and complete.

Proof. Let X be a topological vector space and suppose that A is

a compact subset of X.
We want to show that A is totally bounded and complete.
We have by theorem 2.9 “ Every compact set is totally bounded ”
Therefore A is Totally bounded.
Now we want to show that A is complete.
For that it is enough to show that every cauchy sequence in A is
converges to some element in A.
Let [xα : α ∈ D] be any cauchy set in A. Let
Tz = {xα : α ≥ z} for each α ∈ D

But D is a directed set. So, Tz = {xα : α ≥ z} has Finite Intersection

Property. But then,

29
 ∩{T z : z ∈ D} =
̸ ϕ,

by compactness of A. Consequently, there exist an element a ∈ A

such that

a ∈ ∩{Tz : z ∈ D}.

We shall show that {xα } converges to a.

Let U ∈ B. Choose V ∈ B such that V + V ⊂ U .
Since {xα } is Cauchy, there exist α0 such that α ≥ α0 , α′ ≥ α0
implies

(2.10.1) xα − xα0 ∈ V

But a ∈ Tz ∀z ∈ D.
Thus, (a + V ) ∩ Tz ̸= ϕ and accordingly, there exist

(2.10.2) xα ∈ a + V ∀α ≥ z.

Hence for α′ ≥ z, we have from equations (2.10.1.) and (2.10.2)

xα′ − a = xα′ − xα + xα − a ∈ V + V ⊂ U .

Thus {xα } converges to a. Thus A is complete.

Conversely suppose that A is totally bounded and complete.
We want to show that A is compact.
Let C be any collection of closed sets in A with the finite interaction
property
Then there exists an ultra filter F on A with F ⊃ C. But then
F is Cauchy, because A is totally bounded, since K is complete.
F → x ∈ A.
Let S ∈ C, let V be a neighbourhood of 0. Then,

x+V ⊃B

30
for some B ∈ F because F is a filter base.
But, S ∈ F .
Hence B ∩ S ̸= ϕ and hence

(x + V ) ∩ S ̸= ϕ.

Thus, x ∈ S. But S is closed in A and therefore, x ∈ S.

That is, ∩C ̸= ϕ.
This proves that A is compact.

Lemma 2.11. Let A be a convex set. If α ≥ 0, β ≥ 0, then

(α + β)A = αA + βA

Proof. Let A be a convex set and suppose that α ≥ 0, β ≥ 0. We

want to show that (α + β)A = αA + βA.

• Case 1: If α = 0, β = 0.

Then the relation (α + β)A = 0 = αA + βA =⇒ (α + β)A =

αA + βA

• Case 2: Let α > 0, β > 0.

If z ∈ A then

αz + βz ∈ αA + βA.

But

αz + βz = (α + β)z.

Therefore,

(α + β)z ∈ αA + βA, ∀z ∈ A,

31
 that is,

(2.11.1) (α + β)A ⊂ αA + βA

If x, y ∈ A

(2.11.2) αx + βy = (α + β)(α1 x + β1 y),

α β
where α1 = α+β and β1 = α+β .

But α1 > 0, β1 > 0 and α1 + β1 = 1. Thus, since A is convex,

it follows that

(2.11.3) α1 x + β1 y ∈ A

From equations (2.11.2) and (2.11.3) we have

αx + βy ∈ (α + β)A

hence

(2.11.4) αA + βA ⊂ (α + β)A

From equations (2.11.1) and (2.11.4) we observe that

αA + βA = (α + β)A.

Definition 2.12. A topological vector space is said to be normable

if its topology is given by a norm.

Example 2.13. Finite dimensional Hausdorff spaces are normable.

32
Theorem 2.14. (Kolmogoroff ’s Criterion for Normability)
Let X be a topological vector space. Then X is normable if and only
if X is Hausdorff and X contains a bounded convex open neighbour-
hood of 0.

Proof. Suppose that X is normable.

Let V = {x ∈ X : ∥x∥ < 1}.
Then V is a non-empty open subset of X.
Let G be any neighbourhood of 0. Then W ⊂ G, where for some
r≥0

W = {x ∈ X : ∥x∥ < r}.

Also
1
W = {x ∈ X : ∥x∥ ≤ r} =⇒ r {x ∈ X : ∥x∥ ≤ r} = V

i.e., V = 1r W ,

and 1r W ⊂ 1r G
i.e., V ⊂ 1r G
Hence V is a bounded set in X.
Next we would show that V is convex.
Let x, y ∈ V and let 0 ≤ t ≤ 1 we have the following implications:

x ∈ V =⇒ ∥x∥ ≤ 1
y ∈ V =⇒ ∥y∥ ≤ 1.

and hence

∥tx + (1 − t)y∥ ≤ t∥x∥ + (1 − t)∥y∥

≤ t + (1 − t)

=1

33
That is, ∥tx + (1 − t)y∥ ≤ 1.

Thus tx + (1 − t)y ∈ V .
Thus we established that there exists a non empty, open, bounded,
convex set namely V in X. And also we have the result that “ Every
Topological vector space is a Hausdorff space. ”
i.e., X is Hausdorff and X contains a bounded convex open neigh-
bourhood V of 0.
Conversely suppose that X is Hausdorff and X contains a bounded
convex open neighbourhood U of 0.
Then we can find an absolutely convex open neighborhood V of 0
such that V ⊂ U .
Boundedness of U =⇒ Boundedness of V . Let ∥ · ∥ be the
Minkowski’s functional of V , that is,

(2.14.1) ∥x∥ = inf{λ > 0 : x ∈ λV if x ̸= 0}

But then ∥x∥ = 0 if x = 0. Write

A(x) = {λ > 0 : x ∈ λV } , for each x ∈ X

Since V is absorbing, there exists ρ = ρ(x) > 0 Such that, αx ∈ A if

|α| ≤ ρ. Take, α = ρ. Then ρx ∈ V . Therefore x ∈ ρ1 V .
1
This means that ρ ∈ A(x) and hence A(x) ̸= ϕ. But A(x) is bounded
below by 0. Thus
0 ≤ ∥x∥ ≤ ρ1 .

Hence 0 ≤ ∥x∥ < ∞ ∀x ∈ X.

Next we shall prove that,
∥x + y∥ ≤ ∥x∥ + ∥y∥ .

Take any x, y ∈ X. Let ϵ > 0. By the definition of infimum, there

exist λ ∈ A(x) satisfying

34
 λ < ∥x∥ + ϵ.

Similarly, there exist µ ∈ A(x) satisfying

µ < ∥y∥ + ϵ.

Hence

(2.14.2) λ + µ < ∥x∥ + ∥y∥ + 2ϵ.

Also, λ ∈ A(x)µ ∈ A(x) imply x ∈ λV and y ∈ µV . Accordingly,

by using Lemma 2.11, we get

x + y ∈ λV + µV = (λ + µ)V

and hence ∥x + y∥ ≤ λ + µ ≤ ∥x∥ + ∥y∥ + 2ϵ by (2.14.2) But ϵ > 0

is arbitrary. Hence

∥x + y∥ ≤ ∥x∥ + ∥y∥.

Next, if µ > 0 then

∥µx∥ = inf{λ > 0 : µx ∈ λV }

= inf{µ µλ : x ∈ µλ V }

= µ inf{ µλ : x ∈ µλ V }

= µ∥x∥

(2.14.3) ∥µx∥ = µ∥x∥

Also, since V is balanced, λV is also balanced. Therefore,

µx ∈ λV ⇐⇒ |µ|x ∈ λV

and consequently,

∥µx∥ = ∥|λ|x∥ = |µ|∥x∥ by (2.14.3)

35
 Thus

∥µx∥ = |µ|∥x∥ ∀ scalars µ

Further, we claim that

x ̸= 0 =⇒ ∥x∥ > 0

Suppose x ̸= 0 in X. Since X is Hausdorff, there exists a

(2.14.4) balanced neighbourhood W of 0 such that x ∈ W .

Since V is balanced, we can find t > 0 such that V ⊂ tW . Given

any ϵ > 0, by the infimum property, choose λ ∈ A(x) such that,
λ < ∥x∥ + ϵ.

Now we have the following implications:

λ ∈ A(x) =⇒ x ∈ λV
x
=⇒ λV ⊂ tW

=⇒ x ∈ λtW

=⇒ λt > 1 by (2.14.4) and by using the fact that W is balanced

1
=⇒ λ > t

1
=⇒ t < λ < ∥x∥ + ϵ
1
=⇒ t < ∥x∥ + ϵ
1
=⇒ 0 < t < ∥x∥ since ϵ is arbitrary

Thus x ̸= 0 =⇒ ∥x∥ > 0. Or equivalently, ∥x∥ = 0 =⇒ x = 0.

Thus ∥.∥ is a norm on X.

Let τ original topology on X and ξ is the norm topology on X. Let

36
W be any τ - neighbourhood of 0. Since V is bounded, we have
rV ⊂ W for some r > 0

Hence {rV : r > 0} is a base at 0 for τ . But V = {x : ∥x∥ < 1}. So

{rV : r > 0} is also a base at 0 for ξ.

Example 2.15. (Non-normable space): Let ℘ be the vector

space of all complex sequence, (xk ).Define
P∞ 1  |xn −yn | 
d(x, y) = n=1 2n 1+|xn −yn |

where x = (xk ) and y = (yk ) ∈ ℘.

Then d is a metric for ℘.
Consider B(0, r) = {x ∈ ℘ : d(x, 0) < r}, where 0 < r < 1.
Given r > 0, choose n such that 12n < r. Take x = δ k
Let λ be an arbitrary complex [Link]
 
d(λx, 0) = 2n 1+|λ| ≤ 21n < r.
1 λ

Hence λx ∈ B(0, r). Since λ is arbitrary in C, we have Cx ⊂

B(0, r).

Therefore, ℘ has no bounded neighborhood of 0. Consequently, ℘ is

not normable.

37
 APPLICATIONS OF
TOPOLOGICAL VECTOR SPACES

Topological vector spaces have many important applications in

various fields of mathematics, physics, and engineering. Here are a
few examples:
• Quantum mechanics: Topological vector spaces are used ex-
tensively in the mathematical foundations of quantum mechan-
ics, which is the study of the behavior of particles on a micro-
scopic scale. In this context, topological vector spaces provide a
framework for describing the properties of quantum states and
the operators that act on them.

• Functional programming: Topological vector spaces are also

used in functional programming, which is a programming paradigm
that emphasizes the use of functions as the primary building
blocks of programs. In this context, topological vector spaces
provide a framework for studying the behavior of higher-order
functions, which are functions that take other functions as input
or output.

• Partial differential equations: Topological vector spaces are

used to study partial differential equations, which arise in many
areas of science and engineering. For example, the space of
solutions to a partial differential equation can be viewed as a
topological vector space, and the topology on that space can
give information about the behavior of the solutions.

• Differential geometry: Topological vector spaces are impor-

tant in differential geometry, which is the study of geometric

38
structures on manifolds. For example, the tangent space to a
manifold at a point is a topological vector space, and the space
of vector fields on a manifold is a topological vector space as
well.

39
 Bibliography

1. [Link] Rao. Functional Analysis, Second Edition,

Narosa Publishing House, New Delhi.

2. Walter Rudin. Functional Analysis, Second Edition, Tata Mc.

Graw-Hill Publishing Company Limited, New Delhi.

3. Robertson A P, W J Robertson. Topological Vector Space,

Cambridge University Press.

4. Treves F. Topological Vector Space, Distributions And Kernals,

Academic Press.

40
DON BOSCO ARTS & SCIENCE
COLLEGE ANGADIKADAVU

DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

ALGEBRAIC GRAPH THEORY

 DON BOSCO ARTS & SCIENCE COLLEGE
ANGADIKADAVU
DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

ALGEBRAIC GRAPH THEORY

Dissertation submitted in the partial

Fulfillment of the requirement for the award of
MSc Degree in Mathematics of
Kannur University
Name : GIFTY P GEORGE
Roll No. : C1PSMM1906

Examiners:
1.
2.
 KANNUR UNIVERSITY

BONAFIDE CERTIFICATE

Certified that this project report “ALGEBRAIC GRAPH THE-

ORY” is the bonafide work of GIFTY P GEORGE who carried out
the project work under my supervision.

Signature of Head Signature of Supervisor

 DECLARATION

I, GIFTY P GEORGE hereby declare that the Project work enti-

tled ALGEBRAIC GRAPH THEORY has been prepared by me and
submitted to Kannur University in partial fulfillment of requirement
for the award of Master of Science is a record of original work done
by me under the supervision of Mrs. PRIJA V, Assistant Professor,
Department of Mathematics, Don Bosco Arts & Science College,
Angadikadavu. I, also declare that this Project work has not been
submitted by me fully for the award of any Degree, Diploma, Title
or recognition before any authority.

Place: Angadikadavu GIFTY P GEORGE

Date: (C1PSMM1906)
 ACKNOWLEDGEMENT

Introduction is the proper place to begin. But first I bow my head

before the Almighty who is always with me. Also I must express my
deepest gratitude to people along the way. No words can adequately
express the sense of gratitude, still I try to express my heartfelt
thanks through words. The outset, I am deeply indebted to my
project supervisor Mrs. PRIJA V Assistant Professor, Department
of Mathematics, Don Bosco Arts and Science College, Angadika-
davu, for the invaluable guidance, loving encouragement and metic-
ulous care towards me throughout my career. I express my deep
sense of gratitude to all the faculty members of the Department of
Mathematics, Don Bosco Arts and Science College, Angadikadavu.
I can never forget the support and encouragement rendered by the
Principal and the Staff of Don Bosco Arts & Science College, An-
gadikadavu. I could not name many who sincerely supported and
helped for the successful completion of this Project. It is my plea-
sure and duty to thank each and everyone of them who walked with
me.

GIFTY P GEORGE
 CONTENTS

1. INTRODUCTION ................................................................ 01

2. PRELIMINARIES ................................................................ 03
0.1 Graph Theoretic Definitions ........................................ 03
0.2 Definitions Related to Algebra ..................................... 05

3. THE ZERO-DIVISOR GRAPH OF A .................................. 08

COMMUTATIVE RING
1.1 Zero-Divisor Graph ...................................................... 08
1.2 Properties of Γ(R) ........................................................ 10

4. CAYLEY GRAPHS .............................................................. 19

2.1 Cayley Graphs ............................................................... 19
2.2 Types of Cayley Graphs .............................................. 20
2.3 Graphical Regular Representation ................................ 31
2.4 Basic Properties of Cayley Graphs ............................... 33
2.5 Applications of Cayley Graphs .................................... 35

5. AUTOMORPHISM GROUP OF GRAPHS.......................... 37

3.1 The Automorphism Group of Specific Graphs ............ 41

6. BIBLIOGRAPHY................................................................. 44
 INTRODUCTION
Graph theory has become a very popular and rapidly growing area of
discrete mathematics for its numerous theoretical development and
countless applications to practical problems. As a research area,
graph theory is still relatively young, but it is maturing rapidly with
many deep results having been discovered over the last couple of
[Link] meaning of the term graph in graph theory is not same
as the term graph that used to represent a function or statistical
data. The study of graph theory was introduced by Euler in 1736.
For the contribution of Euler towards graph theory, he is known as
the father of graph theory. But the term graph was introduced by
Sylvester in a paper published in 1878. It took 200 years since Eu-
ler’s publication to revive graph theory and the first book on graph
theory was published in 1936 by Denes Konig.

In the last decade of 19th century the literature of Algebraic Graph

Theory has grown enormously and many research papers have ap-
peared in this area of graph [Link] graph theory is a fas-
cinating subject concerned with the interplay between Algebra and
Graph Theory. Algebraic tools can be used to give surprising and
elegant proofs of graph theoretic facts and there are many interesting
algebraic objects associated with graphs. In recent years Algebraic
Graph T heory has become an interesting topic of research where
properties of graphs are studied by translating them into algebraic

1
structures and the results and methods of algebra are used to de-
duce theorems about the graphs. On the other hand, many algebraic
structures can also be studied by translating them into graph and
using the properties of graphs. Algebraic graph theory includes the
study of symmetry and regularity properties of graphs which can
be studied using group theory. Asymmetry property of a graph is
related to the existence of automorphism. The concept of automor-
phism of a graph, which is the permutation of vertices that preserves
adjacency, played an important role in the characterization of graphs.
The existence of automorphism led to the idea of permutation group
of graphs. In recent times many researches are showing their interest
towards the relation of graphs with rings. In this project we mention
some significant results in this field which have been proved in the
last century.

2
 PRELIMINARIES

Like all other areas of mathematics, there is a certain amount of

terminology with which we must be familiar in order to discuss the
graphs and their properties. In this chapter we discuss some defini-
tions and properties related to graph and algebra.

0.1 Graph Theoretic Definitions

Definition 0.1.1.
A graph is an ordered triplet (V (G),E(G), IG ) where V (G) is an
empty set, E(G) is a set disjoint from V (G) and IG is an incidence
relation that associates with each element of E(G) and ordered pair
of element of V (G).
Elements of V (G) are called vertices or points or nodes and elements
of E(G) are called edges or links of G.V (G), and E(G) are called
the vertex set and edge set respectively.
Definition 0.1.2.
Simple Graph:A graph is simple if it has no loops and multiple edges.
Definition 0.1.3.
A simple graph G is complete if every pair of distinct vertices of G
are adjacent in G. A complete graph on n vertices is denoted by Kn .
Note 0.1.4.

n(n−1)
1. A simple graph with n vertices can have at most 2 edges.

2. The cardinality of V (G) is called the order of G and the cardi-

3
 nality of E(G) is called the size of G.

Definition 0.1.5.
The set of all neighbors of vertex u is the open neighborhood of u or
the neighborhood set of u and is denoted by N(u).
N[u] = N(u) ∪ {u} is the closed neighborhood of u.
Definition 0.1.6.
Isomorphism of graphs:If G and H are simple graphs, an isomor-
phism from G to H is a bijection ϕ : V(G) −→ V(H) such that u
and v are adjacent if and only if ϕ(u) and ϕ(v) are adjacent in H.
Definition 0.1.7.
Bipartite graph:A graph is bipartite if its vertex set can be parti-
tioned into two non empty subsets X and Y such that each edge of
G has one end in X and the other end in Y .The pair (X,Y) is called
a bipartition of the bipartite graph.
Definition 0.1.8.
Complete bipartite graph:A simple bipartite graph is complete if
each vertex of X is adjacent to all the vertices of Y .
If |X| = p and |Y| = q then the complete bipartite graph is denoted
by Kp,q .
Definition 0.1.9.
A complete bipartite graph of the form K1,n is called a star graph.
Definition 0.1.10. Subgraph:A graph H is called a subgraph of G
if V(H)⊆ V(G) or E(H) ⊆ E(G).
Definition 0.1.11.
A digraph is a graph with directed edges. If G is a digraph then the

4
elements of the set E(G) are ordered pairs and are called arcs.
For a digraph, there are two types of degree defined. For a vertex
v of a digraph G, the number of edges incident from it is called
out-degree of v and is denoted by deg − (v) and the number of edges
incident to it is called in-degree of v and is denoted by deg + (v).

0.2 Definitions related to Algebra

Definition 0.2.1.
A ring R is a set together with two binary operations(addition and
scalar multiplication) such that

1. R is an abelian group with respect to addition(so that R has

a zero element denoted by 0, and every x ∈ A has an additive
inverse -x).

2. Multiplication is associative and distributive over addition.

3. xy = yx ∀ x, y ∈ R, and have an identity element(denoted by

1).

4. There exists 1 ∈ R such that x1 = 1x = x ∀ x ∈ R. The identity

element is unique.

Here the word ring we shall mean a commutative ring with an iden-
tity element, that is, a ring satisfying above axioms.
Definition 2.2.2.
A ideal r of a ring R is a subset of R which is an additive subgroup
and is such that Rr ⊂ r(i.e.,x ∈ R, and y ∈ r imply xy ∈ r).

5
Definition 0.2.3.
If a and b are two non zero elements of a ring R such that ab = 0,
then a and b are divisors of zero or [Link] particular a is
left zero-divisor and b is a right zero-divisor.
Remark 0.2.4.
There is no distinction between left and right zero-divisors in a com-
mutative ring.
Definition 0.2.5.
An integral domain is a commutative ring with unity containing no
zero-divisors. For example, Z and K[x1 , x2 , ..., xn ] (K a field, xi in-
determinates) are integral domains.
Definition 0.2.6.
An element x ∈ R is nilpotent if xn = 0 for some n>0.
A nilpotent element is a zero-divisor but not conversely.
Definition 0.2.7. A unit in R is an element x which divides 1, that
is an element x such that xy = 1 for some y ∈ R.
Definition 0.2.8.
A field is a ring R in which 1 0 and every non zero element is a
[Link] field is an integral domain but not conversely.
Definition 0.2.9.
A ring R is said to be Noetherian if it satisfies the following three
conditions.

1. Every non empty set of ideals in R has a maximal element.

2. Every ascending chain of ideals in R is stationary.

6
 3. Every ideal in R finitely generated.

Definition 0.2.10.
An Artin ring is one which satisfies the descending chain condition
on ideals.
Definition 0.2.11.
A ring is quasi local if it contains a unique maximal ideal.
Definition 0.2.12.
In a commutative ring R an annihilator(ideal) of an element x, de-
noted ann(x), is the set of those elements y for which xy = [Link] terms
of the zero divisor graph this would be the set of vertices adjacent
to x. Note that x itself may be an element of ann(x).

7
 Chapter 1

THE ZERO-DIVISOR GRAPH

OF A COMMUTATIVE RING
1.1 Zero-Divisor Graph

Definition 1.1.1.
Let R be a commutative ring (with unity 1) and let Z(R) be its set
of zero [Link] zero divisor graph of a ring R is a simple graph
(that is with no loops and multiple edges) whose set of vertices con-
sists of all non zero zero divisors of R , with an undirected edge
between two distinct vertices x and y if and only if xy = 0. The zero
divisor graph ofR is denoted byΓ(R).

Exercise 1.1.2.
Let us draw a graph for the ring R = Z12 , where
Z12 = {0, 1, 2, ..., 11}. The zero-divisors of Z12 are {2, 3, 4, 6, 8, 9,
10} which is the vertex set of the zero divisor graph Γ(Z12 ). Edge
set of this graph contains pair of vertices. For example, {3, 4} is
an edge because 3 × 4 = 0 in Z12 .Note that we do not consider the
loops, so even though 6 × 6 = 0 in Z12 . Therefore {6, 6} is not an
edge. Now we list all edges in the complete edge set E as follows.

E = {{2, 6}, {3, 4}, {3, 8}, {4, 6}, {4, 9}, {6, 8}, , 9}, {6, 10}}.

8
The graph Γ(Z12 ) is,

Figure 1.1

The main objective of this topic is to study the interplay of ring the-
oretic properties of R with graph theoretic properties of Γ(R). This
study helps to illuminate the structure of Z(R).

The idea of a zero-divisor graph of a commutative ring was intro-

duced by [Link],where he was mainly interested in [Link]
investigation of colorings of a commutative ring was then contin-
ued by [Link] and M. Naseer. Their definition was slightly
different than ours; they let all elements of R be vertices and had
distinct x and y adjacent if and only if xy = 0. We will denote their
zero-divisor graph of R by Γ0 (R). In Γ0 (R), the vertex 0 is adjacent
to every other vertex, but non-zero-divisors are adjacent only to 0.
Note that Γ(R) is a (induced) subgraph of Γ0 (R). Our results for
Γ(R) have natural analogs to Γ0 (R); however we feel our definition
better illustrates the zero-divisor structure of R.

9
Remark 1.1.3.
The zero divisor graph Γ(R) is the empty graph if and only if R is
an integral domain.

1.2 Properties of Γ(R)

In this section we show that Γ(R) is always connected and has small
girth and diameter and we determine which complete graphs and
star graphs may be realized as Γ(R).

Theorem 1.2.1.
Let R1 and R2 be two finite commutative rings. If R1 ∼
= R2 , then
ΓR1 ∼
= ΓR2 .

Proof.
Assume R1 ∼
= R2 . Then there exists an isomorphic map
ϕ : R1 → R2 , Since ϕ is a ring isomorphism that preserves zero-
divisor. That is ϕ maps zero-divisor of R1 to a zero-divisor of R2 .
Let ai and bi be the zero divisor of R1 and R2 respectively, where i
= 1, 2, ..., n. Then ϕ (ai ) = bi for some i. Also if ai is adjacent to
aj , then ϕ (ai ) is adjacent to ϕ(aj ), for all i j. Then ϕ is a graph
isomorphism. Hence Γ(R1 ) ∼
= Γ (R2 ).

10
Remark 1.2.2.
Converse of the above theorem is not [Link] this consider the ex-
ample given below.
Exercise 1.2.3.
Below are the zero-divisor graphs for several rings. These examples
show that non isomorphic rings may have the same zero divisor graph
and that the zero divisor graph does not detect nilpotent elements.

Figure1.2

Exercise 1.2.4.
Path graph (Linear graph) cannot always be realised as a zero divisor
graph. It is not possible to find a ring which gives Γ(R) with vertices
{a, b, c, d} and edges a - b, b - c, c - d. That is there is no ring R
such that Γ(R) is,

Figure 1.3

11
We can give a proof for it as follows:
Suppose R is a ring with Γ(R) as shown in the figure above.
Then Z(R)= {0, a, b, c, d}.
Since {a, b} and {b, c} are edges, so ab = bc = 0.
Therefore b(a + c) = 0 → a + c = 0 or a + c ∈ (R) . Since b makes
an edge with a and c only, a + c = 0, a, b, [Link] will show that
the possibilities gives a contradiction. Suppose a + c = 0. Then
ad = (-c)d = [Link] there is no edge between a and d. So a + c ̸=
0. Also a + c cannot be a or c either because then c = 0 or a = 0
by the cancellation property of rings. Now let us assume that a + c
= b. By the symmetry of graph we can assume that b + d = c. Then

(a + c)(b + d) = bc
⇒ (a + c)b + (a + c)d = bc
⇒ ad = bc = 0.

This is a [Link] the proof.

Exercise 1.2.5.
We have seen above that Γ(R) can be a triangle or square. But Γ(R)
cannot be an n-gon for n ≥ 5.
It can be proved as in the above case.

12
 Figure 1.4

First consider the case n = 5.

Let Z(R) = {0, a, b, c, d, e} with ab = bc = cd = de = ae = 0
and no other zerodivisor [Link] (-a)b = 0 and (-a)e = [Link]
-a=a. Similarly -x=x for every x ∈ Z(R).Also (b + e)a = 0 or b +
e = 0, a, b, [Link] we cannot have b + e = b and b + e = e. Also
b + e = 0 is not [Link]
b + e = a and thus a2 = 0. Since we consider simple graph there is
no [Link] x2 = 0 for x ∈ Z(R) is not possible. The case for n
≥ 5 is similar.

Remark 3.2.6.
For each n ≥ 3 there is a zero divisor graph with n-cycle.

Remark 3.2.7.
Let A and B be integral domain and let R = A × B. Then Γ(R) is

13
a complete bipartite graph with | ΓR| = |A| × |B| - 2.

Exercise 1.2.8.
If A = Z12 , then Γ(R) is a star graph with |ΓR|=|B|.
For example, Γ(Fp ×Fq )=K p−1,q−1 and Γ(Z2 ×Fq )=K 1,q−1 .We give two
specific examples.

Figure 1.5

Remark 1.2.9.
ΓR may be [Link] is a ring may have an infinite number of
zero divisors. But ΓR is probably of most interest when it is finite,for
then one can draw Γ(R) .

We will show that Γ(R) is finite only when R itself finite.

Theorem 1.2.10.
Let R be a commutative ring. Then Γ(R) is finite if and only if
either R is finite or an integral [Link] particular, if
1 ≤ |Γ R| < ∞, then R is finite and not a field.

14
Proof.
Suppose that Γ(R) is finite and non empty. Then there are non zero
x, y ∈ R with xy = 0. Let I = ann(x). Then y ∈ I and in fact,ry ∈ I
for all r ∈ R and I ⊆ Z(R). Thus R must be [Link] R is in-
finite with finitely many zero divisors. Since I is a subset of the zero
divisors of R, it is [Link] there exists an i ∈ I such that J = {r ∈
R : ry = i} is [Link] r, s ∈ J, (r s)y = 0. Thus, ann(y) is infinite,
contradicting the fact that there are only finitely many zero divisors.

Definition 1.2.11.
A graph Γ is connected if there is a path between any two distinct
vertices.
For distinct vertices x and y of Γ, let d(x, y) be the length of the
shortest path from x to y (d(x, y) = ∞ if there is no such path).

Definition 1.2.12.
The diameter of Γ is diam(Γ) = sup {d(x, y) : x, y ∈ Γ}.
The girth of Γ, denoted by g(Γ), is defined as the length of the
shortest cycle in Γ (g(Γ)=∞ if Γ contains no cycle). Note that if Γ
contains a cycle, then g(Γ) ≤ 2diam(Γ)+1.

We next show that the zero divisor graphs are all connected and
have exceedingly small(≤ 3) diameter and girth. Thus, for distinct
x, y ∈ Z(R)∗ , either xy = 0, xz = zy = 0 for some z ∈ Z(R)∗ - {x,

15
y}, or xz1 = z1 z2 = z2 y = 0 for some distinct z1 , z2 ∈ Z(R)∗ -{x, y}.

Theorem 1.2.13.
Let R be a commutative ring (not necessarily finite). Then Γ R is
connected. Moreover diam ( Γ R )≤ 3.

Proof.
Let x, y ∈ Γ(R) , with x ̸= y. If xy = 0, then d(x, y) = 1. Suppose
now that xy ̸= 0. If x2 = 0 =y 2 , then x - xy - y is a path of length
two and d(x, y) = 2. Suppose x2 = 0 and y 2 ̸= 0. There exists an
element b ∈ Γ(R) with b ∈ y such that by = 0. If bx = 0, then x -
b - y is a path of length two between x and y. In either case, d(x, y)
= 2. A symmetric argument holds if y 2 = 0 and x2 ̸= 0. Thus we
may suppose that neither x2 and y 2 is zero. Then there exists non
zero zero divisors a, b ∈ Γ(R) (not necessarily distinct) with ax = 0
= by. If a = b, then x - a - y is a path of length two, and hence d(x,
y) = 2. Thus we may assume that a ̸= b. Consider the element ab.
If ab = 0, then x - a - b - y is a path of length 3, and hence d(x, y)
≤ 3. If ab ̸= 0, then x - ab - y is a path of length 2, and hence d(x,
y) = 2. In all of the cases, there is a path between x and y of length
less than or equal to 3, and since x and y were arbitrary it follows
that the diameter of Γ(R) is less than or equal to 3.

16
Remark 1.2.14.
It is clear that the zero divisor graphs of Z4 , Z2 × Z2 and Z6 have
diameter 0,1 and 2 respectively. We can show that the diameter 3 is
also achieved. For this consider the ring Z2 × Z2 × Z2 . The distance
between the elements (1,1,0) and (0,1,1) is three. In fact, a shortest
path is (1,1,0) - (0,0,1) - (1,0,0) - (0,1,1).

The fact that the distance between points is small also constraints
the length of the shortest cycle, that is to say, the girth of the graph.
The following corollary make use of the previous theorem to estab-
lish a bound for the girth of a zero divisor graph.

Corollary 1.2.15.
If R is a ring, then the girth of Γ(R) is less than eight.
Proof.
It is enough to suppose to the contrary that we could find a ring
R such that Γ(R) has a smallest cycle C of length exactly 8, say,
v0 - v1 - v2 - v3 - v4 - v5 - v6 - v7 - v8 = v0 . Let P1 denote the
path v0 - v1 - v2 - v3 - v4 and P2 denote the path v0 - v7 - v6 - v5 -
v4 .To help in visualizing the proof the figure below provides a hypo-
thetical representation of each of the two main considerations which
[Link] observe that for 0 < i < 4 < j < 8, vi and vj are not
connected. Assume the case is otherwise. Then v0 - v1 ...- vi - vj ...-
v0 is a cycle of length less than eight contradicting the assumption
that the girth is eight. Now assume there is a path v0 - x - y - v4

17
by previous theorem(If not, then there is a path v0 - x -v4 . The
proof goes through if in this case we just identify x and y. The im-
possibility of v0 and v4 being adjacent is apparent). The fact just
proved implies that the path v0 - x - y - v4 intersects P1 , or perhaps
P2 , but not [Link], by symmetry we may as well assume that
the path v0 - x - v4 does not intersect P2 . This assumption yields a
cycle v0 − x − y − v4 − v5 − v6 − v7 − v8 = v0 of length seven, the
final contradiction.

Figure 1.6

Exercise 1.2.16.
Let T be a an integral domain, and n ≥ 3 an integer. Define

R = T[X1 ,X2 ,..., Xn ] / (X1 X2 ,X2 X3 ,...,Xn X1 )

and let xi be the coset of Xi in R. Then x1 -x2 ...-xn -x1 - is a cycle of

length n.

18
 Chapter 2

CAYLEY GRAPHS

2.1 Cayley Graphs

Cayley graphs, named after mathematician Arthur Cayley, is an

important concept relating group theory and graph theory. Cay-
ley graphs are frequently used to render the abstract structure of a
group easily visible by a way of representing this structure in graph
form. Properties of a group G, such as its size or number of genera-
tors, become much easier to examine when G is rendered as a Cayley
graph. Though also referred to as group diagrams, the definition of
Cayley graphs is suggested by Cayley’s theorem, which states that
every group G is isomorphic to a subgroup of the symmetric group
acting on G.

Definition 2.1.1.
Let G be a finite group with identity 1. Let S be a subset of G
/ S and S = S −1 . That is, s ∈ S if and only if s−1 ∈ S.
satisfying 1 ∈
The Cayley graph Cay (G, S) on G with connection set S is defined
as follows:
(a) the vertices are the elements of G;
(b) there is an edge joining g and h if and only if h = sg for some s

19
∈ S.
The set of all Cayley graphs on G is denoted by Cay (G).

Note 2.1.2.
When G is an abelian group, we use additive notation for the group
operation. Hence we write S = -S for the connection set and h = s
+ g(for some s ∈ S).
For example consider Cay (Z9 , {1, 3, 6, 8}).

Figure 2.1

2.2 Types of Cayley Graphs

There are three main types of Cayley graphs:

1. Cayley Digraphs

20
 2. Cayley Color Graphs

3. Simple Cayley Graphs

(1)Cayley Digraphs: Let G be a finite group and let S ⊆G. The

Cayley digraph on G with connection set S, denoted D(G, S), is the
digraph with vertex set G and with (g, h) being an arc if and only
if gh−1 ∈ S.
Consider the example: Cay (Z10 , {1, 3, 7,9}) given below

Figure 2.2

(2) Cayley Color Graphs: We can extend the idea of Cayley di-
graph to a Cayley color graph, where S is a generating set for G, each
si ∈ S is assigned a color, and if g = si h, then the arc connecting
them is colored si .
Example given below shows the Cayley color graph of s3 with con-
nection set S = {a, b}, where a = (1 2 3) and b = (1 2).

21
 Figure 2.3

Notice that it is often helpful to represent each group element in

terms of group presentations, thus displaying the arcs more clearly:
Now let us examine the automorphism groups of these Cayley color
graphs. First, let us present another definition.

Definition 2.2.1.
An automorphism ϕ ∈ Aut (D(G, S)) is color preserving if given an
arbitrary arc {g, h}, {g, h} and {ϕ (g) , ϕ (h)} have the same color.

Before presenting the main theorem, we must first present an in-

termediate result.

Proposition 2.2.3.
Let G be a group with generating set S and let ϕ be a color pre-

22
serving permutation on V (D(G, S)). Then ϕ is a color preserving
automorphism of D(G, S) if ϕ (gh) = ϕ (g)h.
Proof.
Suppose that ϕ (gh) = ϕ (g)h. To show that ϕ is color preserving,
we need to show that if gh−1 = s then ϕ(gh−1 )=s.
Suppose gh−1 = s. Then,

ϕ(gh−1 ) = ϕ(g)h−1
= ϕ(g)g −1 s
= ϕ(gg −1 s)
=s

Theorem 2.2.4.
Let G be a nontrivial group with generating set S. Then the group
of color preserving automorphisms of D(G, S) is isomorphic to G.
Proof.
Let G be a group of order n and gi ∈ G for 1 ≤ i ≤ n. Define the
map ϕi : V (D(G, S)) −→ V (D(G, S)) by ϕi (g) = gi g. This map
is surjective, since given any g ∈ V (D(G, S)), g = ϕi (gi−1 g). This
map is also injective since if

ϕi (g1 ) = ϕi (g2 )
⇒ gi g1 = gi g2
⇒ g1 = g2 .

Now let g1 ,g2 ∈ G. By above proposition, ϕi is a color preserving

automorphism since,

23
 ϕi (g1 g2 ) = gi (g1 g2 )
= (gi g1 )g2
= ϕi (g1 ) g2 .

Now let A = {ϕi : 1 ≤ i ≤ n} and define a map α : G −→ A by

α(gi ) = ϕi . We must verify this map is an isomorphism from G to
the groups of color preserving automorphisms. This map is injective
since ϕi ̸= ϕj when i ̸= j. We will show that the map is surjective
by proving that for any color preserving automorphism ϕ, ϕ = ϕi for
some ϕi ∈ A. Let e be an identity element of G and let ϕ(e) = gi .
Given an arbitrary element gj ∈ G, we can write this element as a
product of generators from our generating set s. Let gj = s1 r1 s2 r2 ...
sm rm .Then by above proposition we can write,

ϕ(gj ) = ϕ(egj )
= ϕ(es1 r1 s2 r2 ... sm rm .)
= ϕ(e)s1 r1 s2 r2 ... sm rm .
= gi gj
= ϕi (gj ).

Therefore ϕ = ϕi and α is surjective.

Finally we must show that α preserve the group operation. Since,

ϕij (g) = (gi gj )g

= gi (gj g)
= ϕi (gj g)
= ϕi (ϕj (g))

We can deduce that,

24
 α(gi gj ) = ϕij
= ϕi ◦ ϕj
= α(gi ) ◦ α(gj ).

Thus we have shown that given any group G, we can construct a

colored digraph representation of this group whose color preserving
automorphism group is isomorphic to G itself. We will now turn
our attention to a more generalized version of Cayley graphs and
examine their properties

3) Simple Cayley Graphs: Simple Cayley graphs make up a large

amount of graphs in an important family of graphs called Vertex-
Transitive graphs. They are defined nearly identically as Cayley
digraphs, however they contain an edge set instead of an arc set,and
the connection set S must be closed under inverses. More specifi-
cally, if s ∈ S, then s−1 ∈ S.
To further understand these graphs, let us first consider the motiva-
tion behind this restriction on S. Given a Cayley digraph, we know
that there exists an arc u,v if and only if u = sv. However, if s−1
∈ S then v = s−1 u and therefore v,u is also an arc. Thus, we can
replace these two arcs by a simple edge. This restriction on S thus
replaces every pair of arcs with an edge, turning our digraph D (G,
S) into a simple Cayley graph. We will denote the simple Cayley
graph of a group G on the connection set S by Cay (G, S). To study
these graphs further let us first review some definitions and theo-
rems. Recall that given a group G acting on a set V , x, y ∈ V

25
are G - equivalent if gx = y for some g ∈ G. This G - equivalence
is an equivalence relation on V . Each partition of V into such an
equivalence class is called an orbit of V under G. Also recall that the
stabilisersubgroup of G for an element x ∈ V , denoted Gx , is the
set of all group elements in G that fix x.
That is,
Gx = {g ∈ G : gx = x }

Definition 2.2.5.
A graph Γ is vertex transitive if there exists a single orbit of V (Γ)
under Aut(Γ). That is, given any v, u ∈ V (Γ) there exists a ϕ ∈
Aut(Γ) such that ϕ (v) = u.
Note that here Γ denote simply a graph not a zero divisor graph.

Theorem 2.2.6.
Every Simple Cayley Graph is Vertex-Transitive.
Proof.
Let ρg : v −→ vg for all v ∈ V (Cay (G, S)). Clearly, ρg permutes
the elements of V (Cay (G, S)). To show that ρg ∈ Aut(Cay (G, S))
we must show that {v, u} ∈
E (Cay (G, S)) if and only if {vg, ug} ∈ E (Cay (G, S)).
Suppose {v, u} ∈ E (Cay (G, S)). Then we know that v = su for
some s ∈ S or equivalently, vu−1 = s. But

(vg)(ug −1 ) = vgg −1 u−1

26
 = vu−1 .

Therefore, {v, u} ∈ E (Cay (G, S)) if and only if {vg, ug} ∈ E (Cay
(G, S)) and ρg is the group of Cay(G, S).
Now given any vertices v, u the mapping ρv−1 u maps v to u, since
vv −1 u = u. Thus any Cayley graph is vertex-transitive.
Just as a subgroup of an automorphism group for a Cayley color
graph of G contained a subgroup isomorphic to G, simple Cayley
graph possess a similar property.

Definition 2.2.7.
A group G acting on a set V is semi-regular if Gv = e for all v ∈ V.
If a group is semi-regular and transitive, then we say it is regular.

Theorem 2.2.8.
Let G be a group and S be an inverse closed subset of G. Then
Aut(Cay(G, S)) contains a regular subgroup isomorphic to G.
Proof.
Let G be a group with connection set S and Cay(G, S) be the Cayley
graph for G defined on S. Now consider the mapping ρg : v −→ vg.
We know that ρg ∈ Aut(Cay(G, S)),and it can be easily shown that
H = {ρg : g ∈ G} is a subgroup of G. This group acts regularly on
G, since it is clearly semi-regular and is also transitive by the proof
of the above theorem. The map ϕ : H −→ G defined by ϕ(ρg ) = g
is an isomorphism by Cayley’s theorem. Thus, H ≤ Aut(Cay(G, S))
is isomorphic to G.

27
It is natural to ask whether all vertex-transitive graphs are Cay-
ley graphs. This question was negatively answered by the counter
example of the Petersen graph.

Figure 2.4

The Petersen graph on 10 vertices is the smallest example of a vertex-

transitive graph which is not a Cayley graph.
The following theorem gives criteria for when a graph is indeed a
Cayley graph for some group.

Theorem 2.2.9.
If a group G acts regularly on the vertices of the graph Γ, then Γ is
a Cayley graph for G relative to some inverse closed connection set
S.

28
Proof.
Let Γ be a graph with degree n and let G be a group acts normally
on V (Γ) = {v1 , v2 , ..., vn }. Since G acts normally on Γ, there ex-
ists a unique element gi ∈ G such that gi v1 = vi Now define a set S by,

S = {gi ∈ G : {vi , v1 } ∈ E(Γ)}

Now let x,y be arbitrary elements in V (Γ). Then since gx is an

automorphism of Γ, so{x, y} ∈ E(Γ) if and only if {gx−1 x,gx−1 y} ∈
E(Γ). Since gx u = x, gx−1 x = u. Furthermore, since gy u = y, gx−1 gy u
= gx−1 y .Therefore,

gx−1 gy ∈ S ←→ {u,gx−1 y} ∈ E(Γ) ←→ gx−1 x,gx−1 y ∈ E(Γ) ←→ {x,y}

∈ E(Γ).

So, if we identify every vertex x with the group element gx , then

Γ = Cay(G, S). Since Γ is undirected, S is closed under inverses.

So given any graph Γ and a subgroup G acts regularly on V (Γ)

if and only if Γ is a Cayley graph for G for some connection set S.
Notice that our definition of a Simple Cayley graph does not require
our connection set to be a generating set for G. Our next theorem
shows the consequence of S being a generating set for G. First, let
us present two definitions.

29
Definition 2.2.10.
A path of length r from vertex x to vertex y in a graph is a sequence
of r + 1 distinct vertices starting with x and ending with y such that
consecutive vertices are adjacent.

Definition 2.2.11.
A graph Γ is connected if there is a path between any two vertices
of Γ.

Theorem 2.2.12.
The Cayley graph Cay(G, S) is connected if and only if S is a gen-
erating set for G.

Exercise 2.2.13.
Consider the group G = Z5 , with connection set S = {2, 3}.Since
0 + 2−1 = 3, 0 + 3−1 = 2, 1 + 3−1 = 3, 1 + 4−1 = 2, 2 + 4−1 = 3.
Below is the graph ,Cay(Z5 ,{2, 3}).

Figure 2.5

30
Notice that the graph is connected, since S is a generating set for G.

2.3 Graphic Regular Representation

We know that given a group G and a connection set S, there is

a subgroup of Aut(Cay(G, S)) that is isomorphic to G. Often, this
subgroup is a proper subgroup of Aut(Cay(G, S)), meaning the en-
tire Aut(Cay(G, S)) is not isomorphic to G. Consider for example
G = S3 . In the case of finite abelian groups Aut(Cay(G, S)) is itself
isomorphic to G.

Definition 2.3.1.
A group G admits graphical regular representation if the automor-
phism of Cayley graph is isomorphic to G.

Let us again provide an intermediate result before presenting our

main theorem for this section.

Proposition 2.3.2.
Let Cay(G, S) be a Cayley graph for G defined on the connection
set S. Suppose that ϕ is an automorphism of the group G that fixes
S set-wise. Then, ϕ regarded as a permutation of the vertices of
Cay(G, S) fixes the vertex corresponding to the identity element of
G.

31
Proof.
Let ϕ be a group automorphism. Then ϕ must fix the identity ele-
ment. Let us show that ϕ is a graph automorphism. Suppose that
v, w are adjacent vertices. Then vw−1 ∈ S. Therefore, ϕ(vw−1 ) ∈ S.
But, ϕ(vw−1 ) = ϕ(v)ϕ(w−1 ). So ϕ(v) and ϕ(w) are adjacent. There-
fore, ϕ preserves adjacency and is a graph automorphism.

Theorem 2.3.3. Let Γ be a vertex-transitive graph whose auto-

morphism group G = Aut(Γ) is abelian. Then G acts regularly on
V (Γ), and G is an elementary abelian 2-group.
Proof.
Suppose Γis a vertex-transitive graph with automorphism group
G = Aut(Γ). Let g, h ∈ G. Suppose that g fixes an arbitrary vertex
v ∈ V (Γ). Now,
gh(v) = hg(v) = h(v).

Therefore, g fixes h(v) as well. However, since G acts transitively on

V (Γ) any vertex, u can be written as h(v) = u for some h ∈ G. Thus
g fixes every vertex and the stabilizer is the identity. Therefore, G
acts regularly on V (Γ) and Γ = Cay(G, S) for some connection set S.
Now consider the map g −→ g −1 . This map preserves the adjacency
since Aut(Γ) is abelian and is therefore a graph automorphism. Fur-
thermore, since S is closed under taking inverses, this mapping fixes
S set-wise. Thus, by above proposition, this map must fix vertex 1.
But since G acts regularly on V (Γ), this map must be the identity
[Link] g = g −1 and G is an abelian 2-group.

32
Let us think for a moment what this theorem officially states. By
theorem, If a group G acts regularly on the vertices of the graph Γ,
then Γ is a Cayley graph for G relative to some inverse closed con-
nection set S, we know that if G acts regularly on a graph Γ, then Γ
is a Cayley graph for G. Therefore, we know that if a graph has an
abelian automorphism group, then this abelian group has a graphi-
cal regular representation. Furthermore, we know that this can only
happen when G is an abelian 2-group. Thus, the only abelian groups
that have a graphical regular representation are abelian 2-groups.

2.4 Basic Properties of Cayley Graphs

Cayley diagrams are one of many representations of finite groups.

They provide a means of representing a group diagrammatically
and various properties of groups including commutativity can be
extracted from the graph. The Cayley diagram also provides suffi-
cient information to test for isomorphism between groups, and thus
is a useful tool for recognizing the type of a given group.
Let S = {g1 , g2 , ..., gn } be a set of distinct elements and let G is the
group generated by the set S. We can define a relation ∼ on G such
that a ∼ b if and only if b = gi a, where gi ∈ S. Then the Cayley
digraph is the digraph formed from the relation ∼, where the vertex
set of the graph is the group G.
Here are various simple properties of Cayley digraphs.

1. Let a be a vertex in D(G, S) and |S| = n. Then deg+ (a)

33
 = deg− (a) = n. Its proof is as follows: Let a ∈ G. Then
g1−1 a,g2−1 a,...gn−1 a n a are n distinct elements in G(by closure),
since suppose that

gi−1 a = gj−1 a
⇒ gi−1 = gj−1
⇒ gi = gj .

This is a contradiction to the fact that elements of S are dis-

tinct. Hence gi−1 a ∼ a, i = 1,2,...,[Link] deg− (a) = n.
Similarly g1 a, g2 a, ..., gn a are n distinct elements in G, and a ∼
gi a for all i = 1, 2, ..., [Link],deg+ (a) = n.

2. Cayley digraph is strongly connected(That is, there is a path

from a to b and from b to a whenever a and b are vertices in
the graph.

3. If G ̸= {e} then ∼ is irreflexive. That is Cayley digraphs has

no loops.

4. If for all g ∈ S, g −1 ∈ S, then ∼ is a symmetric relation. That

is, Cayley digraph is an undirected graph.

5. Let G1 = S1 and G2 = S2 be isomorphic groups,

| S1 | ≤ | S2 |, then D(G1 , S1 ) is isomorphic to a subgraph of

34
 D(G2 , S2 ).

2.5 Applications Of Cayley Graphs

There are many meaningful applications of Cayley graphs and we

conclude with a quick outline of a few of them.

1. Any problem for which graphs are used as a model and the
use of edges is being optimized provides a natural setting for
Cayley [Link] example, suppose that we are to construct
a network for which the number of direct links we may use is
restricted, but we want to maximize the probability that the
network remains connected after some links or vertices of the
network are deleted. Then there is a strong tendency for the
graph to be either vertex-transitive or close to [Link]
Cayley graphs in particular the K-dimensional cube QK have
been extensively studied by researchers working with networks.

2. It is possible to tell from a Cayley digraph whether or not the

corresponding group is commutative.

3. Another application of Cayley diagrams is in that on binary

representation of data, namely Gray codes. A gray code of
length n is a sequence of n bit binary strings, with the property
that consecutive words differ by at most one element. Gray
codes are useful in mechanical encoders since a slight change in
location only affects one bit. Using a typical binary code , up to

35
n bits could change, and slight misalignments between reading
elements could cause high levels of error since flipping a bit will
increase or decrease the value by a power of two. For example,
an error flipping the MSB of an 8-bit word will change the value
by 27 .
Gray codes can be represented by the direct product (C2 )n .The
difference between Gray code and normal binary code is the
ordering of the elements. In gray code the greater than or equal
to relation is defined as follows:
For a, b ∈ (C2 )n , a ≥ b if and only if a → b. The fact that
((C2 )n , ≥) is a totally ordered set follows from that we can
always find a Hamiltonian path in ((C2 )n , S)(since every two
elements in the path are comparable). Thus ((C2 )n , ≥) is a
well ordered set by choosing the starting vertex(element) of the
Hamiltonian path as the least element.

36
 Chapter 3

AUTOMORPHISM GROUP OF
GRAPHS

Definition 3.0.1.
A graph automorphism is an isomorphism from a graph to itself. In
other words, an automorphism of a graph G is a bijection
ϕ : V (G) −→ V (G) such that uv ∈ E(G) if and only if ϕ(u)ϕ(v) ∈
E(G).

This definition generalizes to digraphs and graphs with loops. Here

we will assume that all graphs are undirected graphs with no loops or
multiple edges. Let Aut(G) denote the set of all automorphisms on
a graph G. Note that this forms a group under function composition.
In other words,

1. Aut(G) is closed under function composition.

2. Function composition is associative on Aut(G). This follows

from the fact that function composition is associative in gen-
eral.

3. There is an identity element in Aut(G). This is mapping e(v)

= v for all v ∈ V (G).

4. For every σ ∈ Aut(G), there is an inverse element σ −1 ∈ Aut(G).

37
 Since σ is a bijection, it has an inverse. By definition, this is an
automorphism.

Thus, Aut(G) is the automorphism group of G. Consider the graph

illustrated below.

Figure 3.1

An automorphism of G can leave every vertex fixed, this is the iden-

tity automorphism e. An automorphism of G can swap vertices a
and c and leave the others alone. This is the automorphism α =
(a, c). Similarly we can swap vertices b and d while leaving a and c
fixed. This results in the automorphism β = (b, d). Finally, we can
swap vertices a and c and swap vertices b and d. This results in the
automorphism αβ = (a, c) (b, d). Hence, Aut(G) is isomorphic to
the Klein - 4 - group, V4 =Z2 × Z2 .
We can use the automorphism group to define a relation between
two vertices in G. Let u, v ∈ V (G), vertex u relates to v if there

38
exists ϕ ∈ Aut(G) such that ϕ (u) = v. We claim that this is an
equivalence relation.

1. Reflexive: Note that e(u) = ufor all u ∈ V (G), where eis the
identity automorphism.

2. Symmetric: If ϕ (u) = v, then ϕ−1 (v) = u.

3. Transitive: If ϕ (u) = v and σ (v) = w, then σ(ϕ(u)) = w.

Thus the relation is an equivalence relation. Like all equivalence

relations, this induces a partition of the vertex set into equivalence
classes. These classes are usually called automorphism classes. If all
the vertices of the graph are in the same automorphism class, then
we say that the graph is vertex transitive.
Some facts about the automorphism of a graph.

Proposition 3.0.2.
Let G be a graph.

1. (Degree preserving)For all u ∈ V (G) and for all ϕ ∈ Aut(G),

deg(u) = deg(ϕ(u)).

2. (Distance preserving)For all u, v ∈ V (G) and for all ϕ ∈ Aut(G),

d(u, v) = d (ϕ(u), ϕ(v)).

3. The automorphism group of G is equal to the automorphism

group of the complement G.

39
Proof.

1. Let u ∈ V (G) with neighbors u1 , u2 , ..., uk . Let ϕ ∈ Aut(G).

Since ϕ preserves adjacency, it follows that ϕ(u1 ), ϕ(u2 ), ...,ϕ(uk )
are neighbors of ϕ(u). Therefore, deg(ϕ(u)) ≥ k. If v ∈
/ {u1 , u2 , ..., uk }
is a neighbor of ϕ(u), then ϕ−1 (v) is a neighbor of u. Therefore,
the neighbors of ϕ(u) are precisely ϕ(u1 ), ϕ(u2 ), ..., ϕ(uk ).Hence
deg(u) = deg(ϕ(u)).

2. Let u, v ∈ V (G) and let ϕ ∈ Aut(G). Suppose that the distance

from u to v is d(u, v) = d. Further, let u = u0 , u1 , ..., ud−1 , ud = v
be a shortest path from u to v. Since ϕ preserves adjacency, ϕ(u)
= ϕ(u0 ), ϕ(u1 ), ..., ϕ(ud−1 ), ϕ(ud ) = ϕ(v) is a path from ϕ(u)
to ϕ(v). Thus, d(ϕ(u), ϕ(v)) ≤ d = d(u, v). Suppose that ϕ(u),
v1 , ..., vm−1 ,ϕ(v) is a shortest path from ϕ(u) to ϕ(v).It follows
that u, ϕ−1 (v1 ),...,ϕ−1 (vm−1 ), v is a shortest path from u to v.
It follows that d(u, v) ≤ d(ϕ(u) ,ϕ(v)). Hence, have equality.

3. We will proceed by showing set inclusion in both directions.

First suppose that we have some permutation σ ∈ Aut(G), and
an edge e ∈
/ EG . By the definition of the complement of a graph,
it follows that e ∈ EG . Similarly, we know from the definition of
a graph automorphism that σ(e) ∈
/ EG , and hence we find that
σ(e) ∈ EG . Thus we have shown that σ is an automorphism
of G, and thus σ ∈ Aut(G).Note that G is isomorphic to the
complement of G, so we can simply interchange G and G and
find that if τ ∈ Aut(G), then τ ∈ Aut(G). Thus we have shown

40
 that Aut(G) = Aut(G).

3.1 The Automorphism Group of Specific Graphs

In this section, we give the automorphism group for several families
of [Link] the vertices of the path, cycle, and complete graph on
n vertices be labelled v0 , v1 , ..., vn−1 in the obvious way.

Theorem 3.1.1.

1. For all n ≥ 2, Aut(Pn ) ∼

= Z2 , the second cyclic group.

2. For all n ≥ 3, Aut(Cn ) ∼

= Dn , the nth dihedral group.

3. For all n, Aut(Kn ) ∼

= Sn , the nth symmetric group.

Proof.

1. As in the proof of the above proposition, any automorphism ϕ

∈ Aut(Pn ) must either map a vertex of degree one to a vertex
of degree one. Thus either ϕ(v0 ) = v0 and ϕ(vn−1 ) = v0 or
ϕ(v0 ) = vn−1 . In either case, the orbit of the remaining vertices
is precisely determined by their distance from v0 . In the first
case, ϕ(vi ) = vi for all i. This results in the identity automor-
phism. In the second case, ϕ(vi ) = vn - 1 - i for all i. Thus,
Aut(Pn ) ∼
= Z2 .

2. Consider the mapping ρ(vi ) = vi+1 , where the computation on

the indices is computed modulo n. Since vi vi+1 is an edge in the
graph, ρ is an automorphism. If n is even, then consider the

41
 mapping τ (vi ) = vn−1−i and τ (vn−1−i ) = vi for i = 0, 1, ..., n2 − 1.
If n is odd, then consider the mapping τ (v0 ) = v0 , τ (vi ) = vn −i,
and τ (vn − i) = vi for i = 0, 1, ..., n−1
2 . In both cases, vi vi+1

and vn − 1 − ivn − 2 − i are both edges in Cn . Thus, τ is an

automorphism. Note that ρn = τ 2 = e and ρτ = τ ρn−k . Hence ρ
and τ generate the nth dihedral group, Dn . Since we can think
of Cn as a regular n - gon, we have that Aut(Cn ) ∼
= Dn .

3. Since S1 is the trivial group, the result holds for n = 1. For

the remainder of the proof, let n ≥ 2. Let x and y be distinct
vertices of Kn . Consider the mapping ϕ(x) = y, ϕ(y) = x and
ϕ(v) = v for all other v ∈ V (Kn ). Since x and y are both
adjacent to every vertex, ϕ is an automorphism of Kn . Thus,
every transposition of two vertices is an automorphism. Since
the set of all transpositions generates Sn , the result follows.

Theorem 3.1.2.
For the complete bipartite graph, Kn,m , if n > m, then
Aut(Kn,m ) ∼
= Sn × Sm .
Proof.
By above theorem, Aut(Kn ) ∼
= Sn . By proposition 3.0.2 Aut(K n ) ∼
=
Sn . Thus, any automorphism of the form (xi , xj 4 ) or of the form
(yk , yl ) is in Aut(Kn,m ). Thus, Sn × Sm is a subgroup of Aut(Kn,m ).
Now suppose that n > m. Since deg(xi ) = m and deg(yj ) = n,it
follows from the proof of proposition 3.0.2 that there is no automor-

42
phism ϕ such that ϕ(xi ) = yj . Thus, Aut(Kn,m ) ∼
= Sn × Sm .

Definition 3.1.3.
The double star is the tree with two adjacent non-leaf vertices x and
y such that x1 , x2 , ..., xn are the leafs adjacent to x and y1 , y2 , ..., ym
are the leafs adjacent to y. This graph is denoted Sn,m .

Example,
The double star S5,4 is,

Figure 3.2

43
Bibliography

1. Godsil, Chris. Algebraic Graph Theory, Springer (2001).

2. Biggs, Norman. Algebraic Graph Theory, Cambridge (1974).

3. David F Anderson and Philip S Livingston. The Zero Divisor

Graph of a commutative ring (1999).

44
DON BOSCO ARTS & SCIENCE
COLLEGE ANGADIKADAVU

DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

BILINEAR FORMS
 DON BOSCO ARTS & SCIENCE COLLEGE
ANGADIKADAVU
DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

BILINEAR FORMS

Dissertation submitted in the partial

Fulfillment of the requirement for the award of
MSc Degree in Mathematics of
Kannur University
Name : SWATHI SUDHAKARAN
Roll No. : C1PSMM1909

Examiners:
1.
2.
 KANNUR UNIVERSITY

BONAFIDE CERTIFICATE

Certified that this project report “BILINEAR FORMS” is the

bonafide work of SWATHI SUDHAKARAN who carried out the
project work under my supervision.

Signature of Head Signature of Supervisor

 DECLARATION

I, SWATHI SUDHAKARAN hereby declare that the Project work

entitled BILINEAR FORMS has been prepared by me and submitted
to Kannur University in partial fulfillment of requirement for the
award of Master of Science is a record of original work done by me
under the supervision of Mrs. RIYA BABY , Assistant Professor,
Department of Mathematics, Don Bosco Arts & Science College,
Angadikadavu. I, also declare that this Project work has not been
submitted by me fully for the award of any Degree, Diploma, Title
or recognition before any authority.

Place: Angadikadavu SWATHI SUDHAKARAN

Date: (C1PSMM1909)
 ACKNOWLEDGEMENT

Introduction is the proper place to begin. But first I bow my head

before the Almighty who is always with me. Also I must express
my deepest gratitude to people along the way. No words can ad-
equately express the sense of gratitude, still I try to express my
heartfelt thanks through words. The outset, I am deeply indebted
to my project supervisor Mrs. RIYA BABY Assistant Professor, De-
partment of Mathematics, Don Bosco Arts and Science College, An-
gadikadavu, for the invaluable guidance, loving encouragement and
meticulous care towards me throughout my career. I express my deep
sense of gratitude to all the faculty members of the Department of
Mathematics, Don Bosco Arts and Science College, Angadikadavu.
I can never forget the support and encouragement rendered by the
Principal and the Staff of Don Bosco Arts & Science College, An-
gadikadavu. I could not name many who sincerely supported and
helped for the successful completion of this Project. It is my plea-
sure and duty to thank each and everyone of them who walked with
me.

SWATHI SUDHAKARAN
 CONTENTS

1. INTRODUCTION .......................................................... 01

2. PRELIMINARY RESULTS ......................................... 03

3. BILINEAR FORMS ........................................................ 09

4. SYMMETRIC AND SKEW SYMMETRIC BILINEAR

FORMS .............................................................................. 27

5. GROUP PRESERVING BILINEAR FORMS...............38

6. CONCLUSION ................................................................. 42

7. BIBLIOGRAPHY ........................................................... 43
 INTRODUCTION

Linear algebra is that branch of mathematics which treats the

common properties of algebraic system which consists of a set, to-
gether with a resonable notation of a linear combination of elements
in the [Link] project is a brief survey of bilinear forms which is a
generalization of so called inner products on real or complex spaces
to any arbitrary field K.

Inner products are maps which are not completely linear, in the
sense that they are linear in its first argument and conjugate linear in
the second argument. So naturally a question arises is there any sim-
ilar map which includes inner products as well in its collection and
can be considered as a generalization of inner [Link] bilin-
ear form include certain types of inner products in its collection and
these are linear in both of its [Link] form has established
an important role through out many of the topics of mathematics
such as Linear Algebra and Coding [Link] forms also showed
importance in different fields like Engineering,Biology,Economics This
made me to read and study extra in bilinear forms.

The second chapter includes the basic definitions, notations and

examples which are needed to support the study of bilinear forms
and the other chapters deal with types of bilinear forms.

1
 The main aim of this project is to provide an introduction to
bilinear forms and some of its basis properties and characterisation
.

2
 CHAPTER 1

PRELIMINARY RESULTS

Vector space (1.1)

Let V be a non empty set with two operation addition and scalar
multiplication for each α, β ∈ V and F be a field of scalars such that

α + β ∈ V ∀ α, β ∈ V

α ∈ V, a ∈ F → aα ∈ V

Then V is called a vector space over the field F, if and only if the
following axioms are satisfied ∀ α, β ∈ V

1) Commutativity: α + β = β + α

2)Associativity: α + (β + γ) = (α + β) + γ

3)Existence of identity : There exist a unique vector 0 in V such

that

α+0=0+α = 0

4) Existence of inverse : There exist a unique vector -α ∈ V such

that

α + (−α) = 0 = (−α) + α

3
The element (-α) in V is called the additive inverse of α in V

5) k(α + β) = kα + kβ, k ∈ F

6) (k1 + k2 )α = k1 α + k2 α , k1 and k2 ∈ F

7) For any unit scalar 1 in F , 1 α = α

8) (k1 k2 )α = k1 (k2 α), k1 and k2 ∈ F

Linear combination of vector space(1.2)

Let V(F) be a vector space if a1 , a2 , ...an ∈ V then any element of

the form a1 α1 + a2 α2 + ... + an αn where α1 , α2 , ..., αn ∈ F is called a
linear combination of the vectors α1 , α2 , ..., αn .

Basis(1.3)

A basis for a vector space V is linearly independent set of vectors in

V which spans V

Dimension(1.4)

The number of elements in a basis set is called the dimension of the

vector space

Finite dimensional vector space(1.5)

A vector space V(F) is called a finite dimensional vector space , if

there exist a finite subsystem of the same which generate the whole
space

4
Linear transformation(1.6)

Let V and W be two vector space over the field F . A linear trans-
formation from V into W is a function T from V into W such that
, T(cx+y)= cT(x)+T(y) ∀ x, y ∈ V and c ∈ F

Vector space isomorphism (1.7)

Let U(F) and V(F) be two vector spaces over the same field F. Then
a mapping T: U → V is called an isomorphism from U to V ,if

1) T is one-one and onto

2) T(α + β) = T (α) + T (β), ∀ α, β ∈ U
3) T(aα) = aT (α)

condition 2 and 3 united to give

T(aα + bβ) = aT (α) + bT (β), ∀ α, β ∈ U and a,b ∈ F
Then U(F) and V(F) are said to be isomorphic vector spaces and
symbolically we write U(F) ∼
= V(F).

Transpose of a matrix(1.8)
h i
Let A = aij then the matrix obtained by interchanging rows
m×n
and columns of A is called the transpose of A and is denoted by At

Dual space and Dual basis(1.9)

If V is a vector space ,the collection of all linear functionals on V

forms a vector space and it is denoted by V ∗ , call it the dual space

5
of V ,
V ∗ = L(V,F)
The basis B ∗ = {f1 , f2 , ...fn } for V ∗ is called the dual basis of
B = {a1 , a2 , ..., an } the basis for V.

Sum of two vector space(1.10)

LetV and W be vector spaces over a field F then the sum V + W

is the space of all sums α + β such that α ∈ V and β ∈ W That is,

V + W = {α + β : α ∈ V, β ∈ W}

Direct sum(1.11)

Let V be a vector space over the field F and W1 and W2 be two

subspaces of V then V is said to be direct sum of W1 and W2 denoted
by V = W1 ⊕ W2 , if V = W1 + W2 and w1 ∩ w2 = {0}.

Group(1.12)

A group consists of the following

1) A set G ; a rule (operation) which associates with each pair of

elements X,Y in G an element XY in G

2)X(YZ) = (XY)Z, ∀ X,Y and Z ∈ G

3) There is an element e in G such that eX = Xe = e,∀ X ∈ G

4) To each element X in G there corresponds an element X −1 in G

such that

6
 XX −1 = X −1 X = e

Field(1.13)

A field is a non empty K along with function +:K×K → K and

· : K × K → K such that
1) (K,+) is an abelian group , that is

a) k1 + k2 = k2 + k1 ∀k1 , k2 ∈ K
b) k1 + (k2 + k3 ) = (k1 + k2 ) + k3 , ∀k1 , k2 , k3 ∈ K

c) There exist a unique element 0 called the zero element of K such

that

k+0 = 0+k ,∀ k ∈ K

d) To every k ∈ K , there corresponds a unique element -k ∈ K such

that

k+(-k) = 0 = (-k)+k

2) (K-{0},·) is an abelian group , that is

a) k1 · k2 = k2 · k1 , ∀ k1 , k2 ∈ K
b)k1 · (k2 · k3 ) = (k1 · k2 ) · k3 , ∀ k1 , k2 , k3 ∈ K

c) There exist a unique element 1 called the unit element of K such

that

k · 1 = k , for every scalar k ∈ K

d) To every k ∈ K , there corresponds a unique element k −1 ∈ K

such that

7
 k −1 · k = k ·k −1

3) · is distributive with respect to +, that is

k1 · (k2 + k3 ) = k1 · k2 + k1 .k3 ∀ k1 , k2 , k3 ∈ K

8
 CHAPTER 2

BILINEAR FORMS

Definition(2.1)

Let V be a vector space of a finite dimension over the field F, a

bilinear form on V is a mapping f:V×V → F ,which satisfies,

(1)f(aα1 + bα2 , β)=a f(α1 , β)+b f(α2 , β)

(2) f(α, aβ1 + bβ2 )=a f(α, β1 )+b f(α, β2 )

for every a,b ∈ F and α, β, α1 , β1 , α2 , β2 ∈ V(F)

We express condition 1 by saying f is linear in the first variable and
condition 2 by saying f is linear in the second variable.

Results

(1) Addition of bilinear forms

Suppose f1 and f2 are two bilinear forms on a finite dimensional

vector space V(F). The sum of two bilinear forms f1 and f2 are
f1 + f2 and is such that (f1 + f2 )(α, β) = f1 (α, β)+f2 (α, β)
∀α, β ∈ V(F)
For, it is easy to verify that f1 + f2 is also bilinear form on V
We have,
(f1 + f2 )(aα1 + bα2 , β) = f1 (aα1 + bα2 , β) + f2 (aα1 + bα2 , β)

9
 = [af1 (α1 , β)+bf1 (α2 , β)]+[af2 (α1 , β)+bf2 (α2 , β)]
(since f1 and f2 are bilinear )
= a[f1 (α1 , β)+f2 (α1 , β)]+b[f1 (α2 , β)+f2 (α2 , β)]
= a[(f1 + f2 )(α1 , β)] + b[(f1 + f2 )(α2 , β)]
Also
(f1 + f2 )(α, aβ1 + bβ2 ) = f1 (α, aβ1 + bβ2 ) + f2 (α, aβ1 + bβ2 )
= [af1 (α, β1 )+bf1 (α, β2 )]+[af2 (α, β1 )+bf2 (α, β2 )]
(since f1 and f2 are bilinear )
= a[f1 (α, β1 )+f2 (α, β1 )]+b[f1 (α, β2 )+f2 (α, β2 )]
= a[(f1 + f2 )(α, β1 )] + b[(f1 + f2 )(α, β2 )]
=⇒ f1 + f2 is a bilinear form on V

(2) Scalar multiplication of bilinear form

Let f and g are bilinear forms on V and c is a scalar , we define cf

as follows,

(cf)(α, β) = cf (α, β) ∀ α, β ∈ V

Thus,

(cf)(aα1 + bα2 , β) = cf (aα1 + bα2 , β)

= c[af(α1 , β) + bf (α2 , β)]

= a (cf)(α1 , β) + b(cf )(α2 , β)
Similarly,

(cf)(α, aβ1 + bβ2 ) = a(cf )(α, β1 ) + b(cf )(α, β2 )

10
Then cf is a bilinear [Link] have g is also a bilinear [Link]
fore cf+g is a bilinear form.
That is any linear combination of bilinear forms on V is again a
bilinear form.

NOTE

The set of all bilinear forms on V is closed under addition and scalar
multiplication.[follows from the above result]
The set of all bilinear forms on V(F) is generally denoted by L(V,V,F)

(3) Zero function is a bilinear form

A function f : V×V → F is said to be a zero function ,

if f(α, β) = 0 ∈ F ∀ α, β ∈ V such a function is denoted by 0̂ .

0̂(α, β) = 0 ∈ F ∀ α, β ∈ V

Now,
0̂(aα1 + bα2 , β)= 0 = 0+0
= a 0+b 0
= a 0̂(α1 , β) + b 0̂(α2 , β)
Also
0̂(α, aβ1 + bβ2 ) = 0 = 0+0
= a 0+b 0
= a 0̂(α, β1 )+ b 0̂(α, β2 )
∀a, b ∈ F and α, β, α1 , β1 , α2 , β2 ∈ V

11
 =⇒ 0̂ is a bilinear form

(4) If f is a bilinear form on V then -f is also a

bilinear form on V

Where ,

(-f)(α, β)= - [ f (α, β)]

We have,
-f (aα1 + bα2 , β)= - [f(aα1 + bα2 , β)]
= - [af(α1 , β) + bf (α2 , β)]
= -af (α1 , β)- bf (α2 , β)
= a(-f) (α1 , β)+b(-f)(α2 , β)
Similarly,

-f (α, aβ1 +bβ2 ) = a(-f)(α, β1 )+b(-f)(α, β2 )

=⇒ -f is a bilinear form on V

(5)Vector space of bilinear forms

Let B(W) denote the set of all bilinear form on W .

Then it can be easily shown that
B(W)={ f : f is a bilinear form on W } is a vector space over the
field F.
Note that , it is closed under addition and scalar multiplication.
That is ,
f+g(α, β) = f (α, β) + g(α, β) ∀ (α, β) ∈ V× V

12
 (cf)(α, β) = cf (α, β) ∀ (α, β) ∈ V× V and a ∈ F

Associative and commutative property follows from the fact that

the elements of the field F are both associative and commutative
in the field F for [Link] we can verify the laws of scalar
multiplication
For example,
a(f+g) = af + bg
Now , a(f+g)(α, β) = a[(f + g)(α, β)]
= a[f(α, β)+g(α, β)]
= af(α, β)+ag(α, β)
= (af)(α, β) + (ag)(α, β)
= (af+ag)(α, β)
Since above hold for all α, β ∈ W
a(f+g) = af+ag
Similarly we can prove that
(a+b) f = a f+b f
(ab) f = a (bf)
1f=f

Examples

(1) Let V(F) be a vector space over the field F , T be a linear

transformation on V and f a bilinear form on V. Suppose g is a
function from V × V into F given by , g(α, β) = f (Tα , Tβ ). Then
g is a bilinear form on V.

13
Solution :

We have ,
g(aα1 + bα2 , β) = f [T (aα1 + bα2 ), Tβ ]
= f(aTα1 +b Tα2 , Tβ )
= af(Tα1 , T β)+bf(Tα2 , T β)
= ag(α1 , β)+bg(α2 , β)
Also ,
g(α, aβ1 + bβ2 )= f[Tα , T (aβ1 + bβ2 )]
= f(Tα, aT β1 + bT β2 )
= af(Tα , Tβ1 ) + bf (Tα , Tβ2 )
= ag(α, β1 )+bg(α, β2 )

=⇒ g is a biinear form on V

2) Let L1 and L2 are linear functional on V(F).Prove that the

function f : V×V → F defined as f(α, β) = L1 (α)L2 (β) is a bilinear
form on V
Solution:

We have,
f(aα1 + bα2 , β) = L1 (aα1 + bα2 )L2 (β)
= [a L1 (α1 ) + bL1 (α2 )]L2 (β)
= a L1 (α1 )L2 (β) + bL1 (α2 )L2 (β)
= a f(α1 , β) + bf (α2 , β)
Also ,
f(α, aβ1 + bβ2 ) = L1 (α)L2 (aβ1 + bβ2 )

14
 = L1 (α)[(aL2 (β1 ) + bL2 (β2 )]
= a L1 (α)L2 (β1 ) + bL1 (α)L2 (β2 )
= a f(α, β1 ) + bf (α, β2 )
=⇒ f is a bilinear form on V

3) Let F be a field and let m and n be two integers . Let V be the

vector space of all m × n matrices over F , Let A be a fixed m × m
matrix over F and let fA be a function from V × V into F defined
by

fA (X, Y ) = tr(X t AY ) where X,Y ∈ V .

Then fA is a bilinear form on V

solution:

fA (aX + bZ, Y ) = tr[(aX + bZ)t AY)]

= tr [aX t AY ] + tr [bZ t AY]
= a tr[X t AY ] + b tr[Z t AY]
= a fA (X, Y )+ b fA (Z,Y)
[since trace function and transpose operation are linear]
Also,
fA (X, aY + bZ) = tr[X t , A(aY+bZ)]
= tr [aX t AY + bX t AZ]
= tr [aX t AY ] + tr [bX t AZ]
= a tr[X t AY ] + b tr[X t AZ]
= a fA (X, Y )+ b fA (X,Z)

There fore fA is a bilinear form on V .

15
4) Let V be a vector space of all ordered n tuples over the field F
, that is V = Vn (F) . If α = (a1 , a2 , ...an ) and β = (b1 , b2 ...bn ) be
any two elements of V and let f be a scalar valued function on V,
defined by ,

f(α, β) = a1 b1 + a2 b2 + ... + an bn

then f is a bilinear form on V

solution:
Let
α = (a1 , a2 , ..., an )

β = (b1 , b2 , ..., bn )
γ = (c1 , c2 , ..., cn )
Then,
k1 α + k2 γ = k1 (a1 , a2 , ...an ) + k2 (c1 , c2 ...cn )
= (k1 a1 , k1 a2 , ..., k1 an ) + (k2 c1 , k2 c2 , ..., k2 cn )
= (k1 a1 + k2 c1 , k1 a2 + k2 c2 , ..., k1 an + k2 cn )
k1 β + k2 γ = k1 (b1 , b2 , ...bn ) + k2 (c1 , c2 ...cn )
= (k1 b1 , k1 b2 , ..., k1 bn ) + (k2 c1 , k2 c2 , ..., k2 cn )
= (k1 b1 + k2 c1 , k1 b2 + k2 c2 , ...k1 bn + k2 cn )
Thus we have,

f(k1 α + k2 γ, β) = [k1 a1 + k2 c1 , k1 a2 + k2 c2 , ...k1 an + k2 cn , (b1 , b2 , ..., bn )]

= [(k1 a1 +k2 c1 )b1 +(k1 a2 +k2 c2 )b2 +...+(k1 an +k2 cn )bn ]
= k1 (a1 b1 +a2 b2 +...+an bn )+k2 (c1 b1 +c2 b2 +...+cn bn )
= k1 f (α, β) + k2 f (γ, β)

16
 =⇒ f (k1 α + k2 γ, β) = k1 f (α, β) + k2 f (γ, β)
Also,
f(α,k1 β +k2 γ) = [(a1 , a2 , ..., an )(k1 b1 +k2 c1 , k1 b2 +k2 c2 , ..., k1 bn +k2 cn )]
= a1 (k1 b1 +k2 c1 )+a2 (k1 b2 +k2 c2 )+...+an (k1 bn +k2 cn )
= k1 (a1 b1 +a2 b2 +...+an bn )+k2 (a1 c1 +a2 c2 +...+an cn )
= k1 f (α, β) + k2 f (α, γ)
=⇒ f (α, k1 β + k2 γ) = k1 f (α, β) + k2 f (α, γ)
=⇒ f is a bilinear form on V

Exercise

Which of the following functions f defined on vectors α = (x1 , x2 )

and β = (y1 , y2 ) in R2 are bilinear forms ?
1) f(α, β) = x1 y2 − x2 y1
2) f(α, β) = (x1 − y1 )2 + x2 y2

Solution:
Let α = (x1 , x2 )
β = (y1 , y2 )
γ = (z1 , z2 )
Then,
aα + bβ = a(x1 , x2 ) + b(y1 , y2 )
= (ax1 , ax2 ) + (by1 , by2 )
= (ax1 + by1 , ax2 + by2 )

1) f(α, γ) = f [(x1 , x2 ), (z1 , z2 )]

= x1 z2 − x2 z1

17
 f(β, γ) = f [(y1 , y2 ), (z1 , z2 )]
= y1 z2 − y2 z1
f(γ, α) = f [(z1 , z2 ), (x1 , x2 )]
= z1 x2 − z2 x1
f(γ, β) = f [(z1 , z2 ), (y1 , y2 )]
= z1 y2 − z2 y1

There fore, f(aα + bβ, γ) = f [a(x1 , x2 ) + b(y1 , y2 ), γ]

= f[(ax1 + by1 , ax2 + by2 )(z1 , z2 )]
= (ax1 + by1 )z2 − (ax2 + by2 )z1
= a(x1 z2 − x2 z1 ) + b(y1 z2 − y2 z1 )
= a f(α, γ) + bf (β, γ)
Also,
f(γ, aα + bβ) = f [(z1 , z2 ), a(x1 , x2 ) + b(y1 , y2 )]
= f[(z1 , z2 ), (ax1 +by1 , ax2 +by2 )]
= z1 (ax2 + by2 ) − z2 (ax1 + by1 )
= a(z1 x2 − z2 x1 ) + b(z1 y2 − z2 y1 )
= a f(γ, α) + bf (γ, β)
=⇒ f is a bilinear form on R2

2) f(α, γ) = f [(x1 , x2 ), (z1 , z2 )]

= (x1 − z1 )2 +x2 z2
f(β, γ) = (y1 − z1 )2 + y2 z2

f(a α+bβ, γ) = f[(ax1 + by2 , ax2 + by2 ), (z1 , z2 )]

= (ax1 + by2 − z1 )2 + (ax2 + by2 )z2 and

18
 af(α, γ) + bf (β, γ) = a(x1 − z1 )2 + ax2 z2 + b(y1 − z1 )2 + by2 z2
= a(x1 − z1 )2 + b (y1 − z1 )2 + (ax2 +by2 )z2
That is f(aα + bβ, γ) ̸= af (α, γ) + bf (β, γ)
=⇒ f is not a bilinear form on R2

Definition(2.2)

Let B = { α1 , α2 , ..., αn } be an ordered basis for a finite dimensional

vector space V(F) and f be a bilinear form on V then the matrix of
h i
f with respect to basis B that is [f ]B is the matrix A = ai,j such
n×n
that
A = f(αi , αj ) = ai,j [i=1,2,3,...,n ; j=1,2,3,...,n]
Pn
Let α = i=1 xi αi

 
x
 1
 x2 
 
Therefore coordinate matrix of α = X =  
 .. 
.
 
xn

h i
t
Therefore X = x1 , x2 , . . . , xn

Pn
β= j=1 yj αj

 
y
 1
 y2 
 
There fore coordinate matrix of β = Y =  
 .. 
.
 
yn

19
Now,

f(α, β) = f ( ni=1 xi αi , nj=1 yj αj )

P P

Pn Pn
= i=1 j=1 xi yj f (αi , αj )

 
y
 1
ih i  y2 
h 
= x1 , x2 , . . . , xn ai,j  
 .. 
.
 
yn
= X t AY
h it h i
= α A β
B B

Theorem(2.3)

Let V be a finite dimensional vector space over the field F. for each
ordered basis B of V , the function which associates with each bilin-
ear form on V it’s matrix in the ordered basis B is an isomorphism
of the space L(V,V,F) onto the space of n × n matrices over the
field F .

Proof:

Let S be the vector space of all n × n matrices over the field F and
let ϕ be a function from L(V,V,F) into S
h i
ϕ(f ) = f ∀f ∈ L(V,V,F)
B

20
Let B = { α1 , α2 , ..., αn } be the basis of V .
Then for a,b ∈ F and f,g ∈ L(V,V,F) we have ,
h i
ϕ(af + bg) = af + bg
B
Also,
(af+bg) (αi , αj ) = (af)(αi , αj ) + (bg)(αi , αj )
= af(αi , αj ) + bg(αi , αj )
(i = 1,2, ...,n ; j = 1,2,...,n)
h i h i h i
=⇒ af + bg = a f + b g
B B B

=⇒ ϕ(af + bg) = aϕ(f ) + bϕ(g)

=⇒ ϕ is a linear transformation.
To prove ϕ is one − one
Let f,g ∈ L(V,V,F)
Then,
ϕ(f ) = ϕ(g)
h i h i
=⇒ f = g
B B

=⇒ f = g
=⇒ ϕ is one-one
Next to prove that ϕ is onto
h i
Let A = ai,j be any n × n matrix over F , then there exist a linear
h i
form f on V such that f =A
B

That is ϕ(f) = A

=⇒ ϕ is onto
Hence ϕ is an isomorphism of L(V,V,F) onto the space of all n × n

21
matrices over F .

Corollory(2.4)

If B = { α1 , α2 , ..., αn } is an ordered basis for V and

B ∗ = {L1 , L2 , ..., Ln } is the dual basis for V∗ , then the n2 bilinear
form fi,j (α, β) = Li (α)Lj (β), 1 ≤ i ≤ n ; 1 ≤ j ≤ n form a basis for
the space L(V,V,F). In particular the dimension of L(V,V,F) is n2
.

Proof:

The dual basis defined by the fact that Li (α) is the ith coordinates
of α in the ordered basis B . Also fij defined by
fij (α, β) = Li (α)Lj (β) .
We have to prove that this is a bilinear form .
fi,j (aα1 + bα2 , β) = Li (aα1 + bα2 )Lj (β)
= [aLi (α1 ) + bLi (α2 )]Lj (β)
= a Li (α1 )Lj (β) + bLi (α2 )Lj (β)
= a fi,j (α1 , β) + bfi,j (α2 , β)
Also
fi,j (α, aβ1 + bβ2 ) = Li (α)Lj (aβ1 + bβ2 )
= Li (α)[aLj (β1 ) + bLj (β2 )]
= a Li (α)Lj (β1 ) + bLi (α)Lj (β2 )
= a fi,j (α, β1 ) + bfi,j (α, β2 )
=⇒ f is a bilinear form .
If α = x1 α1 + x2 α2 + · · · + xn αn

22
 β = y1 α 1 + y2 α 2 + · · · + yn α n
Then fi,j (α, β) = xi yj
Let f be any bilinear form V and let A be the matrix of f in the
ordered basis B
Then,
P
f(α, β) = i,j Ai,J xi yj
P
=⇒ f = i,j Ai,J fi,j
Therefore n2 forms fij comprise a basis for L(V,V,F)
That is dim( L(V,V,F) = n2

Theorem(2.5)

Let B1 = {α1 , α2 , ..., αn } and B2 = {β1 , β2 , ..., βn } be two ordered

basis of a finite dimensional vector space V(F) and f is a bilinear
form on V , then there exist an invertible n × n matrix C over the
field F such that
h i h i
t
f =C f C , where C t denotes the transpose of matrix C
B2 B

Example

Let f be a bilinear form on R2 defined by f((x1 , y1 ), (x2 , y2 )) = x1 x2 +

y1 y2 . Find the matrix of f for the following basis.
B1 = {(1, 0)(0, 1)} = {α1 , α2 }
B2 = {(1, −1)(1, 1)} = {β1 , β2 }
Also find the transition matrix C from B1 to B2 and verify that
h i h i
t
f =C f C
B2 B

23
Solution:
h i h i
let f = ai,j
B1 B1

So that f(αi , αj ) = ai,j

a1,1 = f (α1 , α1 ) = f((1,0)(1,0)) = 1+0 = 1
a1,2 = f (α1 , α2 ) = f((1,0)(0,1)) = 0+0 = 0
a2,1 = f (α2 , α1 ) = f((0,1)(1,0)) = 0+0 = 0
a2,2 = f (α2 , α2 ) = f((0,1)(0,1)) = 0+1 = 1

Therefore
   
h i a1,1 a1,2 1 0
f = = 
B1 a2,1 a2,2 0 1

h i h i
let f = bi,j
B2 B2

So that f(βi , βj ) = bij

b1,1 = f (β1 , β1 ) = f((1,-1)(1,-1)) = 1+1 = 2
b1,2 = f (β1 , β2 ) = f((1,-1)(1,1)) = 0+0 = 0
b2,1 = f (β2 , β1 ) = f((1,1)(1,-1)) = 1-1 = 0
b2,2 = f (β2 , β2 ) = f((1,1)(1,1)) = 1+1 = 2

There fore
 
h i 2 0
f = 
B2 0 2

24
Let T be a linear operator on V such that T(αi ) = βi
h i
i = 1, 2, . . . and Cij be the matrix of T relative to basis B .
Now,
T(αi ) = βi [by definition]
∴ T(1,0) = (1,-1) =(1,0) + (-1) (0,1)

Or

T(α1 ) = α1 − α2
T(0,1) = (1,1) =(1,0) + (0,1)

Or

T(α2 ) = α1 − α2
Then,
 
1 1
C= 
−1 1
 
1 −1
=⇒ C t =  
1 1

  
t
h i 1 −1 1 0 1 1
=⇒ C f C=   
B1 1 1 0 1 −1 1
 
2 0
= 
0 2
h i
= f
B2

25
 h i h i
t
=⇒ f =C f C
B2 B1

Definition(2.6)

The rank of a bilinear form f on V , written rank(f) is defined to be

the rank of any matrix representation .

Definition(2.7)

A bilinear form f on a vector space V is said to be degenerate , if for

all non zero α in V , f(α, β) = 0 for every β in V and for all nonzero
β in V, f(α, β) = 0 for every α in V
A bilinear form is clled non degenerate if it is not degenerate .
That is,
if for all non zero α in V there is a β in V such that f(α, β) ̸= 0 and
for all non zero β in V there is a α in V such that
f(α, β) ̸= 0

26
 CHAPTER 3

SKEW SYMMETRIC AND

SYMMETRIC BILINEAR
FORMS

3.1 SYMMETRIC BILINEAR FORMS

Definition(3.1.1)

A bilinear form on the vector space V is said to be symmetric if,

f(α, β) = f (β, α) ∀ α, β ∈ V

Result

If V is a finite dimensional vector space then the bilinear form f on

V is symmetric if and only if its matrix A in some ordered basis is
symmetric
That is,

At = A

27
 Proof : Let the matrices of α, β with respect to a certain basis B
h i
be X and Y and A = f
B
Now,
h i
t
f(α, β) = X f Y
B

= X t AY
h i
t
f(β, α) = Y f X
B

=Y t AX
Now, f will be a symmetric bilinear form if and only if f(α, β) =
f (β, α)
∀ α, β ∈ V Or X t AY = Y t AX ∀ matrices X and Y .
Now X t AY is a 1 × 1 matrix and as such it is equal to its transpose
.
t
X t AY = Y t AX
t
⇒ Y t At (X t ) = Y t AX
⇒ Y t At X = Y t AX
t
(since (X t ) = X)
⇒ At = A
That is A is symmetric
Hence bilinear form will be symmetric if and only if its matrix is
symmetric .

28
 Theorem(3.1.2)

Let V be a vector space over the field F , whose characteristic is not

2, that is 1+1 ̸= 0 , then every symmetric bilinear form is uniquely
determined by the curresponding quadratic form .
proof:
Let f be any bilinear form on V and let q be the corresponding
quadratic form associated with f , then for all α, β ∈ V
We have ,
q(α + β) = f (α + β, α + β)
= f(α, α + β) + f (β, α + β)
= f(α, α) + f (α, β) + f (β, α) + f (β, β)
= q(α) + f (α, β) + f (α, β) + q(β)
(since f(α, β) = f (β, α) , that is symmetric)
= q(α) + (1 + 1)f (α, β) + q(β)
(1+1) f(α, β) = q(α + β) − q(α) − q(β).......(1)
Since 1+1 ̸= = 0 , f(α, β) is uniquely determined by the correspond-
ing quadractic form.

Further extension

Let F be a subfield of the complex numbers ,then ......(1) yields the

polarization identity
That is,

1 1
f(α, β) = 4 q(α + β) - 4 q(α − β)

29
From above theorem we have ,
2f(α, β) = q(α + β) − q(α) − q(β).......(1)
Also

q(α − β) = f (α − β, α − β)
= f(α, α − β) − f (β, α − β)
= f(α.α) − f (α, β) − f (β, α) + f (β, β)
= q(α) + q(β) − 2f (α, β)
(since f is symmetric)
2 f (α, β) = q(α) + q(β) − q(α − β) .......(2)
Adding .......(1) and ......(2) we get,
4 f (α, β) = q(α + β) − q(α − β)
1 1
⇒ f (α, β) = 4 q(α + β) - 4 q(α − β)

Theorem (3.1.3)

Let V be a finite dimensional vector space over a field of character-

istic zero and let f be a symmetric bilinear form on V , then there is
an ordered basis for V in which f is represented by a diagonal matrix
.
proof:
We want to prove that there is a basis B = {α1 , α2 , ..., αn } for V
h i
such that f(αi , αj ) = aij = 0 for i ̸= j
If dim V = n = 1 or f = 0 then the theorem is true
We may suppose f ̸= 0 and dim V = n > 1
If q is the quadratic form associated with f and if f(α, α) = 0

30
for each α ∈ V
Then q(α) = 0
Thus by polarization identity ,

f(α, β) = 41 q(α + β) − 14 q(α − β)

We get f(α, β) = 0 ∀ α, β ∈ V
Thus f = 0 which controdicts our assumption and hence there must
exist a vector α ∈ V such that f(α, α) = q(α) ̸= 0
Let U be the one dimensional subspace of V which is spanned by
vectors α and W be the set of all vectors β such that f(α, β) = 0
Then W is a subspace of V

claim: V = U ⊕ W

1) To show that U ∩ W = {0}

Let δ ∈ U ∩ W
⇒ δ ∈ U and δ ∈ W
δ ∈ U ⇒ δ = kα for some scalar k
Also,
δ ∈ W ⇒ 0 = f (α, δ) = f (α, kα) = kf (α, α)
But f(α, α) ̸= 0
Therefore,
k =0 and so δ = kα=0
⇒ U ∩ W = {0}
2) To show that V = U+W
Let γ ∈ V

31
 f (γ,α)
Put β = γ − f (α,α) α

We have,
f (γ,α)
f(α, β) = f (α, γ − f (α,α) α)

f (γ,α)
= f(α, γ) − f (α,α) f (α, α)

= f(α, γ) − f (γ, α)
=0
Hence β ∈ W
f (γ,α)
Thus γ = f (α,α) α +β

That is γ ∈ U+W
Thus by (1) ans (2) we get V = U ⊕ W
⇒ dimW = dimV − dimU
= n -1
Let h be the restriction of f from V to W . It is bilinear form on W
and dim W = n-1 < n
Thus by mathematical induction we assume there exist a basis {α2 , α3 , ...αn }
of W such that f(αi , αj ) = 0 , i ̸= j (i≥ 2,j≥ 2)
Again {α1 } is the basis of U and we have V = U ⊕ W
Therefore it follows that {α1 , α2 , ...αn } is the basis of V such that
f(αi , αj ) = 0 , i ̸= j
That is the matrix of f with respect to the basis B is a diagnal ma-
trix.

Corollory(3.1.4)

Let F be a subfield of complex numbers and let A be an n×n sym-

32
metric matricx over F then there is an invertible n×n matrix C over
F such that C t AC is diagonal
Proof:
Let V be a finite dimensional vector space and let X = {α1 , α2 , ..., αn }
be an ordered basis of V
h i
Then f = A , where f is bilinear form on V
X

But as A is symmetric matrix so the bilinear form f is also symmet-

ric.
Therefore by preceding theorem there exist an ordered basis Y of V
h i
such that f = C t AC
Y

⇒ C t AC is also a diagonal matrix

Hence the proof

33
 3.2 SKEW SYMMETRIC BILINEAR FORMS

Definition(3.2.1)

If f is a bilinear form on the vector space V then it is said to be skew

symmetric if,

f(α, β) = −f (β, α) ∀ α, β ∈ V

Theorem(3.2.2)

Every bilinear form on a vector space V over a subfield F of the com-

plex numbers can be uniquely expressed as the sum of a symmetric
and skew symmetric bilinear forms
Proof:
Let f be a bilinear form on a vector space V
Let
f1 (α, β) = 1/2[f (α, β) + f (β, α)] ......(1)
f2 (α, β) = 1/2[f (α, β) − f (β, α)]......(2)
It is easy to prove that both f1 and f2 are bilinear forms on V
Further we also have ,
1
f1 (β, α) = 2 [ f(β, α) + f (α, β)]
= f1 (α, β)
⇒ f1 is symmetric
Then
1
f2 (β, α) = 2 [ f(β, α) − f (α, β)]

34
 = - 1/2 [ f(α, β) − f (β, α)]
= -f2 (α, β)
⇒ f2 is skew symmetric bilinear form
From (1) and (2) we have
f1 (α, β) + f2 (α, β) = f (α, β)
⇒ (f1 + f2 )(α, β) = f (α, β) ∀ α, β ∈V
⇒ f1 + f2 = f
Uniqueness:
Further suppose that f = f3 + f4 where f3 is symmetric and f4 is
skew symmetric
We have,

f (α, β) = (f3 + f4 )(α, β)

= f3 (α, β) + f4 (α, β) .......(3)

Also,

f (β, α) = (f3 + f4 )(β, α)

= f3 (α, β) − f4 (α, β) .......(4)

(since f3 is symmetric and f4 is skew symmetric)

Hence adding (3) and (4) we get ,

f (α, β) + f (β, α) = 2f3 (α, β)

⇒ f3 (α, β) = 12 [f (α, β) + f (β, α)]
= f1 (α, β)
⇒ f3 = f1

35
Substracting (4) from (3) we also get ,

f (α, β) − f (β, α) = 2f4 (α, β)

⇒ f4 (α, β) = 12 [f (α, β) − f (β, α)]

= f2 (α, β)
⇒ f4 = f2
Hence the solution f = f1 + f2 is unique

Theorem (3.2.3)

Let V be a finite dimensional vector space then the bilinear form f

on V is skew symmetric if and only if its matrix A in some ordered
basis is skew symmetric
That is At = -A
Proof:
Let B be an ordered basis for V and let α, β be any two vectors in
V
Then f(α, β) = X t AY and f(β, α) = Y t AX , where f is a bilinear
form on V and X and Y are co-ordinate matrices of α and β
Also A is the matrix of f in the ordered basis B
Now f will skew symmetric if and only if X t AY =-Y t AX for all
column matrix X and Y
But X t AY is 1×1 matrix
Therefore,
X t AY = (X t AY )t
= Y t At (X t )t

36
 = Y t At X
The bilinear form f will skew symmetric if and only if Y t At X = -
Y t At X forall column matrices X and Y
That is At = -A
That is A is skew symmetric

37
 CHAPTER 4

GROUP PRESERVING
BILINEAR FORMS

Definition(4.1)

Let f be a bilinear form on V and let T be a bilinear operator on the

same vector space V over the field F then we say that T preserves f
if f(Tα , Tβ ) = f (α, β) ∀α, β ∈ V

Remarks

1)The identity operator I preserves every bilinear form

For , f(Iα, Iβ) = f (α, β), ∀α, β ∈ V, f being any bilinear form
2)Let S and T be linear operators which preserve f , then the product
ST also preserve f
For f(STα , STβ ) = f (Tα , Tβ ) (since S preserves f )

= f(α, β) (since T preserves f)

⇒ ST preserves f

3) Let G be the set of all linear operators on V which preserve f then

we say that G is closed with respect to the operation of product of
two bilinear operators .

38
 Theorem(4.2)

Let f be a non-degenerating bilinear form on a finite dimensional

vector space V then the set G of all bilinear operators on V which
preserve f is a group under the operation of composition.
Proof:
Let G = { T : T preserves f }
Let T and S be two linear operators in G which preserve bilinear
form
Therefore

f(Tα , Tβ ) = f (α, β)
f(Sα , Sβ ) = f (α, β)

Closure property:

f(T Sα , T Sβ ) = f (Sα , Sβ ) (since T preserves f)

= f(α, β) (since S preserves f)

Therefore TS also preserves [Link] TS ∈ G

Therefore G is closed for the given composition.

Associative:
For each T,S,P ∈ G we have,

(T(SP))(α) = T (SP )(α)

= T[S(P(α))]
= (TS)(P(α))
= (TS)P(α)
⇒ T(SP) = (TS)P

39
Hence it is associative.

Identity :

f(Iα, Iβ) = f (α, β) (since Iα = α and Iβ = β)

Therefore the identity operator I also preserves f and as such belong
to G .
Hence G has an identity element.

Inverse:

Let T ∈ G so that,
f(T α, T β) = f(α, β), ∀α, β ∈ V
We have to show that T is invertible and T −1 ∈ G
Since V is finite dimensional it will be sufficient to show that T is
nonsingular
That is

Tα = 0
⇒α=0

Choose a vector α in the nullspace of T , so that T α = 0

Then for any vector β in V
We have,

f(α, β) = f (T α, T β) [ since T preserves f ]

= f(0,T β )
=0

40
Now f(α, β) = 0, ∀β ∈ V and f is non degenerate .That is for all non
zero α there exist β ∈ V such that f(α, β) ̸= 0
Therefore

α=0

Thus

Tα = 0
⇒α=0
⇒ T is non singular
⇒ T is invertible and T −1 be inverse of T

So T T −1 = T −1 T = I Now,we have shown the existence of T −1 and

we have to show that T −1 is in G
That is , it preserves f
f (T −1 (α), T −1 (β)) = f (T T −1 α, T T −1 β) (since T preserves f)

= f(I α, Iβ)
= f(α, β)
Which shows that T −1 also preserves f.
That is T −1 ∈ G
Thus inverse of each element of G exists
Hence G is a group.

41
 CONCLUSION

By doing this project we learned about bilinear forms on finite di-

mensional vector spaces. In the first section we treats the space of
bilinear forms on a vector space of dimension n .The matrix of bi-
linear form in an ordered basis is introduced, and the isomorphism
between the space of forms and the space of n × n matrices is es-
[Link] rank of bilinear form is defined,and non- degenerate
bilinear forms are introduced. The second section discusses sym-
metric bilinear forms and skew-symmetric bilinear forms. The third
section discusses the group preserving a bilinear forms.

42
 REFERENCE

1. Linear Algebra − Gupta Sharma

2. Linear Algebra − Kenneth Hoffman
3. Linear Algebra − M.L. Khanna

43
DON BOSCO ARTS & SCIENCE
COLLEGE ANGADIKADAVU

DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

FUZZY ALGEBRA
 DON BOSCO ARTS & SCIENCE COLLEGE
ANGADIKADAVU
DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

FUZZY ALGEBRA

Dissertation submitted in the partial

Fulfillment of the requirement for the award of
MSc Degree in Mathematics of
Kannur University
Name : ANJU JAYARAJ
Roll No. : C1PSMM1904

Examiners:
1.
2.
 KANNUR UNIVERSITY

BONAFIDE CERTIFICATE

Certified that this project report “FUZZY ALGEBRA” is the

bonafide work of ANJU JAYARAJ who carried out the project work
under my supervision.

Signature of Head Signature of Supervisor

 DECLARATION

I, ANJU JAYARAJ hereby declare that the Project work entitled

FUZZY ALGEBRA has been prepared by me and submitted to Kan-
nur University in partial fulfillment of requirement for the award of
Master of Science is a record of original work done by me under the
supervision of Mrs. RIYA BABY Assistant Professor, Department
of Mathematics, Don Bosco Arts & Science College, Angadikadavu.
I, also declare that this Project work has not been submitted by me
fully for the award of any Degree, Diploma, Title or recognition be-
fore any authority.

Place: Angadikadavu ANJU JAYARAJ

Date: (C1PSMM1904)
 ACKNOWLEDGEMENT

Introduction is the proper place to begin. But first I bow my head

before the Almighty who is always with me. Also I must express
my deepest gratitude to people along the way. No words can ad-
equately express the sense of gratitude, still I try to express my
heartfelt thanks through words. The outset, I am deeply indebted
to my project supervisor Mrs. RIYA BABY Assistant Professor, De-
partment of Mathematics, Don Bosco Arts and Science College, An-
gadikadavu, for the invaluable guidance, loving encouragement and
meticulous care towards me throughout my career. I express my deep
sense of gratitude to all the faculty members of the Department of
Mathematics, Don Bosco Arts and Science College, Angadikadavu.
I can never forget the support and encouragement rendered by the
Principal and the Staff of Don Bosco Arts & Science College, An-
gadikadavu. I could not name many who sincerely supported and
helped for the successful completion of this Project. It is my plea-
sure and duty to thank each and everyone of them who walked with
me.

ANJU JAYARAJ
 CONTENTS

1. INTRODUCTION .......................................................... 01

2. PRELIMINARIES .......................................................... 03

3. FUZZY RELATION ....................................................... 07

3.1 Fuzzy Equivalence........................................................ 09

4. FUZZY GROUPS............................................................. 13
4.1 Normal fuzzy subgroups and Cosets ............................ 23

5. FUZZY SUBRINGS ........................................................ 36

5.1 Fuzzy Ideals.................................................................. 40

6. CONCLUSION ................................................................. 43

7. BIBLIOGRAPHY ........................................................... 44
 INTRODUCTION
The concept of sets and functions to represent problem were defined
in the 20th century. This way of representing problems is more
rigid. In many circumstances the solutions using this concepts are
meaningless. This difficulty was overcome by the introduction of
fuzzy concept.
In 1965, L.A Zadeh mathematically formulated the fuzzy subset
concept. He introduced the notion of a fuzzy subset of a set as a
method for representing uncertainity. He defined the fuzzy subset of
a non-empty set as a collection of objects with grade of membership
in a continuum, with each object being assigned a value between
0 and 1 by a membership function. Fuzzy set theory was guided
by the assumption that classical sets were not natural, appropriate
or useful notion in describing the real life problems, because every
object encountered in this real physical world carries some degree
of fuzziness. Further the concept of grade of membership is not a
probabilistic concept.
As know to us, a relation is a subset of the Cartesian product of
two sets. A relation is fuzzified while a subset is fuzzified. Infact
whether two objects have a relation is not always easy to determine.
For example, the relation ”greater than” on the set of real numbers
is a crisp one because we can determine the order relation of any
two real numbers without vagueness. However the relation ”much
greater than” is a fuzzy one because it is impossible for us to figure

1
out the exact minimum difference of two numbers satisfying this
relation. In real world problems, there exists a lot of such relations.
For example, ”being friend of” and ”being confident in” between
some people. These relations will be termed as fuzzy relations.
After the concept of fuzzy sets first introduced by Zadeh, the
theory of fuzzy mathematics has been widely used in Mathematics
and many more areas.
In 1971, Rosenfeld introduced the concept of fuzzy subgroup,
which spreads the area of fuzzy algebra. Fuzzy algebra plays an
important role in the field of computer, such as fuzzy codes, fuzzy fi-
nite state machines, regular fuzzy languages and codes and so on, so
it is necessary for us to do further research on fuzzy algebra theory.
In 1981, W.J Liu introduced the concept of fuzzy ideals of a ring.
It was followed by the studies of Mukherjee and Sen who defined
and examined fuzzy prime ideals of a ring. Fuzzy ideals were fur-
ther investigated by Malik and Mordeson and they gave complete
characterization of fuzzy prime ideals of an arbitrary ring.

2
 PRELIMINARIES

Definition 1.0.1
Let X be a nonempty set. A fuzzy set A in X is characterised by
membership function µA : X → [0, 1] and µA (x) is interpreted as a
degree of membership of element x in fuzzy set A, for each x ∈ X.
It is clear that A is completely determined by the set of tuple

A = {(x, µA (x) : x ∈ X)}

i.e., A mapping A : X → [0, 1] is called a fuzzy set on X. The

value A(x) of A at x ∈ X stands for the degree of membership of x
in A.

• The set of all fuzzy sets on X will be denoted by F (X).A(x) = 1

means full membership, A(x) = 0 means non-membership and
intermediate values between 0 and 1 means partial membership.

A(x) is referred to as a membership function as x varies in X.

Proposition 1.0.2
Let X be a non empty set. Then there exists an isomorphism be-
tween (P (X), ∩, ∪c ) and (Ch(X), ∨, ∧,c ), where P (X) is the power-
set of X and Ch(X) is the set of two valued characteristic function
on X. - It follows from the above proposition that every subset of
X may be regarded as a mapping from X to {0, 1}

3
 In this sense an ordinary set is also a fuzzy set whose member-
ship function is just its characteristic function. Accordingly we shall
identify the membership degree A(x) with the value χA (x) of the
characteristic function χA at x when A is an ordinary set.
For the two extreme cases ∅ and X, the membership functions
are defined by

∀x ∈ X, ∅(x) = 0 and X(x) = 1

• In contrast with fuzzy sets, ordinary sets are sometimes termed

as crisp sets.

Example 1.0.3
A realter wants to classify the house he offer to his client. One indi-
cator of comfort of these house is the number of bedrooms in it. Let
X = {1, 2, . . . , 10} be the set of available types of houses described
by x = the number of bedrooms in house. Then the fuzzy set com-
fortable type of house for a four person family may be described as
A = {(1, 0.2), (2, 0.5)(3, 0.8)(4, 1)(5, 0.7)(6, 0.3)}.
In the set of ordered pair, the first element denote element and
second element denote the degree of membership.

• If ∀x ∈ X, A(x) ≤ B(x), then we call that A is a subset of B or

A is contained in B, denoted by A ⊆ B.

If ∀x ∈ X, A(x) = B(x), then A and B are called equal, denoted

by A = B. Obviously, A = B if and only if A ⊆ B and B ⊆ A.

4
 If A ̸= ∅, A ⊆ B and ∃x ∈ X such that A(x) < B(x), then we
say that A is properly contained in B, denoted by A ⊂ B, where
A, B ∈ F (X).

Proposition 1.0.4
∀A, B, C, D ∈ F (X)

1. A∩B ⊆ A and A ⊆ A∪B 2. A ⊆ B ⇔ A∪B = B ⇔ A∩B = A

2. A ⊆ B and C ⊆ D ⇒ A ∪ C ⊆ B ∪ D and A ∩ C ⊆ B ∩ D

3. A ⊆ B ⇒ B c ⊆ Ac

• (F (X), ∩, ∪, c ) is a soft algebra, i.e., F (X) satisfies ∀A, B, C, D ∈

F (X)

1. idempotency : A ∪ A = A, A ∩ A = A

2. commutativity : A ∪ B = B ∪ A, A ∩ B = B ∩ A

3. associativity : (A∪B)∪C = A∪(B∪C), (A∩B)∩C = A∩(B∩C)

4. absorption law : A ∪ (A ∩ B) = A, A ∩ (A ∪ B) = A

5. distributivity : A ∪ (B ∩ C) = (A ∪ B) ∩ (A ∪ C)

A ∩ (B ∪ C) = (A ∩ B) ∪ (A ∩ C)

6. the existence of the greatest and least elements : ∅ ⊆ A ⊆ X

7. involution : (Ac )c = A

5
 8. De- Morgan’s law : (A ∪ B)c = Ac ∩ B c , (A ∩ B)c = Ac ∪ B c

From the above property (F (X), ∪, ∩, c ) are largely dependent on

properties of ([0, 1], max, minc c ) = ([0, 1], ∨, ∧,c )

• The set {x | A(x) = 1} is said to be the kernel of A, Ker A.

• The set {x | A(x) > 0} is called the support of A, denoted by

supp(A).

• The number
W
x∈X A(x) is called the height of A, hgt(A).

• The number
V
x∈X A(x) is referred to as the plinth of A, denoted
by plt(A).

• If Ker(A) = ∅, then A is called a fuzzy set.

Definition 1.0.5
Let A be a fuzzy set on X. For α ∈ [0, 1], the α cut Aα of A is
defined as Aα = {x | A(x) ≥ α}, and the strong α cut Âα of A is
defined as Âα = {x | A(x) > α}

1. A ∩ B ⊆ A and A ⊆ A ∪ B

6
Chapter 1
FUZZY RELATION

Definition 2.0.1
Let X and Y be two non-empty sets. A mapping R : X× Y →
[0, 1]] is called a fuzzy(binary) relation from X to Y . For (x, y) ∈
X × Y, R(x, y) ∈ [0, 1] is referred to as the degree of relationship
between x and y.
Particularly, a fuzzy relation from X to X is called a fuzzy(binary)
relation on X
By definition, a fuzzy relation R is a fuzzy set on X × Y , i.e.,
R ∈ F (X × Y )

Definition 2.0.2
Let R be a fuzzy relation from X to Y . The R-afterset xR of x(x ∈
X) is a fuzzy set on Y defined by, ∀y ∈ Y, (xR)(y) = R(x, y).
The R-foreset Ry of y(y ∈ Y ) is a fuzzy set on X defined by,
∀x ∈ X, (Ry)(x) = R(x, y)
Since fuzzy relations are fuzzy sets, they have the same set-theoretic
operations as fuzzy sets.
Let R and S be fuzzy relations from X to Y.R is contained in S,
denoted R ⊆ S if and only if ∀(x, y) ∈ X × Y, R(x, y) ≤ S(x, y); R is
equal to S, denoted R = S, if and only if ∀(x, y) ∈ X × Y, R(x, y) =

7
S(x, y). Clearly, R = S if and only if R ⊆ S and S ⊆ R The union
R ∪ S ∈ F (X × Y ) of R and S is defined by, ∀(x, y) ∈ X × Y ,

(R ∪ S)(x, y) = R(x, y) ∨ S(x, y)

The intersection R ∩ S ∈ F (X × Y ) of R and S is defined by

∀(x, y) ∈ X × Y ,

(R ∩ S)(x, y) = R(x, y) ∧ S(x, y)

The complement Rc ∈ F (X × Y ) of R is defined by ∀(x, y) ∈

X ×Y,

(Rc ) (x, y) = 1 − R(x, y)

The inverse R−1 ∈ F (X × Y ) of R is defined by ∀(x, y) ∈ X × Y ,

R−1 (y, x) = R(x, y)

S
In addition, if Ri ∈ F (X × Y ) for i ∈ I indexing set, then i∈I Ri
is defined by, ∀(x, y) ∈ X × Y

!
[ _
Ri (x, y) = Ri (x, y)
i∈I i∈I
T T
and i∈I Ri is defined by ∀(x, y) ∈ X × Y, i∈I (Ri ) (x, y) =
∧i∈i Ri (x, y)

8
 Proposition 2.0.8 (R ∪ S)−1 = R−1 ∪ S −1

Proposition 2.0.9 (R ∩ S)−1 = R−1 ∩ S −1

c
Proposition 2.0.10 (Rc )−1 = R−1

A fuzzy relation also has the concept of (strong) α cut. The crisp
relation, Rα = {(x, y) | R(x, y) ≥ α} for α ∈ [0, 1] will be called the
α-cut relation of R. Rα = {(x, y) | R(x, y) > α} for α ∈ [0, 1] will
be called the strong α-cut relation of R.

Fuzzy Equivalence

Definition 2.1.1
If R(x, x) = 1, ∀x ∈ X, then R is called a reflexive(fuzzy) relation.

Definition 2.1.2
R is reflexive if and only if ∀α ∈ [0, 1], Rα is reflexive.
Proof. If R is reflexive, then ∀α ∈ [0, 1], R(x, x) = 1 ≥ α. Hence
(x, x) ∈ Rα . Thus Rα is reflexive.
Conversely, assume that ∀α ∈ [0, 1], Rα is reflexive.
Particularly, R1 is reflexive.
Hence ∀x ∈ X, (x, x) ∈ R1 or R(x, x) = 1
It follows from that R is reflexive if and only if R1 (1-cut relation
of R )is reflexive.

9
 Definition 2.1.3
If ∀x, y ∈ X, R(x, y) = R(y, x), then R is called a symmetric( fuzzy)
relation.
Obviously, R is symmetric if and only if R = R−1
Definition 2.1.3
R is symmetric if and only if ∀α ∈ [0, 1], Rα is a symmetric relation.
Definition 2.1.4
Proof. If R is symmetric and (x, y) ∈ Rα , then R(y, x) = R(x, y) ≥
α.
Hence (y, x) ∈ Rα , which proves the symmetry of Rα
Conversely, assume that ∀α ∈ [0, 1], Rα is symmetric.
For any x, y ∈ X, take α = R(x, y). Then (x, y) ∈ Rα and hence
(y, x) ∈ Rα due to the symmetry of Rα
Therefore, R(y, x) ≥ α = R(x, y)
Next, x, y ∈ Rα and hence (y, x) ∈ Rα implies R(x, y) ≥ α and
R(y, x) ≥ α. We can take R(y, x) = α so that R(x, y) ≥ R(y, x).
Combining the two inequalities yields, R(x, y) = R(y, x)
Definition 2.1.5
If R ⊇ R2 , then R is said to be a transitive(fuzzy)relation
Definition 2.1.6
R is transitive if and only if ∀x, y, z ∈ X, R(x, z) ≥ R(x, y)∧ R(y, z)
Proof.

10
 R is transitive ⇔ R ⊇ R2

⇔ ∀x, z ∈ X, R(x, z) ≥ R2 (x, y)

⇔ ∀x, y, z ∈ X, R(x, z) ≥ R(x, y) ∧ R(y, z)

Definition 2.1.7
R is transitive if and only if ∀α ∈ [0, 1], Rα is transitive.
Proof. Suppose that R is transitive. Let (x, y), (y, z) ∈ Rα , for
any fixed α ∈ [0, 1]. It follows that R(x, y) ≥ α and R(y, z) ≥ α.
Then,

R(x, y) ∧ R(y, z) = α ≥ R(x, z) ⇒ (x, z) ∈ Rα

.Conversely, let ∀α ∈ [0, 1], Rα is transitive.

We prove ∀x, y, z ∈ X, R(x, z) ≥ R(x, y) ∧ R(y, z)
By letting, R(y, x) ∧ R(y, z) = α, we have R(x, y) ≥ α and
R(y, z) ≥ α,
so that (x, y) ∈ Rα and (y, z) ∈ Rα .
Hence R(x, z) ≥ R(y, x) ∧ R(y, z) = α.
Since Rα is transitive.
Definition 2.1.8
If R is reflexive, symmetric and transitive, then R is called a fuzzy
equivalence relation. R is a fuzzy equivalence relation if and only if
∀α ∈ [0, 1], Rα is an equivalence relation.
Definition 2.1.9

11
Let R be a fuzzy equivalence relation on X. A fuzzy set [a]R for
a ∈ X defined by;

∀x ∈ X, [a]R (x) = R(a, x)

is called the fuzzy equivalence class of a by R. The set X/R =

{[a]R | a ∈ X} of all fuzzy equivalence classes is called the fuzzy quo-
tient set of X by R.
Definition 2.1.10
If R is a fuzzy equivalence relation, then [a]R = [b]R if and only if
R(a, b) = 1
Proof. Suppose that [a]R = [b]R . R(a, b) = [a]R (b) = [b]R (b) =
R(b, b) = 1 Conversely let R(a, b) = 1. Then ∀x ∈ X,

[a]R (x) = R(a, x) ≥ R(a, b) ∧ R(b, x) = R(b, x) = [b]R (x)

.Since R is transitive.
Similarly, [b]R (x) = R(b, x) ≥ R(b, a) ∧ R(a, x) = R(a, x) =
[a]R (x).
We have [b]R (x) ≥ [a]R (x)
Consequently [a]R = [b]R .

12
 Chapter 2
FUZZY GROUPS
Definition 3.0.11
Definition 4.0.11. Let G be a group. A fuzzy subset A on G is called
a fuzzy subgroup of G if it satisfies the following conditions:

1. A(xy) ≥ A(x) ∧ A(y) for any x, y ∈ G

2. A x−1 ≥ A(x) for any x ∈ G.

Definition 3.0.12
Let A be a fuzzy subset of a set S. For t ∈ [0, 1], the set

At = {x ∈ S | A(x) ≥ t}

is called a level subset of the fuzzy subset A.

Note 3.0.13
At is a subset of S in the ordinary sense. The terminology ’level set’
was introduced by Zadeh.
Proposition 3.0.14
Let A be a fuzzy subgroup of G. For any x ∈ G,

1. A(x) ≤ A(e)

2. A x−1 = A(x) 3. A (xn ) ≥ A(x), where n is an arbitrary

integer.

Proof. 1. A(e) = A xx−1 ≥ A(x) ∧ A x−1 ≥ A(x) ∧ A(x) =

A(x)

13
  
−1 −1
≥ A x−1



2. A(x) = A x

We have from the definition of fuzzy subgroup that, A x−1 ≥

A(x)
Therefore, A x−1 = A(x).

3. The proof is by mathematical induction on n.

Let n = 2. Then we have from the definition of fuzzy subgroup

A x2 ≥ A(x) ∧ A(x) = A(x)

i.e. the result holds for n = 2.

Assume it holds for k.
Now,

A xk+1 ≥ A(x) ∧ A xk

≥ A(x) ∧ A(x) [ by induction hypothesis ]

= A(x)

Hence the result is true for any arbitrary integer n.

Proposition 3.0.15
Let A ∈ F (G). Then A is a fuzzy subgroup of G if and only if

A xy −1 ≥ A(x) ∧ A(y)

holds for any x, y ∈ G. Proof. Assume that A is a fuzzy subgroup

of G. Then we have,

14
 A xy −1 ≥ A(x) ∧ A y −1

= A(x) ∧ A(y)

Conversely, suppose that A xy −1 ≥ A(x) ∧ A(y), holds for any

x, y ∈ G. Then for any x ∈ G,

A(e) = A xx−1 ≥ A(x) ∧ A(x) = A(x)

i.e.,

A(x) ≤ A(e)

Thus for any x ∈ G,

A x−1 = A ex−1 ≥ A(e) ∧ A(x) = A(x)

i.e.,

A x−1 ≥ A(x)

Meanwhile for any x, y ∈ G

 
−1 −1
≥ A(x) ∧ A y −1 ≥ A(x) ∧ A(y)



A(xy) = A x y

Therefore,
A is a fuzzy subgroup of G.

15
 Proposition 3.0.16
Let G be a group and A be a fuzzy subgroup of G. Then the level
subset At , for t ∈ [0, 1], t ≤ A(e), is a subgroup of G, where e is the
identity of G.
Proof. We have,

At = {x ∈ G | A(x) ≥ t}

Since A(e) ≥ t, e ∈ At
Clearly, At ̸= ∅
Let x, y ∈ At . Then A(x) ≥ t and A(y) ≥ t
Since A is a fuzzy subgroup of G,

A(xy) ≥ A(x) ∧ A(y) ≥ t

i.e.,

A(xy) ≥ t

Hence xy ∈ At
Again x ∈ At ⇒ A(x) ≥ t
Since A is a fuzzy subgroup,

A x−1 ≥ A(x)

and hence A x−1 ≥ t i.e.

16
 x−1 ∈ At

Therefore, At is a subgroup of G

Proposition 3.0.17
Proposition 4.0.17. Let G be a group and A be a fuzzy subset of G
such that At is a subgroup of G for all t ∈ [0, 1], t ≤ A(e), then A is
a fuzzy subgroup of G.
Proof. Assume At is a subgroup of G for all t ∈ [0, 1], t ≤ A(e).
Let x, y ∈ G and let A(x) = t1 and A(y) = t2 . Then, x ∈ At1 and
y ∈ At2
Let us assume that t1 < t2 . Then it follows At2 ⊆ At1 .
So y ∈ At1 .
Thus x, y ∈ At1 and since At1 is a subgroup of G, by hypothesis,
xy ∈ At1 Therefore,

A(xy) ≥ t1 = A(x) ∧ A(y)

Next, let x ∈ G and let A(x) = t. Then x ∈ At

Since At is a subgroup, x−1 ∈ At
Therefore,

A x−1 ≥ t

And hence,

17
 A x−1 ≥ A(x)

Thus, A is a fuzzy subgroup of G.

Definition 3.0.18
(Level Subgroup) Let G be a group and A be a fuzzy subgroup of G.
The subgroups At , t ∈ [0, 1] and t ≤ A(e), are called level subgroups
of A.
Particularly,

AA(e) = {x ∈ G | A(x) = A(e)}

is a subgroup of G if A is a fuzzy subgroup of G. We shall denote

this subgroup by A∗ .
The binary multiplicative operation in G can be extended to F (G)
using the Zadeh’s extension principle.

Definition 3.0.19
Let A, B ∈ F (G). Then A ◦ B is defined by:
for any z ∈ G,

_
(A ◦ B)(z) = (A(x) ∧ B(y))
z=xy

In addition, for every A ∈ F (G), we shall define A−1 ∈ F (G) by:

for any x ∈ G, A−1 = A x−1 . With these notions, we present an

18
equivalent statement of a fuzzy subgroup.

Propostion 3.0.20
Let A ∈ F (G). Then A is a fuzzy subgroup of G if and only if

A ◦ A−1 = A

Definition 3.0.21
Let A be a fuzzy subgroup of G, and let f be an epimorphism of G
onto a group G′ . Then f (A) is a fuzzy subgroup of G′ .

Definition 3.0.22
Let f be a homomorphism from a group G to a group G′ , and let B
be a fuzzy subgroup of G′ . Then f −1 (B) is a fuzzy subgroup of G.
Proof. For any x, y ∈ G,

f −1 (B) xy −1 = B f xy −1

= B f (x)f y −1

≥ B(f (x)) ∧ B f (y)−1

= B(f (x)) ∧ B f y −1

≥ B(f (x)) ∧ B(f (y))

= f −1 (B)(x) ∧ f −1 (B)(y)

19
 Definition 3.0.23
Let A1 , A2 , . . . , An be fuzzy subgroups of G1 , G2 , . . . , Gn respectively.
Then the Cartesian product A1 × A2 × . . . × An is a fuzzy subgroup
of G1 × G2 × . . . × Gn
Proof. Form a tuple for xi , yi ∈ Gi . Since we have,

Ai xi yi−1 ≥ Ai (xi ) ∧ Ai (yi )

Then,

n
^
x1 y1−1 , x2 y2−1 , . . . , xn yn−1 Ai xi yi−1



(A1 × A2 × . . . × An ) =
i=1
^n
≥ Ai (xi ) ∧ Ai (yi )
i=1

= (A1 × A2 × . . . × An ) (x1 y1 , x2 y2 ,

Suppose G is a finite group, then the number of subgroups of G

is finite whereas the number of level subgroups of a fuzzy subgroup
A appears to be infinite. But since every level subgroups is indeed
a subgroup of G, not all these level subgroups are distinct.
The next proposition characterises this aspect:

Proposition 3.0.24
Let G be a group and A be a fuzzy subgroup of G. Two level
subgroups At1 , At2 (with t1 < t2 )of A are equal if and only if there

20
is no x ∈ G such that t1 < A(x) < t2 .
Proof. Let us consider At1 − At2 . Suppose there exists x ∈ G such
that t1 < A(x) < t2 , then At2 ⊂ At1 , since x belongs to At1 but not
in At2 , which contradicts the hypothesis.
Conversely, let there be no x ∈ G such that t1 < A(x) < t2 . Since
t1 < t2 , we have At2 ⊆ At1
Let x ∈ At1 , then A(x) ≥ t1 , and hence A(x) ≥ t2 , since A(x)
does not lie between t1 and t2 .
Therefore, x ∈ At2
So, At1 ⊆ At2
Thus, At1 = At2

Lemma 3.0.25
Let G be a finite group of order n and A be a fuzzy subgroup of G.
Let

Im(A) = {ti | A(x) = ti , for some x ∈ G}

Then {Ati } are the only level subgroups of A.

Proposition 3.0.26
Any subgroup H of a group G can be realized as a level subgroup of
some fuzzy subgroup of G.

21
 Proposition 3.0.27
Let Ā be the collection of all fuzzy subgroups of group G and B̄ be
the collection of all level subgroups of members of Ā. Then there
is a one to one correspondence between the subgroups of G and the
equivalence classes of level subgroups (under a suitable equivalence
relation on B̄ )

Definition 3.0.28
Let G be a group of prime power order. Then G is cyclic if and only
if there exists a fuzzy subgroup A of G such that for x, y ∈ G,

1. if A(x) = A(y), then < x >=< y > 2. if A(x) > A(y), then
< x >⊂< y >

Proposition 3.0.29
Let G be a cyclic p-group of order pn , where p is a prime. Let A be
a fuzzy subgroup of G then for x, y ∈ G

1. if O(x) > O(y), then A(y) ≥ A(x)

2. if O(x) = O(y), then A(x) = A(y)

Definition 3.0.30
Let A and B be two fuzzy subgroups of a group G. Then A and B
are said to be conjugate fuzzy subgroups of G if for some x ∈ G,

A(y) = B x−1 yx , ∀y ∈ G

22
 Proposition 3.0.31
If A and B are conjugate fuzzy subgroups of the group G, then
O(A) = O(B)

Normal fuzzy subgroups and Cosets

Now we introduce the notion of fuzzy middle coset of a group G.

Definition 3.1.1
Let A be a fuzzy subgroup of a group G. Then for any x, y ∈ G, a
fuzzy middle coset xAy of the group G is defined by,

(xAy)(a) = A x−1 ay −1 , ∀a ∈ G

Proposition 3.1.2
If A is a fuzzy subgroup of a group G, then for any x ∈ G the fuzzy
middle coset xAx−1 of the group G is also a fuzzy subgroup of the
group G

Proposition 3.1.3
Let A be any fuzzy subgroup of a group G and xAx−1 be a fuzzy

23
middle coset of the group G. Then

O xAx−1 = O(A)

for any x ∈ G.
Proof. Let A be a fuzzy subgroup of a group G and x ∈ G. Then
by the above proposition,the fuzzy middle coset xAx−1 is a fuzzy
subgroup of the group G. By the definition of a fuzzy middle coset
of the group G,

xAx−1 (a) = A x−1 ax

∀a ∈ G
Hence for any x ∈ G, A and xAx−1 are conjugate fuzzy subgroups
of the group G as there exists x ∈ G such that

xAx−1 (a) = A x−1 ax , ∀a ∈ G.

Thus O xAx−1 = O(A), for any x ∈ G.

Proposition 3.1.4
Let A be a fuzzy subgroup of a finite group G. Then O(A) divides
O(G).
Proof. Let A be a fuzzy subgroup of a finite group G with e as
its identity element. Clearly,

H = {x ∈ G | A(x) = A(e)}

24
 is a subgroup of the group G for H is a t-level subset of the group
G, where t = A(e)
By Lagrange’s theorem,

O(H) | O(G)

Thus by the definition of order of a fuzzy subgroup of the group

G,

O(A) | O(G)

Proposition 3.1.5
Let A and B be any two improper fuzzy subgroups of a group G.
Then A and B are conjugate fuzzy subgroups of the group G if and
only if A = B

Proposition 3.1.6
Let A and B be two fuzzy subsets of an abelian group G. Then
A and B are conjugate fuzzy subsets of the group G if and only if
A = B.
Proof. Let A and B be conjugate fuzzy subsets of the group G.
Then for some x ∈ G, we have

A(a) = B x−1 ax , ∀a ∈ G

= B x−1 xa , ∀a ∈ G

= B(a), ∀a ∈ G

25
 Hence A = B.
Conversely if A = B, then for the identity element e of the group
G, we have

A(a) = B e−1 ae , ∀a ∈ G

Hence A and B are conjugate fuzzy subsets of the group G.

Proposition 3.1.7
A fuzzy subgroup A of G is called normal if

A(xy) = A(yx)

holds for any x, y ∈ G.

Proposition 3.1.8
A fuzzy subgroup A of G is normal if and only if

A xyx−1 = A(y)

holds for any x, y ∈ G. Proof. Let A be a fuzzy subgroup of G.

Suppose A is normal. By definition, for any x, y ∈ G,

26
 A xyx−1 = A (xy)x−1

= A x−1 (xy)

= A x−1 xy

= A(y)

Conversely, suppose that A xyx−1 = A(y) holds for any x, y ∈

[Link],

xy = x(yx)x−1

A(xy) = A x(yx)x−1

= A(yx)

i.e., A is normal.

Proposition 3.1.9
A ∈ F (G) is a normal fuzzy subgroup of G if and only if

A ◦ A−1 = A and A ◦ B = B ◦ A

holds for all B ∈ F (G). Proof. For any fuzzy subgroup, A◦A−1 =
A. Take

27
 _
(A ◦ B)(z) = A(x) ∧ B(y)
z=xy
_
A zy −1 ∧ B(y)


=
y∈G
_
A y −1 z ∧ B(y)


=
y∈G
_
= A(x) ∧ B(y)
z=yx
_
= B(y) ∧ A(x)
z=yx

= (B ◦ A)(z)

Conversely, A ◦ A−1 = A implies A is a fuzzy subgroup. To show

that A is normal, take B = x−1 . Then


x−1 ◦ A (y)



A(xy) =

= A ◦ x−1 (y)


_
A(s) ∧ x−1 (t)

=
y=st

= A(yx)

Proposition 3.1.10
A ∈ F (G) is a normal fuzzy subgroup of G if and only if At is a
normal subgroup of G for any t ∈ [0, 1], t ≤ A(e)
Proof. A is a subgroup if and only if At is one. For normality,
take x ∈ G and y ∈ At . It follows that A xyx−1 = A(y) ≥ t.

Hence, xyx−1 ∈ At , and thus At is normal.

28
 Conversely, take x, y ∈ G and t = A(y). Then, t ∈ {A(x) | x ∈ G}
and y ∈
At . Hence xyx−1 ∈ At . Consequently

A xy −1 x ≥ t = A(y)

. As a result, A is a normal fuzzy subgroup of G.

Particularly, A∗ is a normal subgroup of G if A is a normal fuzzy
subgroup of G.

Definition 3.1.11
Let A be a fuzzy subgroup of G. For every x ∈ G, define xA, Ax ∈
F (G) by; ∀y ∈ G, (xA)(y) = A x−1 y and (Ax)(y) = A yx−1 .

Then xA and Ax are called the left coset and right coset of A with
respect to x respectively. Clearly, xA = Ax holds for any x ∈ G,
if A is a normal fuzzy subgroup of G. In this case, we simply call
xA(= Ax) a coset. Write G/A = {xA | x ∈ G}

Lemma 3.1.12
Let A be two normal fuzzy subgroups of G. Then xA ◦ yA = (xy)A
holds for any two cosets xA, yA ∈ G/A.
Proof. On the one hand, for any z ∈ G

29
 _
(xA ◦ yA)(z) = ((xA) (z1 ) ∧ (yA) (z2 ))
z=z1 z2

≥ (xA)(x) ∧ (yA) x−1 z

= A x−1 x ∧ A y −1 x−1 z

= A(e) ∧ A y −1 x−1 z

= A (xy)−1 z

= ((xy)A)(z)

We have the following result concerning (G/A, ◦).

Proposition 3.1.13
Let A be a normal fuzzy subgroup of G. Then

1. (G/A, ◦) is a group

2. G/A is isomorphic to G/A∗

Proof. 1
Clearly, the operation ◦ is associative, A is the identity of G/A and
the inverse of xA is x−1 [Link], (G/A, ◦) is a group.

2. For any x ∈ G, let f : xA → xA∗ . Then, for any x, y ∈ G,

f (xA ◦ yA) = f (xyA)

= xyA∗

= xA∗ yA∗

= f (xA)f (yA)

30
 Hence, f is a homomorphism. In order to prove that f is injective,
suppose that xA = yA. Then A x−1 z = A y −1 z for all z ∈ G.

Particularly, A x−1 y = A(e) when z = y. Thus x−1 y ∈ A∗ . As a

result, xA∗ = yA∗ . Hence f is injective. It is clear that f is surjective.

In summary, f is an isomorphism between G/A and G/A∗ .
G/A will be called the quotient group of G by a normal fuzzy
subgroup A of G.

Definition 3.1.14
Let A be a fuzzy subgroup of a group G and x ∈ G. Then the pseudo
fuzzy coset (xA)p is defined by,

(xA)p (y) = p(x)A(y), ∀y ∈ G

and for some p ∈ P .

Proposition 3.1.15
Let A be a fuzzy subgroup of a group G. Then the pseudo fuzzy
coset (xA)p is a fuzzy subgroup of G, ∀x ∈ G.
Proof. Let A be a fuzzy subgroup of a group G.∀x, y ∈ G,

(aA)p xy −1 = p(a)A xy −1

≥ p(a) ∧ {A(x), A(y)}

^
= {p(a)A(x), p(a)A(y)}
^
= {(aA)p (x), (aA)p (y)}

31
 i.e.

^
(aA)p xy −1 ≥ {(aA)p (x), (aA)p (y)} , ∀x, y ∈ G

This proves that (aA)p is a fuzzy subgroup of the group G.

Proposition 3.1.16
Let A be a normal fuzzy subgroup of G. Define Ā : G/A → [0, 1] by

∀xA ∈ G/A, Ā(xA) = A(x)

Then Ā is a normal fuzzy subgroup of G/A. Proof. Firstly, for

any xA ∈ G/A,

Ā (xA)−1 = Ā x−1 A

= A x−1

= A(x)

= Ā(xA)

and for any xA, yA ∈ G/A,

Ā(xAyA) = Ā(xyA)

= A(xy)

≥ A(x) ∧ A(y)

= Ā(xA) ∧ Ā(yA)

Hence, Ā is a fuzzy subgroup of G/A. Next, for any xA, yA ∈

G/A,

32
 Ā(xA ◦ yA) = Ā(xyA)

= A(xy)

= A(yx)

= Ā(yxA)

= Ā(yA ◦ xA).

Hence Ā is a normal fuzzy subgroup of G/A.

Proposition 3.1.17
Let A be a normal fuzzy subgroup of G and let f be an epimorphism
of G onto a group G. Then f (A) is a normal fuzzy subgroup of G.
Proof. We have, A is a normal fuzzy subgroup of G and f be an
epimorphism of G onto G. Then f (A) is a fuzzy subgroup of G. Let
u, v ∈ G. Then there exists x ∈ G such that f (x) = u since f is
surjective. Hence, we obtain successively

33
 _
f (A) uvu−1 =


A(z)
f (z)=uvu−1
_
= A(z)
f (z)=f (x)v(f (x))−1
_
= A(z) (f is a homomorphism )
f (x−1 zx)=v
_
A xyx−1


=
f (y)=v
_
= A(y)
f (y)=v

=f (A)(v)

Hence f ( A) is a normal fuzzy subgroup of G.

Proposition 3.1.18
Let f be a homomorphism from G to a group G and let B be a normal
fuzzy subgroup of G. Then f −1 (B) is a normal fuzzy subgroup of G.
Proof. We have, if f is a homomorphism from G to G and B is
a fuzzy subgroup of G, then f −1 (B) is a fuzzy subgroup of G. Now,
let x, y ∈ G. Then

34
 f −1 (B)(xy) = B(f (xy))

= B(f (x)f (y))

= B(f (y)f (x))

= B(f (yx))

= f −1 (B)(yx)

Hence, f −1 (B) is a normal fuzzy subgroup of G.

Proposition 3.1.19
Let A1 , A2 , . . . , An be normal fuzzy subgroups of G1 , G2 , . . . , Gn re-
spectively. Then the Cartesian product ni=1 Ai is a normal fuzzy
Q

subgroup of G1 × G2 × . . . × Gn Proof. We have, if A1 , A2 , . . . , An

be fuzzy subgroups of G1 , G2 , . . . , Gn respectively. Then the Carte-
sian product ni=1 Ai is a fuzzy subgroup of G1 × G2 × . . . × Gn .
Q

Furthermore, ∀ (x1 , x2 , . . . , xn ) , (y1 , y2 , . . . , yn ) ∈ G1 × G2 × . . . × Gn

n
! n
!
Y Y
Ai (x1 , x2 , . . . , xn ) (y1 , y2 , . . . , yn ) = Ai (x1 y1 , x2 y2 , . . . , xn yn )
i=1 i=1
n
^
= Ai (xi yi )
i=1
^n
= Ai (yi xi )
i=1
n
!
Y
= Ai (y1 x1 , y2 x2 , . . . , yn xn )
i=1

35
 Chapter 3
FUZZY SUBRINGS
In this and next subsection, we assume (R, +, ◦) is a ring. For
convenience, we write xy instead of x ◦ y for x, y ∈ R.

Definition 4.0.1
A ∈ F (R) is called a fuzzy sub ring of R if A satisfies that

1. ∀x, y ∈ R, A(x − y) ≥ A(x) ∧ A(y)

2. ∀x, y ∈ R, A(xy) ≥ A(x) ∧ A(y)

Proposition 4.0.2
A ∈ F (R) is a fuzzy sub ring of R if and only if At is a sub ring of
R for every t ∈ [0, 1], t ≤ A(e)
By proposition,

A∗ = {x | A(x) = A(0)}

is a sub ring of R. The operations on R can be extended to F (R)

as follows: ∀A, B ∈ F (R), ∀z ∈ R

_
(A + B)(z) = (A(x) ∧ B(y)
x+y=z
_
(A − B)(z) = (A(x) ∧ B(y)
x−y=z
_
(A ◦ B)(z) = (A(x) ∧ B(y)
xy=z

36
 Remark 4.0.3
A ∈ F (R) is a fuzzy sub ring of R if and only if A − A ⊆ A and
A◦A⊆A
Proof. Let A be a fuzzy sub ring of R. Since A is a fuzzy group
under addition, A − A ⊆ A. Moreover,

_
∀z ∈ R, A ◦ A(z) = (A(x) ∧ A(y) ≤ A(xy) = A(z)
xy=z

i.e.,

A◦A⊆A

Conversely, suppose that A − A ⊆ A and A ◦ A ⊆ A. Then

∀x, y ∈ R,

A(x − y) ≥ (A − A)(x − y)
_
= (A(s) ∧ A(t))
s−t=x−y

≥ A(x) ∧ A(y)

Similarly,

A(xy) ≥ (A ◦ A)(xy)
_
= (A(s) ∧ A(t))
xy=st

≥ A(x) ∧ A(y).

Consequently, A is a fuzzy sub ring of R.

37
 Proposition 4.0.4
Let A be a fuzzy sub ring of R and let f be an epimorphism of R
onto a ring R. Then f (A) is a fuzzy sub ring of R.
Proof. Let u, v ∈ R. Then there exists x, y ∈ R such that f (x) =
u and f (y) = v since f is surjective. Hence, we obtain successively

  
_ _
f (A)(u) ∧ f (A)(v) = A(x) ∧  A(y)
f (x)=u f (y)=v
_
= A(x) ∧ A(y)
f (x)=u,f (y)=v
_
≤ A(x − y) (A is a fuzzy subring of R)
f (x)=u,f (y)=v
_
≤ A(x − y) (f is a homomorphism )
f (x)−f (y)=u−v
_
= A(z)
f (z)=u−v

=f (A)(u − v)

Similarly, f (A)(uv) ≥ f (A)(u) ∧ f (A)(v)

Hence f (A) is a sub ring of R.

Proposition 4.0.5
Let f be a homomorphism from R to a ring R and let B be a fuzzy
sub ring of R. Then f −1 (B) is a sub ring of R. Proof. For any
x, y ∈ R,

38
 f −1 (B)(xy) = B(f (xy))

= B(f (x)f (y))

≥ B(f (x)) ∧ B(f (y))

= f −1 (B)(x) ∧ f −1 (B)(y)

Similarly,

f −1 (B)(x − y) ≥ f −1 (B)(x) ∧ f −1 (B)(y)

Thus f −1 (B) is a fuzzy sub ring of R.

For a fixed element k ∈ [0, 1], the fuzzy set k̂ on R defined by
k̂(x) = k, ∀x ∈ R is called a constant fuzzy set on R.
Obviously all constant fuzzy sets on R satisfy both the axioms of
a fuzzy ring. Hence all constant fuzzy sets on R are fuzzy rings on
R.
Now turning to non-constant fuzzy sets on R, we get the following
proposition.

Proposition 4.0.6
A non-constant fuzzy set A on R is a fuzzy ring on R if and only if
At is a sub ring of R, ∀t ∈ [0, 1]

• If R1 , R2 are rings, f : R1 → R2 is a function, A is a fuzzy set

on R1 and B is a fuzzy set on R2 , then the image of A under f
is the fuzzy set f (A) on R2 defined by,

39
 _
f (A)(y) = A(x) | x ∈ f −1 (y)

= 0, iff

The pre image of B under f is the fuzzy set f −1 (B) on R1 defined

by,

f −1 (B)(x) = B[f (x)], ∀x ∈ R1

Proposition 4.0.7
Let R1 , R2 be rings, f : R1 → R2 be a ring homomorphism. A be a
fuzzy ring on R1 and B be a fuzzy ring on R2 . Then f (A) is a fuzzy
ring on R2 and f −1 (B) is a fuzzy ring on R1 .

Fuzzy Ideals

Definition 4.1.1
A fuzzy ring A on a ring R is said to be a fuzzy left ideal if

A(xy) ≥ A(y), ∀x, y ∈ R

and fuzzy right ideal if

A(xy) ≥ A(x), ∀x, y ∈ R

A is called a fuzzy ideal if it is both a fuzzy left ideal and a fuzzy

40
right ideal. In other words, a fuzzy set A on R is a fuzzy ideal if
∀x, y ∈ R.

V
1. A(x − y) ≥ {A(x), A(y)}
W
2. A(xy) ≥ {A(x), A(y)}

Proposition 4.1.2
A fuzzy ring on R is a fuzzy ideal if and only if At is an ideal of
R, ∀t ∈ [0, 1] Proof. Suppose that A is a fuzzy ideal of R. Then At ,
for t ∈ [0, 1] is a sub ring of R.
Let x, y ∈ At and z ∈ R. Then

A(x − y) ≥ A(x) ∧ A(y) ≥ t

and

A(zx) ≥ A(z) ∨ A(x) ≥ A(x) ≥ t

Hence x − y ∈ At and zx ∈ At . Thus At is an ideal of R.

Conversely suppose that At is an ideal of R for every t ∈ [0, 1].
Then A is a fuzzy subring of R. Let x, y ∈ R and t = A(x). Then
t ∈ [0, 1] and x ∈ At . Since At is an ideal xy ∈ At .
Hence A(xy) ≥ t = A(x).
Similarly, A(xy) ≥ A(y)
Therefore A(xy) ≥ A(x) ∨ A(y)
Thus A is a fuzzy ideal of R.
Particularly,

41
 A∗ = {x | A(x) = A(0)}

is an ideal of R if A is a fuzzy ideal of R.

Proposition 4.1.3
Let A be a fuzzy ideal of R and let f be an epimorphism of R onto
a ring R. Then f (A) is a fuzzy ideal of R
If A is a fuzzy subring on R, we shall use the notation A0 for
{x ∈ R | A(x) = A(0)}

42
 CONCLUSION

We conclude this project by knowing the different aspect of fuzzy

algebra. This project is all about fuzzy set., fuzzy relation, fuzzy
subring. From this we get most of the ideas of fuzzy algebra and
relation between fuzzy sets. Fuzzy sets can be applied in the follow-
ing fields: Engineering, Psychology, medicine, Artificial intelligence,
Ecology, Decision making theory, Sociology, Computer science, Man-
ufacturing and others.

43
 Bibliography

[1] [Link] Kandasamy. Smarandache fuzzy Algebra Amer-

ican Research Press(2003)
[2] [Link]. Fuzzy Groups and Level Subgroups Journal of Math-
ematical Analysis and Applications (1981)
[3] Mustafa Akgul. Some properties of Fuzzy Groups Journal of
Mathematical Analysis and Applications(1988)
[4] Souriar Sebastian, [Link],[Link] and Divya Mary
Daise. On Fuzzy Ideals and Rings
[5] Mukherjee.T.K and sen. On Fuzzy Ideals of a ring, Fuzzy sets
and systems(1987)

44
DON BOSCO ARTS & SCIENCE
COLLEGE ANGADIKADAVU

DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

FUZZY MATRICES
 DON BOSCO ARTS & SCIENCE COLLEGE
ANGADIKADAVU
DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

FUZZY MATRICES

Dissertation submitted in the partial

Fulfillment of the requirement for the award of
MSc Degree in Mathematics of
Kannur University
Name : AMRUTHA P M
Roll No. : C1PSMM1902

Examiners:
1.
2.
 KANNUR UNIVERSITY

BONAFIDE CERTIFICATE

Certified that this project report “FUZZY MATRICES ” is the

bonafide work of AMRUTHA P M who carried out the project work
under my supervision.

Signature of Head Signature of Supervisor

 DECLARATION

I, AMRUTHA P M hereby declare that the Project work entitled

FUZZY MATRICES has been prepared by me and submitted to
Kannur University in partial fulfillment of requirement for the award
of Master of Science is a record of original work done by me under
the supervision of Mrs. PRIJA V, Assistant Professor, Department
of Mathematics, Don Bosco Arts & Science College, Angadikadavu.
I, also declare that this Project work has not been submitted by
me fully for the award of any Degree, Diploma, Title or recognition
before any authority.

Place: Angadikadavu AMRUTHA P M

Date: (C1PSMM1902)
 ACKNOWLEDGEMENT

Introduction is the proper place to begin. But first I bow my head

before the Almighty who is always with me. Also I must express my
deepest gratitude to people along the way. No words can adequately
express the sense of gratitude, still I try to express my heartfelt
thanks through words. The outset, I am deeply indebted to my
project supervisor Mrs. PRIJA V Assistant Professor, Department
of Mathematics, Don Bosco Arts and Science College, Angadika-
davu, for the invaluable guidance, loving encouragement and metic-
ulous care towards me throughout my career. I express my deep
sense of gratitude to all the faculty members of the Department of
Mathematics, Don Bosco Arts and Science College, Angadikadavu.
I can never forget the support and encouragement rendered by the
Principal and the Staff of Don Bosco Arts & Science College, An-
gadikadavu. I could not name many who sincerely supported and
helped for the successful completion of this Project. It is my plea-
sure and duty to thank each and everyone of them who walked with
me.

AMRUTHA P M
 CONTENTS

1. INTRODUCTION .......................................................... 07

2. PRELIMINARIES .......................................................... 09
2.1 OPERATION ON MATRICES .................................. 11

3. FUNDAMENTAL CONCEPTS .................................. 17

2.1 FUZZY VECTOR ...................................................... 18
2.1 STANDARD BASIS..................................................... 21

4. FUZZY MATRICES ........................................................ 29

4.1 ADDITION OF MATRICES......................................... 30
4.2 Max-Min COMPOSITION OF MATRICES .............. 33
4.3 COMPARABLE MATRICES.........................................39

5. RANKS ............................................................................... 43

6. GREEN’S RELATION ..................................................... 53

7. CONCLUSION ................................................................. 58

8. BIBLIOGRAPHY ........................................................... 59
 INTRODUCTION

We deal with fuzzy matrices, that is matrices over the fuzzy algebra
F= [0,1] under max-min operations (+,·) defined as a+b = max{a,b}
and a · b =min{a,b} for all a,b ∈ F and the standard order ‘≤’ of
real numbers. Let Fmn be the set of all m×n matrices over F. In
short Fn denotes Fnn . For A ∈ Fn , AT , R(A), C(A), ρr (A), ρc (A)
and ρ(A) denotes the transpose, row space, column space, row rank,
column rank and rank of A. Algebraic operations on matrices are
max-min operations, which are different from that of the standard
operations on real matrices. In practice, fuzzy matrices have been
proposed to represent fuzzy relations in a system based on fuzzy sets
theory, the behaviour of the dynamic fuzzy systems depends heavily
on the products of fuzzy matrices in the matrix representations of
the system. A ∈ Fmn is regular if there exists X such that AXA = A;
X is called a generalized (g − ) inverse of A and is denoted by A− . A− {
1} denotes the set of all g-inverses of a regular matrix A. A regular
matrix as one that has a generalized inverse lays the foundation in the
study on fuzzy relational equations. Regular fuzzy matrices play an
important role in estimation and inverse problem in fuzzy relational
equations and in fuzzy optimization problems. This motivates us
to develop the study on generalized regular fuzzy matrices. The
power of a fuzzy matrix are either convergent to a fuzzy matrix (or)
oscillating with finite period. For a fuzzy matrix A, Ak+d = Ak
for some integers k,d ≥ 0. Therefore, all fuzzy matrices have an

7
index and a period. On the other hand, most matrices over the non
negative real numbers will not have an index and a period. Spectral
inverses, such as group inverse and Drazin inverse are defined for
fuzzy matrices, analogous to that for complex matrices. For A ∈ Fn ,
The Drazin inverse of A is a solution of the equations : Ak XA =
Ak , XAX = X, AX = XA, for some positive integer k. Group inverse
is the solution of the equations : AXA = A, XAX = X, AX = XA.
Hence Drazin inverse and group inverse are identical when k = 1.
We define the regularity index of a matrix A ∈ Fn as a generalization
of the index of A. A characterization of a matrix whose regularity
index coincides with the index of A is established. It is shown that
for a matrix, regularity index is less than (or) equal to the index of
a matrix. Numerical examples are provided to illustrate the relation
between the regularity index and index of a fuzzy matrix.

8
 CHAPTER 1

PRELIMINARIES
Matrices are one of the most important tools in mathematics. Also
the Matrices are not only used as a representation of the coefficients
in system of linear equations, but utility of matrices far exceeds that
use.
Definition 1.0.1

A matrix is an ordered rectangular array of numbers or functions.

The numbers or functions are called the elements or the entries of

the matrix.
We denote matrices by capital letters. The following are some ex-
amples of matrices:

   
−2 5 2 + i 3 −1/2
 √   
A=0 B =  3.5 −1
   
5 2 
  √ 
3 6 3 5 5/7
 
3
1+x x 3
C= 
cosx sinx + 2 tanx

In the above examples, the horizontal lines of elements are said

to constitute, rows of the matrix and the vertical lines of elements

are said to constitute, columns of the matrix. Thus A has 3 rows

9
and 2 columns, B has 3 rows and 3 columns while C has 2 rows and
3 columns.
Definition 1.0.2

Order of a matrix-A matrix having m rows and n columns is called

a matrix of order mn.

Note 1.0.3

1. We shall follow the notation, namely A = [aij ]mn to indicate that

A is a matrix of order m × n.
2. We shall consider only those matrices whose elements are real

numbers or functions taking real values.

Definition 1.0.4

Identity matrix-A square matrix in which elements in the diagonal

are all 1 and rest are all zero is called an identity matrix. In other
words, the square matrix A = [aij ]n×n is an identity matrix, if
aij = 1 if i = j

0 if i ̸= j

We denote the identity matrix of order n by In. When order is clear

from the context, we simply write it as I. For example

 
1 0
 
0 1

10
 1.1 Operation on Matrices
we shall introduce certain operations on matrices, namely, addition

of matrices, multiplication of a matrix by a scalar, difference and

multiplication of matrices.
Definition 1.1.1

Addition of matrices - The operation of addition of matrices are

defined only if the matrices which are being added are of the same
order. If A and B are two (m × n) matrices with elements aik and bik
respectively, then their sum A + B is the (m × n) matrix C whose
elements cik are given by cik = aik + bik

   
1 2 3 3 4 1
A=  and B= 
−1 1 2 0 2 −1
 
4 6 4
then A+B =  
−1 3 1
Two matrices are said to be conformable to addition if they are of

the same order.

Definition 1.1.2

Multiplication of a matrix - Two matrices A and B can be multi-

plied together to form their product BA (in that order) only when
the number of columns of B is equal to the number of rows of A. A
and B are then said to be conformable to the product BA. We shall
see shortly, however, that A and B need not be conformable to the

11
product AB, and that, even when they are, the product AB does not
necessarily equal the product BA. That is, matrix multiplication is
in general non-commutative. Suppose now A is a matrix of order
(m × p) with elements aik and B is a matrix of order (p × n) with
elements bik . Then A and B are conformable to the product AB
which is a matrix C, say, of order (m × n) with elements cik defined
by,
Pp
cik = s=1 ais bsk

For example, if A and B are the matrices

 
  1 2
3 1 2  
A= , B = 3 1 


2 1 3  
2 3
 
10 13
Then, AB =  
11 14

Now the product BA is also defined in this case since the num-

ber of columns of B is equal to the number of rows of A. However,

it is readily found that

 
7 3 8
 
 
11 4 9 
 
12 5 13

Properties of matrix addition

1. Commutative Law If A = [aij ], B = [bij ] are matrices of the same

12
order, say m × n, then A + B = B + A.
2. Associative Law For any three matrices A = [aij ], B = [bij ], C =

[cij ] of the same order, say m × n, (A + B) + C = A + (B + C).

3. Existence of additive identity Let A = [aij ] be an m × n matrix

and O be an m × n zero matrix, then A + O = O + A = A. In other

words, O is the additive identity for matrix addition.
4. The existence of additive inverse Let A = be any matrix, then

we have another matrix as A = [aij ]m×n such that A+(A) = (A)+A

= O. So A is the additive inverse of A or negative of A.
Properties of multiplication of matrices

1. The associative law For any three matrices A, B and C. We have

(AB)C = A(BC), whenever both sides of the equality are defined.

2. The distributive law For three matrices A, B and C.

(i) A(B+C) = AB + AC (ii) (A + B)C = AC + BC, whenever

both sides of equality are defined.
3. The existence of multiplicative identity For every square matrix

A, there exist an identity matrix of same order such that IA = AI

= A.
Properties of scalar multiplication of a matrix

1. k(A + B) = kA + kB

2. (k + l)A = kA + lA

13
 Transpose of a matrix
If A = [aij ] be an m × n matrix, then the matrix obtained by in-

terchanging the rows and columns of A is called the transpose of A.

′
Transpose of the matrix A is denoted by A or (AT ). In other words,
if A = [aij ]m×n , then
′
A = [aij ]n×m . For example,

 
3 5  √ 
√ 
′ 3 3 0
If A =  3 Then A = 
 
1  
  5 1 −1/3
0 −1/3 2×3
3×2

Definition 1.1.3 Rank of a matrix

The rank of a matrix is the maximum number of independent rows

(or, the maximum number of independent columns). A square ma-

trix An×n is non-singular only if its rank is equal to n. For an m ×
n matrix,

• If m is less than n, then the maximum rank of the matrix is m.

• If m is greater than n, then the maximum rank of the matrix is

n.

The rank of a matrix would be zero only if the matrix had no ele-
ments. If a matrix had even one element, its minimum rank would
be one.
Definition 1.1.4

The number of non-zero rows in the row reduced form of a matrix

14
A is called the rank of A, denoted rank(A).
Method of find the rank of a matrix

Reduce the given matrix A into the row Echolon form or row

reduced echolon form by performing elementary operations. Then

rank of A is equal to the number of non zero rows of the row echolon
form is equal to the number of non-zero rows of row reduced form.
Solution of a system of linear equations

Consider the following system

a11 x1 + a12 x2 + · · · a1n xn = b1

a21 x1 + a22 x2 + · · · a2n xn = b2

· · ·
· · ·
· · ·
am1 x1 + am2 x2 + · · · amn xn = bm
The matrix form the system is AX = b where
     
a a12 · · · a1n x b
 11   1  1
 a21 a22 · · · a2n   x2   b2 
     
 , X= , b =  
 .. .. ..   ..   .. 
 . . .  . .
     
am1 am2 · · · amn xn bn
Here number of variables = n.

1. Consistency of the system:

If Rank of (A, b) = rank of A = r ,Then the system has solution.

In this case we say that the system is consistent.

15
 Case-1: If n = r, then the system has a unique solution.

Case-2: If r < n then system has infinite number of solutions.

2. In-consistency of the system:

If Rank of (A, b) ̸= rank of A. Then the system has no solution. In

this case we say that the system is in-consistent.

16
CHAPTER 2

FUNDAMENTAL CONCEPTS

The fundamental concepts of fuzzy matrices; A fuzzy algebra is a

mathematical system (F,+,.) with two binary operations +,. de-
fined on a set F satisfying the following properties:
(P1) Idempotence a+a = a
a.a = a

(P2) Commutativity a+b = b+a

a.b = b.a

(P3) Associativity a+(b+c) = (a+b)+c

a.(b.c) = (a.b).c

(P4) Distributivity a.(b+c) = (a.b)+(a.c)

a+(b.c) = (a+b).(a+c)

(P5) Universal bounds a+0 = a ; a+1 = 1

a.0 = 0 ; a.1 = a

A Boolean algebra with complementarity, that is, for each element

a, there exists an element a, called the complement of a such that
a+a = 1 ; a.a = 0.
Thus we have introduced the concepts of fuzzy algebra as a general-
ization of a Boolean algebra. By a fuzzy matrix, we mean a matrix
over a fuzzy algebra. A Boolean matrix is a special case of a fuzzy
matrix with entries from the set {0,1}.

17
 2.1 Fuzzy vector
A fuzzy vector is an n-tuple of elements of the fuzzy algebra F =

[0, 1].
Definition 2.1.1

Let Vn denote the set of all n-tuples (x1 , x2 , ..., xn ) over F. An

element of Vn is called a fuzzy vector of dimension n. The operations

(+,.) are defined on Vn as follows:
For x = (x1 , x2 , ..., xn ) and y = (y1 , y2 , ..., yn ) in Vn ,

x+y = (x1 +y1 , x2 +y2 , ..., xn +yn )

ax = (ax1 , ax2 ..., axn ), for a ∈ F.

The system Vn together with these operations of component wise

addition and fuzzy multiplication is called fuzzy vector (or) a vector

space over F, as the scalars are restricted to F. This will be sufficient
for our purpose. Let O denote the zero vector (0, 0,...,0).
Definition 2.1.2

Let V n = {xT : x ∈ Vn } where xT is the transpose of the vector

x. For u, v ∈ V n , a ∈ F, define av = (aT V T ) ; u+v = (uT + V T )T .

Then V n is a fuzzy vector space. If we write an element of Vn as a 1
× n matrix, it is called a row vector. The elements of V n are column
vectors.
For any result about Vn there exist a corresponding result about V n .
Thus V n is isomorphic to Vn as a fuzzy algebra.

18
 Definition 2.1.3
A subspace of Vn is a subset W of Vn such that O ∈ W and for x,

y ∈ W, x+y ∈ W .
Example
The set Bn of all n-tuples (a1 , a2 , ...,an ) over the two elements

Boolean algebra B = {0, 1} is a Boolean vectorspace of dimension

n. Bn is a subspace of Vn .
Example

W ={(0,0,0), (1,0,0), (0,1,0), (0,0,1), (1,1,0), (0,1,1), (1,0,1) (1,1,1)}

is a subspace of V3 .

Definition 2.1.4
P
A linear combination of elements of set S is a finite sum ai xi where

xi ∈ S and ai ∈ F. The set of all linear combinations of elements of

S is called the span of S, denoted as <S>.
Definition 2.1.5

For subsets S, W of Vn if <S> = W, then S is called a spanning set

(or) set of generators for W. If W is a subspace of Vn , then

<W > = W
Definition 2.1.6

A basis for a subspace W of Vn is a minimal spanning set for W.

Example

W = {(x, x) : x ∈ F} is a subspace of F × F. The singleton set

19
S = {(1,1)} is the minimal spanning set for W, since every element
(x, x) = x(1,1) with x ∈ F is a linear combination of (1,1) in S, S is
a basis for W.
Definition 2.1.7

A set S of vectors over a fuzzy algebra F is independent if and only

if each element of S is not a linear combination of other elements of

S, that is, no element v ∈ S is a linear combination of S\{v}.(where
\ denotes the set theoretic difference)
Definition 2.1.8

A set S of vectors over F is dependent if it is not a linearly inde-

pendent set.
Proposition 2.1.1

1. The set consisting of zero vector along is linearly dependent.

2. If X ⊂ Y and if X is linearly dependent then so is Y.

3. If X ⊂ Y and if Y is linearly dependent then so is X.
Example

The set of vectors {(0.5, 1), (1, 0.6), (0.7, 0.9)} is a dependent set,

since (0.7, 0.9) = 0.9(0.5, 1) + 0.7(1, 0.6).

Example

Let B = {b1 , b2 } be a subset of V3 , where b1 = (b, 0.3, 0.7) with b

≥ 0.8 and b2 = (0.4, 0.6, 0.5). We claim that the span of B, that is,
<B> contains the set S = {a1 , a2 , a3 , a4 } be a subset of V3 given by
a1 = (0.6, 0.3, 0.6), a2 = (0.4, 0.5, 0.5),

20
a3 = (0.5, 0.6, 0.5) and a4 = (0.8, 0.6, 0.7).
a1 = (0.6, 0.3, 0.6) = 0.6(b, 0.3, 0.7) + 0.3(0.4, 0.6, 0.5),

that is, a1 =(0.6)b1 + (0.3)b2

a2 = (0.4, 0.5, 0.5) = 0.4(b, 0.3, 0.7) + 0.5(0.4, 0.6, 0.5),

that is, a2 = (0.4)b1 + (0.5)b2

a3 = (0.5, 0.6, 0.5) = 0.5(b, 0.3, 0.7) + 0.6(0.4, 0.6, 0.5),

that is, a3 = (0.6)b1 + (0.3)b2

a4 = (0.8, 0.6, 0.7) = 0.8(b, 0.3, 0.7) + 0.6(0.4, 0.6, 0.5),

that is, a4 = (0.8)b1 + (0.6)b2

Thus each ai ( i = 1 to 4 ) is a linear combination of b1 and b1 .

Hence, <B> ⊇ S
Definition 2.1.9

Let A be a subset of Vn and A be the set of all subsets of Vn whose

span contains [Link] Schein rank of the set A, denoted as ρs (A) is

the minimum of the cardinality of the set in A . That is, ρs (A) =
min{ |C| : C ∈ A}

2.2 Standard Basis

The boolean algebra {0,1} any finitely generated subspace of Vn

has a unique basis. We define a standard basis and prove that any
finitely generated subspace of Vn over the fuzzy algebra F = [0,1]
has a unique standard basis.
Definition 2.2.1

A basis C over the fuzzy algebra F is a standard basis if and only

21
 P
if whenever ci = aij cj for ci , cj ∈ C and aij ∈ F then aii ci = ci .
Example

The basis {(0.5,1,0.5), (0,1,0.5), (0,0.5,1)} is not a standard basis

for (0.5,1,0.5) = (0.5)(0.5,1,0.5)+1(0,1,0.5)+(0.3)(0,0.5,1)

but (0.5,1,0.5) ̸= (0.5)(0.5,1,0.5). However the basis {(0.5,0.5,0.5),
(0,1,0.5), (0,0.5,1)} is a standard basis for the same subspace.
Theorem 2.2.1

Over the fuzzy algebra F =[0,1], any two bases for a finitely gen-

erated subspace have the same cardinality. Any finitely generated

subspace over F has a unique standard basis.
Proof

We first show that for any finite basis C, there exists a standard ba-

sis having the same cardinality. Let S be the set of all fuzzy vectors
each of whose entries equals some entry of a vector C. Then S is a
finite set.
P
Suppose C is not a standard basis, then ci = aij cj for some ci ∈
C and aij ∈ [0,1] with ci ̸= aii ci , that is ci ̸= min{aii , ci }, therefore
aii ci < ci .
Let C1 be the set obtained from C by replacing ci by aii ci . Then |C|
= |C1 | and <C> = <C1 > and it can be verified that C1 is indepen-
dent set and all the vectors of C1 are all in S.
Let us define an order relation on finite subsets of S as follows; Let
the weight of a finite subset be the sum of all entries of members of
the subset, regarded as real numbers.

22
We define that F1 ≤ F2 , for finite subsets F1 and F2 of S if weight of
F1 ≤ weight of F2 . Clearly this is a partial order relation on finite
subsets of S, Since aii ci < ci , C1 ≤ C and |C1 | = |C| is finite.
If C1 is a standard basis, then C1 is the required standard basis with
the same cardinality as C. If not, then repeat the process of replacing
C1 by a basis C2 and proceed.
Therefore after replacing bases of the form C by bases of the form
C1 , the process must terminate after a finite number of steps.
This can happen only if we have obtained a standard basis with the
same cardinality as C. This proves that for any finite basis, there
exist a standard basis with the same cardinality.
Next we show that there is only one standard basis. Let C be a
standard basis.
P
Suppose for ck ∈ C, we have ck = j aj , where
aj ∈ <C> , then aj can be expressed as a linear combination of basis
P
vector in C, that is aj = bji ci with bji ∈ F and ci ∈ C. Therefore,

X
ck = aj
j
X
= bji ci
i,j
XX
= ( bji ci )
i j
P
Since C is a standard basis. We have, ( j bjk )ck = ck ;

by using the fact that the fuzzy sum is the maximum, bjk ck = ck for

23
 P
some j and from aj = ij bji ci we get aj ≥ ck .
P
From ck = aj , we have ck ≥ aj .Therefore, it follows that ck = aj
for some j.
P
Thus we conclude that whenever ck = j aj , then ck equals some
summand aj .
Next to prove the uniqueness, if possible, let us assume that C and
′ ′ ′
C are two standard basis with | C | = | C |. Since C is a basis
, each element of C can be expressed as a linear combination of
′
elements of C . By proceeding argument, each element ai of C must
′
be a multiple of some element bj of C . Since fuzzy multiplication is
minimum , it follows that ai ≤ bj .
′
In the same manner, by using that each element of C is a multiple
′
of some element of C, and |C| = |C | it follows that ai = bj and
′
therefore C = C .
This proves the uniqueness of the standard basis. Hence the theorem.
Definition 2.2.2

The dimension of the finitely generated subspace S of a vector space

Vn over the fuzzy algebra F denoted by dim(S) is defined to be the

cardinality of the standard basis of S.
Example

The set {(1,0,0), (0,1,0), (0,0,1)} forms the standard basis for V3 .

dim(V3 ) = 3

24
 Example
Let S = {(0.5, 0.5, 0.5), (0, 1, 0.5), (0, 0.5, 1)} be a standard basis

of V3 . The subspace of V3 generated by S is W = <S> where

W = {(x, y, z) : 0 ≤ x ≤ 0.5 ≤ y, z ≤ 1} ∪ {(x, y, z) : 0 ≤ x ≤ y
= z ≤ 0.5}. Here dim(W) = |S| = 3
Remark 2.2.3

For vector spaces over a field, dim(S) = dim(W) if and only if S =

W. This fails for vector spaces over F. For spaces W = {(x, y, z)

: 0 ≤ x ≤ 0.5 ≤ y, z ≤ 1} ∪ {(x, y, z) : 0 ≤ x ≤ y = z ≤ 0.5}
and the set {(1,0,0), (0,1,0), (0,0,1)} forms the standard basis for V3 ,
dim(W) = dim(V3 ) = 3, but W ̸= V3 .
Example

Let W = {(x, y, z) : 0 ≤ x ≤ 0.5 ≤ y, z ≤ 1} ∪ {(x, y, z) : 0 ≤ x

≤ y = z ≤ 0.5} be the space. The vector (0.4, 0.5, 0.6) ∈ W can be

expressed as a linear combination of the standard basis vectors (0.5,
0.5, 0.5), (0, 1, 0.5), (0, 0.5, 1) as
(0.4, 0.5, 0.6)=α(0.5, 0.5, 0.5) + β(0, 1, 0.5) + γ(0, 0.5, 1) for α =

0.4, β ≤ 0.5 and γ = 0.6 ∈ F. Thus expression for the vector space
x is not unique. However, the unique standard basis of a finitely
generated subspace over F admits a unique representation.
Theorem 2.2.2

Let S be a finitely generated subspace of Vn and {c1 , c2 ,..., cn } be

the standard basis for S. Then any vector x ∈ S can be expressed

25
uniquely as a linear combination of the standard basis vectors.
Proof

Since {c1 , c2 ,..., cn } is the standard subspace for S, x is a linear

combination of the standard basis vectors. Let

x = nj=1 βj cj where βj ∈ F
P

In this expression, the coefficients βj ’s are not unique. If we write

this in the matrix form as x = (β1 , β2 , · · · βn )·C, where C is the

matrix whose rows are the basis vectors, then x = p·C has a solution
(β1 , β2 , · · ·,βn ) . It follows that this equation has a unique maximal
solution (p1 , p2 , · · ·, pn ).Then
x = nj=1 pj cj with pj ∈ F is the unique representation of
P

the vector x.
Theorem 2.2.3

Let S be a vector space over F and be the linear span of the vectors

x1 , x2 ..., xm . If some xi is a linear combination of x1 , x2 ..., xi−1 ,

xi+1 ...,xm , then the vectors x1 , x2 ,....,xi−1 , xi+1 ..., xm also spans S.
Proof

Let W = {x1 , x2 ,..., xm } such that S = <W >, since xi is a linear

combination of x1 , x2 ,...,xi−1 , xi+1 ,....,xm , there exist βj ’s for j = 1

to m and j ̸= i ∈ F, such that
xi = m
P
j=1:j̸ =i βj xj

Since S = <W >, any vector y ∈ S can be expressed as

y = α1 x1 + α2 x2 + ..., αi−1 xi−1 + αi+1 xi+1 +...+ αm xm

26
 Pm
= j=1:j̸ =i αj xj + αi xi
Pm
αj xj + αi ( m
P
= j=1:j̸ =i j=1:j̸ =i βj xj )
Pm
= j=1:j̸ =i γj xj

where, γj = αj + αi βj for j = 1 to m and j ̸= i are elements in F.

Since y is arbitrary vector in S, we have S = <W/{xi }>.

Thus the vectors x1 , x2 ..., xi−1 , xi+1 ..., xm spans S. Hence the the-
orem.
Theorem 2.2.4

Let S be a vector space over F of dimension n and let x1 , x2 ..., xm

be linearly independent vectors in [Link] there exists a basis for S

containing x1 , x2 ..., xm .
Proof

Let y1 , y2 ..., yn be the unique standard basis for S. Then the set

W = {x1 , x2 ..., xm , y1 , y2 ..., yn } is a linearly dependent subset

of [Link] yi for some i is a linear combination of the vectors in
W/{yi }. Since S is a subspace, <S> = S, and W = <S> .Thus S
is a linear span of W.
By the above theorem,W/{yi } also spans S. If the set is linearly in-
dependent set, then we have a basis for S as required. Otherwise we
continue the process until we get a basis containing x1 , x2 ..., xm .
Theorem 2.2.5

Any set of (n + 1) vectors in Vn is linearly dependent.

27
 Proof
If the set of (n + 1) vectors linearly independent,then by the above

theorem, we can find a basis for Vn containing the set. This is a con-
tradiction, since every basis for Vn must contain precisely n vectors.

28
 CHAPTER 3

FUZZY MATRICES
By a fuzzy matrix, we mean a matrix over a fuzzy algebra. Here
we confine with matrices over the fuzzy algebra F = [0,1] under the
max-min operations and with the usual ordering on real numbers.
Fuzzy matrices have quite different properties from matrices over a
field, due to the fact that addition in a fuzzy algebra does not form
a group. Here we will develop that every fuzzy linear transformation
on Vn can be represented by a unique fuzzy matrix. One of the most
important ways to study a fuzzy matrix is to consider its row space,
that is, the subspace of Vn spanned by its rows.
Definition 3.0.4

Let Fmn denote the set of all m × n matrices over F. If m = n,

in short, we write Fn . Elements of Fmn are called as membership

value matrices, binary fuzzy relation matrices (or) in short, fuzzy
matrices. Boolean algebra {0, 1} are special types of fuzzy matrices.
Definition 3.0.5

Let A = (aij ) ∈ Fmn . Then the element aij is called the (i, j) entry

of A. Let Ai∗ (or A∗j ) denote the ith row (column) of A . The row
space R(A) of A is the subspace of Vn generated by the rows {Ai∗ }
of A. The column space C(A) of A is the subspace of Vm generated
by the columns {A∗j } of A. The null space (or) kernel of A is the set

29
{x : xA = 0}. Note that row (column) vector is just an element of
Vn (V n ).
Definition 3.0.6

The n × m zero matrix O is the matrix all of whose entries are zero.
The n × n identity matrix I is the matrix (δij ) such that

1 if i = j

δij =
0 if i ̸= j


The n ×m universal matrix J is the matrix all of whose entries are

1.
3.1 Addition of Matrices

Definition 3.1.1
Let A = (aij ) ∈ Fmn and B = (bij ) ∈ Fmn . Then the matrix A+B

= (sup{aij , bij } ∈ Fmn is called the sum of A and B.

Example
   
0.5 0 1 0.2 0.4 0.6
   
If A = 0.8 0.2 0.3, B =
   
0.5 0.3 0.3
   
0 0.6 0.1 0.7 0.8 0
 
0.5 0.4 1
 
Then, A+B = 0.8 0.3 0.3
 
 
0.7 0.8 0.1
Theorem 3.1.1
The set Fmn is a fuzzy algebra under the component wise addition

and multiplication operation (+ , ⊙ ) defined as follows ;

30
For A = (aij ) and B = (bij ) in Fmn ,
A + B = (sup{aij , bij })

A ⊙ B = (inf{aij , bij })

Proof

The properties (P1) to (P4) of fuzzy algebra are automatically held.

A+0 = A and A ⊙ J = A, for all A∈ Fmn , Hence the zero matrix O

is the additive identity and the universal matrix J is the multiplica-
tive identity.
Thus identity element relative to the operations + and ⊙ exist. Fur-
ther A+J = J and A.0 = 0. Therefore (P6) holds.
For A = (aij ), B = (bij ) and C = (cij ) ∈ Fmn , Since (aij ) , (bij ) and
(cij ) are all real numbers in [0, 1] they are comparable.
If aij ≤ bij (or) cij then in both cases,
inf{aij , sup{bij , cij }} = aij and sup{inf{aij , bij }, inf{aij , cij } =
aij .
Therefore ij th entry of A ⊙(B+C) = ij th entry of (A ⊙ B) + (A ⊙
C).
If aij ≥ bij and cij then we have two cases ,
aij ≥ bij ≥ cij (or) aij ≥ cij ≥ bij then,
inf { aij , sup{bij ,cij }} = bij = sup{inf{aij ,bij }, inf{aij , cij }} (or)
inf{aij , sup{bij , cij }} = cij = sup{inf{aij , bij }, inf{aij , cij }}. There-
fore
ij th entry of A ⊙(B+C) = ij th entry of (A ⊙ B) + (A ⊙ C).
Hence A ⊙(B+C) = (A ⊙ B) + (A ⊙ C).

31
Similarly, if aij ≥ bij (or) cij , then both cases sup{aij , inf{bij , cij }}
= aij and inf{sup{aij ,bij }, sup{aij ,cij }} = aij . Therefore ij th
entry of A + (B ⊙ C) = ij th entry of (A + B) ⊙ (A + C).
If aij ≤ bij and cij , then we have two cases aij ≤ bij ≤ cij or aij ≤
cij ≤ bij then,
sup{aij , inf{bij , cij }} = bij = inf{sup{aij , bij }, sup{aij , cij }} (or)

sup{aij , inf{bij , cij }} = cij = inf{sup{aij , bij }, sup{aij , cij }}.

Therefore ij th entry of A + (B ⊙ C) = ij th entry of (A + B) ⊙ (A
+ C).
Hence, A + (B ⊙ C) = (A + B) ⊙ (A + C) can be proved.
Thus the property (P5) of distributivity holds. Thus Fmn is a fuzzy
algebra with the operation +, ⊙ .
Definition 3.1.2

Let A = aij ∈ Fmn and c ∈ F then the fuzzy multiplication, that

is, scalar multiplication with scalar restriction to F is defined as

cA = (inf {c, aij }) ∈ Fmn (3.1)

For the universal matrix J, by the definition ,

cJ = (inf{c, 1}) is the constant matrix all of whose entries are c.
Further under componentwise multiplication,
cJ ⊙ A = (inf {c, aij }) = cA (3.2)

Proposition 3.1.2

The set Fmn is a fuzzy vector space under the operations defined as

A + B = (sup{aij , bij }) and

cA = (inf{c, aij }) for A = (aij ), B = (bij ) ∈ Fmn and c ∈ F.

32
 Proof

For A, B, C ∈ Fmn ,

A + B = B + A ∈ Fmn (commutativity)

A + (B + C) = (A + B) + C (Associativity)

For all A ∈ Fmn , there exist an element O ∈ Fmn such that A+ O

= A.
For c ∈ F,
c(A+B) = cJ ⊙ (A+B)

= (cJ ⊙ A)+(cJ ⊙ B) (By Theorem 3.1.1)

= cA+cB (By (3.2))

For c1 , c2 ∈ F,

(c1 + c2 )A = ( c1 + c2 )J ⊙ A (by(3.2))

= (c1 J + c2 J) ⊙ A

= (c1 J) ⊙ A + (c2 J) ⊙ A

= c1 A + c2 A

Hence Fmn is a vector space over F. In particular for m = 1, it

reduces to definition.
3.2 Max-Min Composition of Matrices

Definition 3.2.1
For A = (aij ) ∈ Fmp and B = (bij ) ∈ Fpn , the max-min product

AB = (supk {inf{aik , bkj }}) ∈ Fmn

The product AB is defined if and only if the number of columns of

A is same as the number of rows of B ; A and B are said to be

33
conformable for multiplication.
Example
   
0.8 0.1 0.6 0.5
A= , B= 
0.2 1 0.4 0.3
Then,
    
0.6 0.5
(0.8 0.1)   (0.8 0.1)  
0.4 0.3 
 

AB = 
     
 0.6 0.5 
 (0.2 1)   (0.2 1)   
0.4 0.3
 
sup{inf {0.8, 0.6}, inf {0.1, 0.4} sup{inf {0.8, 0.5}, inf {0.1, 0.3}
= 
sup{inf {0.2, 0.6}, inf {1, 0.4} sup{inf {0.2, 0.5}, inf {1, 0.3}

 
sup{0.6, 0.1} sup{0.5, 0.1}
= 
sup{0.2, 0.4} sup{0.2, 0.3}

 
0.6 0.5
=  
0.4 0.3

Remark 3.2.2
P2
For elements a1 , a2 ∈ F, the sum i=1 ai = a1 + a2 = sup{a1 , a2 }

and product ab = inf{a, b}, hence the matrix product AB can be

expressed as AB = ( pk=1 aik bkj ).
P

In vector notation AB = (Ai∗ B∗j ) where Ai∗ is the ith row of A and
B∗j is the j th column of B.

34
Proposition 3.2.1

For any three matrices A, B, C over F of order m × n, n × p ,

p × q respectively, (AB)C = A(BC)

Proof

With the given type of matrices (AB)C and A(BC) are both defined

and are of type m × q. Let A = (aij ) , B = (bjk ), C = (ckl ) such that

the ranges of the suffixes i, j, k, l are 1 to m, 1 to n, 1 to p and 1 to
q respectively. Now (ik th ) element of the product AB = nj=1 aij bjk .
P

The (ilth ) element in the product (AB)C is the sum of products of

the corresponding elements in the (ith ) row of AB and (lth ) column
of C with k common. Thus
(ilth ) element of (AB)C = pk=1 {( nj=1 aij bjk )ckl }
P P

= pk=1 nj=1 aij bjk ckl

P P
(3.3)
Pp
Now (jlth ) element of the product BC = k=1 bjk ckl . Again the
ilth element in the product A(BC) is the sum of products of the
corresponding elements in the ith row of A and lth column of BC.
(ilth ) element of A(BC) = nj=1 aij ( pk=1 bjk ckl )
P P

= nj=1 pk=1 aij bjk ckl

P P
(3.4)

From (3.3) and (3.4), we see that (AB)C = A(BC).

Proposition 3.2.2

For any three matrices A, B, C over F of order m × n, n × p , p

× q respectively, A(B + C) = AB + AC

35
 Proof
Let A = (aij ), B = (bjk ), C = (cJk ) such that the ranges of the

suffixes i, j, k are 1 to m, 1 to n, 1 to p respectively. Now

(jk th ) element of B + C = bjk + cjk = sup{bjk , cjk }
(ik th ) element in the product of A and (B + C), that is, of A(B +
C) is the sum of the products of the corresponding elements in ith
row of A and k th column of (B + C).
= nj=1 aij (bjk + cjk )
P

= nj=1 aij bjk + nj=1 aij cjk

P P

= (ik th ) element of AB + (ik th ) element of AC

= (ik th ) element of (AB + AC)

Thus A(B + C) = AB + AC

Definition 3.2.3

We use the notation A2 to designate the product AA, A3 = AA2 =

A2 A ; and in general Ak = [Link]−1 = Ak−1 .A, for any positive integer

(k)
k. The matrix Ak is called the k th power of A. The notations aij
(k)
and Aij denote the (ij)th entry of Ak and (ij)th block of Ak . The
notation (Aij )k means the k th power of the (ij)th block of A.
Remark 3.2.4

We note that the matrix multiplication is not in general commuta-

tive, that is, AB ̸= BA,Further AB = 0, need not imply A = 0 (or)

B = 0 as in the case of real matrices.

36
 Example

     
0.8 0.1 0.6 0.5 0.6 0.2
Let A =  , B =  , C =  
0.2 1 0.4 0.3 0.7 0.3
 
  
0.6 0.5
(0.8 0.1)   (0.8 0.1)  
0.4 0.3 
 

AB = 
     
 0.6 0.5 
 (0.2 1)   (0.2 1)   
0.4 0.3
 
sup{0.6, 0.1} sup{0.5, 0.1}
= 
sup{0.2, 0.4} sup{0.2, 0.3}
 
0.6 0.5
= 
0.4 0.3
   
0.8 0.1
(0.6 0.5)   (0.6 0.5)   
0.2 1 
 

BA = 
     
 0.8 0.1 
(0.4 0.3)   (0.4 0.3)  
0.2 1
 
sup{0.6, 0.2} sup{0.1, 0.5}
= 
sup{0.4, 0.2} sup{0.1, 0.3}
 
0.6 0.5
= 
0.4 0.3

Therefore AB = BA

37
  
  
0.6 0.2
(0.6 0.5)   (0.6 0.5)   
0.7 0.3 
 

BC = 
     
 0.6 0.2 
(0.4 0.3)   (0.4 0.3)  
0.7 0.3
 
sup{0.6, 0.5} sup{0.2, 0.3}
= 
sup{0.4, 0.3} sup{0.2, 0.3}
 
0.6 0.3
= 
0.4 0.3
    
0.6 0.5
(0.6 0.2)   (0.6 0.2)   
0.4 0.3 
 

CB = 
     
 0.6 0.5 
(0.7 0.3)   (0.7 0.3)  
0.4 0.3
 
sup{0.6, 0.2} sup{0.5, 0.3}
= 
sup{0.6, 0.3} sup{0.5, 0.3}
 
0.6 0.5
= 
0.6 0.5

Therefore BC ̸= CB

Example
   
0.5 0 0 0
Let A =  , B= 
0.3 0 1 0.5

38
  
0 0
But, AB =  
0 0
AB = 0
 
0 0
BA =  
0.5 0
BA ̸= 0

Definition 3.2.5

A square matrix is called a permutation matrix if every row and

every column contains exactly one 1 and all the other entries are 0.
Let Pn denotes the set of all n × n permutation matrices.
Definition 3.2.6

For A ∈ Fmn , the transpose is obtained by interchanging its rows

and columns and is denoted by AT .

Note that if A ∈ Pn , then AAT = AT A = In , the identity matrix of
order n.
3.3 Comparable Matrices

Definition 3.3.1

Let A = (aij ) and B = (bij ) ∈ Fmn . We write A ≤ B if aij ≤ bij

for all i, j and we say that A is dominated by B (or) B dominates A.

A and B are said to be comparable, if either A ≤ B (or) B ≤ A.
Example

0 ≤ A ≤ J for all A ∈ Fmn .

39
 Proposition 3.3.1

Let A, B ∈ Fmn . Then A ≤ B ⇐⇒ A + B = B

Proof

If A ≤ B, then

A + B = (sup{aij , bij })
= (bij )

=B

Conversely, if A + B = B , then aij ≤ bij , that is, A ≤ B

Thus A ≤ B ⇐⇒ A + B = B
Proposition 3.3.2

Let A, B ∈ Fmn . If A ≤ B then for any C ∈ Fnp , AC ≤ BC and

for any D ∈ Fpm , DA ≤ DB.

Proof

A ≤ B, aik ≤ bik for all i = 1 to m and k = 1 to n. By fuzzy

multiplication, aik ckj ≤ bik ckj for j = 1 to p.

P P
By fuzzy addition we get, k aik ckj ≤ k bik ckj

Thus AC ≤ BC.
Similarly, Since A ≤ B, aik ≤ bik , for all i = 1 to n and k = 1 to p.
By fuzzy multiplication, aik ckj ≤ bik ckj for j = 1 to m.
P P
By fuzzy addition we get, k dik akj ≤ k dik bkj
Thus DA ≤ DB.

40
Proposition 3.3.3
Let A1 and A2 ∈ Fmn ; B1 and B2 ∈ Fnp . If A1 ≤ A2 and B1 ≤ B2

, then A1 B1 ≤ A2 B2 .
Proof

Let A1 = (aij ), A2 = (a′ij ) , B1 = (bjk ), and B2 = (b′jk ) . Since A1

≤ A2 and B1 ≤ B2 , we have aij ≤ a′ij and bjk ≤ b′jk for all i = 1 to

m , j = 1 to n , k = 1 to p. Therefore, aij bjk ≤ a′ij b′jk . Hence
P P ′ ′
j aij bjk ≤ j aij bjk .

Thus A1 B1 ≤ A2 B2

Example
   
0.8 0.5 1 1
Let A1 =   and A2 =  
0.4 1 1 1
A1 ≤ A2
   
0.6 0.5 0.6 0.2
B1 =   and B2 =  
0.4 0.3 0.7 0.3

B1 ≰ B2
   
0.6 0.5 0.6 0.2
A1 B1 =  ≰  = A2 B2
0.4 0.4 0.7 0.3
Proposition 3.3.4

For A ∈ Fmn , AAT A ≥ A.

Proof

Let A = (aij ) for i = 1 to m and j = 1 to n. Then (ilth ) element of

41
AAT A =
P P
j( k aik ajk )ajl

Since sum in F is the maximum, and multiplication is minimum, this

expression is greater than or equal to each term in the summation.
Therefore , for k = l, j = i, we have (ilth ) element of
AAT A ≥ aik aik ail = a3il = ai
Thus (ilth ) element of AAT A ≥ (ilth ) element of A.
Hence AAT A ≥ A.

42
 CHAPTER 4

RANKS
Rank is one of the fundamental concepts for the development of
fuzzy matrix theory as it is for field based matrices. However the row
rank and column rank of a fuzzy matrix need not be equal. There is
also a third rank concept, called fuzzy rank(or) Schein rank of great
importance in fuzzy matrix theory.
Definition 4.0.2

The row space R(A) of an m×n matrix A is the subspace of Vn

generated by the rows of A. The row rank ρr (A) is the smallest
possible size of a spanning set of R(A). The column space C(A) and
column rank ρc (A) are defined in dual fashion. Thus C(A) = R(AT )
and ρc (A) = ρr (AT ) .
Definition 4.0.3

Let A ∈ Fmn . The fuzzy rank ρf (A) is the smallest integer t such
that A = BC where B ∈ Fmt and C ∈ Ftn . This decomposition is
called the fuzzy rank factorization.
Remark 4.0.4

The row rank, column rank and the fuzzy rank of a zero matrix is
0. For a finite matrix A, ρr (A) is the maximum number linearly
independent rows of A.

43
 Example  
0 0.5
Let A= 
0.5 1
Any element (x, y) in R(A) is of the form
(x, y) = α(0, 0.5) + β(0.5, 1) for α, β ∈ F
= (0, α ∧ 0.5) + (β ∧ 0.5, β). where, ∧ denotes the minimum.

x = β ∧ 0.5 and y = max{α ∧ 0.5, β}

Therefore 0 ≤ x ≤ 0.5 and y ≥ α ∧ 0.5 and

β =⇒ y ≥ β ≥ β ∧ 0.5 ≥ x. Hence

R(A) = {(x , y): 0 ≤ x ≤ y ≤ 0.5} ∪ {(x , y): 0.5 = x ≤ y ≤ 1}.

Here A = AT . Hence C(A) = R(A) and ρf (A) = ρr (A) = ρc (A) = 2
Example
 
1 0.8 0
 
Let A = 0.8 0.7 0
 
 
0.7 0.6 0

(0.8, 0.7, 0.6)T and (1, 0.8, 0.7)T are linearly independent. Hence
ρc (A) = 2. Each row of A is not a linear combination of the other
two rows, that is, all the three rows are linearly independent. Hence
ρr (A) = 3

 
1 0.8  
  1 0.6 0
A = 0.8 0.7   = BC
 
  0.7 0.8 0
0.7 0.6

where, B ∈ F32 and C ∈ F23 ; A = BC is the fuzzy rank

44
factorization. Hence ρf (A) = 2.
Proposition 4.0.5

For A, B ∈ Fmn , we have the following:

1.R(B) ⊆ R(A) if and only if B = XA for some X ∈ Fm

2.C(B) ⊆ C(A) if and only if B = AY for some Y ∈ Fn

Proof

1. If B = XA, then by definition For A = (aij ) ∈ Fmp and B = (bij )

∈ Fpn , the max-min product

AB = (sup {inf {aik , bkj }}) ∈ Fmn

The product AB is defined if and only if the number of columns of

A is the same as the number of rows of B; A and B are said to be
conformable for multiplication, the ith row of XA, that is
P
(XA)i∗ = J xij Ai∗ ∈ R(A)

Hence R(B) ⊆ R(A).

Conversely, suppose R(B) ⊆ R(A), then each row of B is a linear

combination of the rows of A.

P
Hence Bi ∗ = xij Aj∗ and from which it follows that B = XA.
2. By using the facts C(A) = R(AT ) and

(XA)T = AT X T . Then by 1 we get the result.

Proposition 4.0.6

For a pair of matrices A and B if the product AB is defined, then

R(AB) = R(A)B ⊆ R(B) and C(AB) ⊆ C(A).

45
 Proof

Any vector y ∈ R(AB) is of the form y = uAB = xB where x =

uA ∈ R(A). Hence y ∈ R(B).

Thus,
R(AB) = R(A)B ⊆ R(B)

C(AB) = R(AB)T = R(B T AT ) ⊆ R(AT ) = C(A).

Remark 4.0.6

For matrices over a field, R(AB) ⊆ R(B) implies ρr (AB) ≤ ρr (B).

However, this fails for fuzzy matrices.

Proposition 4.0.7

Let A ∈ Fmn with ρr (A) = r. Then there exist matrices B ∈

Fmr and C ∈ Frn such that ρr (A) = ρr (C) = r and A = BC. This
decomposition is called a row rank factorization of A.
Proof

Since row rank of A is r, if R(A) is generated by the rows of an r ×

n matrix C, then ρr (C) = r and there exist an m × r matrix B such

that A = BC. Thus A = BC as required with ρr (A) = ρr (C) = r.
Proposition 4.0.8

Let A ∈ Fmn with ρc (A) = s. Then there exist matrices B ∈ Fms

and C ∈ Fsn such that ρc (A) = ρc (B) = s and A = BC. This de-
composition is called a column rank factorization of A.

46
Proof

Since column rank of A is s, if C(A) is generated by the rows of an

s × n matrix C, then ρc (C) = s and there exist an m × s matrix B

such that A = BC. Thus A = BC as required with ρc (A) = ρc (C)
= s.
Proposition 4.0.9

Let A ∈ Fmn with ρ(A) = ρr (A) = ρc (A) = r, then there exist

matrices B ∈ Fmr and C ∈ Frn such that A = BC with ρ(A) =

ρc (B) = ρr (C) = r. This is called a rank factorization of A.

Example
 
1 0.8 0
 
For the matrix, A = 0.8 0.7 0 , ρf (A) = ρc (A) = 2 and ρr (A)
 
 
0.7 0.6 0
= 3.
The decomposition A = BC with B ∈ F32 , C ∈ F23 is a column
rank factorization and also the fuzzy rank factorization. However A
= BC is not a row rank factorization.
Remark 4.0.7

For field based matrices A, B if R(A) ⊆ R(B) and ranks are equal

then R(A) = R(B). However this fails for fuzzy matrices.

47
 Example
   
0 0.5 0.5 0 0.5 1
Let A =   and B =  ;
0.3 0.5 1 1 0.5 0
  
0.5 0 0 0.5 1
A=   = XB.
1 0.3 1 0.5 0
Hence by result, R(A) ⊆ R(B). Here , (1, 0.5, 0) ̸= α(0, 0.5, 0.5)

+ β(0.3, 0.5, 1) for α, β ∈ [Link] fore 2nd row of B is not a linear

combination of the rows of [Link] R(B) ⊈ R(A) but ρr (A) = ρr (B)
= 2 and ρc (A) = ρc (X) = 2. Hence, A = XB is a rank factorization
of A.
Definition 4.0.8

In a row (column) rank factorization A = BC if R(A) = R(C) (C(A)

= C(B)) then it is said to be a full row (column) rank factorization

of A. Further,if A = BC with R(A) = R(C) and C(A) = C(B) then
it is said to be a full rank factorization.
Remark 4.0.9

For matrices A , B and X in previous example

  
0.5 0 0 0.5 1
A=   = XB is a rank factorization of A. In
1 0.3 1 0.5 0
A = XB, since C(A) = C(X), it is full column rank factorization,
but not a full rank factorization. However, for field based matrices,
rank factorization is the same as full rank factorization.

48
 Definition 4.0.10

For vectors v ∈ Vm and w ∈ Vn the cross-vector c(v, w) is the matrix

A = (aij ) ∈ Fmn such that aij = vi wj , that is, A = v T w.

Proposition 4.0.10

A non-zero matrix A is a cross-vector if and only if ρr (A) = 1 if and

only if ρc (A) = 1 if and only if ρf (A) = 1

Proof

If A is a cross-vector, then we have A = v T w, since A is non-zero,

v and w are non-zero vectors. Hence ρr (w) = 1, ρc (v T ) = ρr (w) =

1. Therefore, ρr (A) = 1 and ρc (A) = 1.
Conversely, if ρr (A) = 1, then by the result we get, A has a row
rank factorization of the form A = BC where B T and C are row
vectors. Hence A is a cross-vector. Similarly, when ρc (A) = 1, then
by the result, we can see that A is a cross-vector. Again if A is a
cross-vector, then A = v T w is a fuzzy rank factorization of A and
ρf (A) = 1. Conversely, if ρf (A) = 1, then A has a fuzzy rank fac-

torization A = BC with BT and C as row vectors. Thus A is a

cross-vector.
Proposition 4.0.11

Let A ∈ Fmn . The fuzzy rank ρf (A) satisfy the following properties.

1. ρf (A) ≤ inf{ρr (A), ρc (A)}

2. ρf (P AQ) ≤ ρf (A), if the matrix P AQ is defined.

49
 3. ρf (A) is the smallest size of a set S of vectors such that

R(A) ⊆ <S>.
4. ρf (A) is the least number of rank 1 matrices whose sum is A.

Proof

1. Let ρr (A) = r, then A has a row rank factorization A = BC with

ρr (A) = ρr (C) = r ; B ∈ C ∈ Frn .

Then by definition we get, ρf (A) ≤ ρr (A). Similarly by proposition,
using column rank factorization we get ρf (A) ≤ ρc (A). Therefore
ρf (A) ≤ inf{ρr (A), ρc (A)}. Thus 1 holds.
2. Let ρf (A) = t, then by definition, t is the least integer such that

A = BC with B ∈ Fmt , C ∈ Ftn is the fuzzy rank factorization of

A.
Then for P ∈ Fpm Q ∈ Fnq ,PAQ = (P B)(CQ) = V W , where V ∈
Fpt and W ∈ Ftq . Thus we have a decomposition for PAQ. There-
fore by definition, ρf (PAQ) ≤ t = ρf (A). Thus ρf (P AQ) ≤ ρf (A)
and 2 holds.
3. By definition of fuzzy rank, A = BC is a decomposition with B

∈ Fmt and C ∈ Ftn where t is the least integer.

By the result, R(A) = R(BC) ⊆ R(C) = <S> where S is the small-
est spanning set of R(C). Thus 3 holds.
4. Let s be the least number of rank 1 matrices Ai ’s such that A

= A1 + A2 + . . . As . Since each Ai is of rank 1, By the above

proposition each Ai is a cross-vector, that is, there exist vectors vi ∈
Vm and wi ∈ Vn such that Ai = viT wi for each i = 1 to s.

50
Let V ∈ Fmt and W ∈ Ftn be defined such that the ith column of V
is viT and ith row of W is wi .
Then A = V W is a fuzzy rank decomposition, that is ρf (A) = s.
Thus 4 holds.
Corollary 4.0.12

Let A ∈ Fmn and B ∈ Fnp . Then the following hold:

1. ρf (AB) ≤ min{ρr (A), ρr (B)}

2. ρf (AB) ≤ min{ρc (A), ρc (B)}

3. ρf (AB) ≤ min{ρr (A), ρr (B), ρc (A), ρc (B)}

Proof

1. ρf (AB) ≤ ρf (A) By above proposition condition 2

≤ ρr (A) By above proposition condition 1

ρf (AB) ≤ ρf (B) By above proposition condition 2

≤ ρr (B) By above proposition condition 1

Hence ρf (AB) ≤ min{ρr (A), ρr (B)}. Thus 1 holds.

2. ρf (AB) ≤ ρf (A) By above proposition condition 2

≤ ρc (A) By above proposition condition 1

ρf (AB) ≤ ρf (B) By above proposition condition 2

≤ ρc (B) By above proposition condition 1

Hence ρf (AB) ≤ min{ρc (A), ρc (B)}. Thus 2 holds.

3. From 1 and 2, we get ρf (AB) ≤ min{ρr (A), ρr (B), ρc (A), ρc (B)}

51
Definition 4.0.11

The Schein rank ρs (A) of a matrix A is defined as the least number

of rank 1 matrices whose sum is M.

Remark 4.0.12

For a Boolean matrix, Schein rank is defined as the least number of

cross vectors whose sum is A.

52
 CHAPTER 5

GREEN’S RELATION
The fuzzy matrix under max-min composition, that is fuzzy mul-
tiplication form a semigroup, that is associative law holds. A fuzzy
matrix is invertible if and only if it is a permutation matrix. We
introduce Green’s equivalence classes for fuzzy matrices. Green de-
fined five equivalence relations R, LH, D, J in any semigroup. For
fuzzy matrices these five equivalence relations can be characterized
in terms of row and column spaces.
Definition 5.0.13

For any two elements x, y of a semigroup S, we define the following

relations: 1. xRy if x = y or if there exist a, b ∈ S such that xa =

y and yb = x
2. xLy if x = y or if there exist a, b ∈ S such that

ax = y and by = x
3. xHy if xRy and xLy

4. xDy if there exist z such that xRz and zLy

5. xJy if there exist a, b, c, d such that axb = y, cyd = x (or)

if xRy (or) xLy. It is well known that the above relations are all
equivalence relations.
Proposition 5.0.15

For n × n matrices X, Y over a commutative semiring R with

53
identities 0 and 1, we have the following:
1. ARB if and only if C(A) = C(B)

2. ALB if and only if R(A) = R(B)

3. AHB if and only if C(A) = C(B) and R(A) = R(B)

4. ADB if and only if there exist matrices X, Y such that R(A)X

= R(AX) = R(B) and AXY = A

5. AJB if and only if there exist matrices X, Y such that

R(A)X ⊃ R(B) and R(B)Y ⊃ R(A).

Proof

1. By proposition, C(B) ⊆ C(A) =⇒ B = Ay for some y ∈ Fn By

definition, ARB if A = B or if there exist A, B ∈ S such that Ay

= B or Bx = A
That is,
ARB if Ay = B ⇐⇒ C(B) ⊆ C(A)

ARB if Bx = A ⇐⇒ C(A) ⊆ C(B)

=⇒ ARB ⇐⇒ C(B) = C(A)

2. By proposition, R(B) ⊆ R(A) ⇐⇒ B = xA for some x ∈ Fm

By definition, ALB if A = B or if there exist A, B ∈ S such that

Ax = B or By = A
That is,
ALB if Ax = B ⇐⇒ R(B) ⊆ R(A)

ALB if By = A ⇐⇒ R(A) ⊆ R(B)

54
 =⇒ ALB ⇐⇒ R(B) = R(A)

By definition we have, AHB if ARB and ALB

Then from 1 and 2,
ARB ⇐⇒ C(A) = C(B)

ALB ⇐⇒ R(A) = R(B)

=⇒ AHB ⇐⇒ C(A) = C(B) and R(A) = R(B)
4. Suppose ADB, then ARC and CRB for some matrix C.

Therefore, R(B) = R(C). Further by proposition, R(B) ⊆ R(A) ⇐⇒

B = XA for some X ∈ Fm ,
AX = C, CY = A for some matrices X and Y
R(A)X = R(AX) = R(C) = R(B)
AXY = CY = A
Conversely, let R(AX) = R(B) and AXY = A

Then, R(AX) = R(B) and (AX)Y = A

By 2 it follows that AXLB , by the result - C(B) ⊆ C(A) if and
only if B = AY for some Y ∈ Fn we get,
C(A) = C(AXY ) ⊆ C(AX) ⊆ C(A)
Therefore, C(A) = C(AX) by 1 It follows that, AR(AX).
Hence, [Link] 4 holds.
5. Suppose, AJB then there exist Z, W, X, Y such that ZAX =

B, W BY = A.
R(B) = R(ZAX) ⊆ R(AX) ⊆ R(A)X and
R(A) = R(W BY ) ⊆ R(BY ) ⊆ R(B)Y
Conversely, if R(B) ⊆ R(A)X and R(A) ⊆ R(B)Y ,

55
then R(B) ⊆ R(AX) and R(A) ⊆ R(BY ) again by the result - R(B)
⊆ R(A) if and only if B = XA for some Y ∈ Fm , we get there exist
matrices W, Z such that ZAX = B and W BY = A. Hence,AJB.
Definition 5.0.14

Two subspace V and W of Vn are isotopic if and only if there exist

matrices X, Y such that V X ⊆ W and W Y ⊆ V ; both XY and

Y X are identity mappings restricted to V and W respectively.
Example

The subspaces {(0, x) : x is a real number} and {(x, 0) : x is a real

number} are isotopic by the permutation matrices

 
0 1
X =Y = .
1 0

Proposition 5.0.16

Two matrices are D-equivalent if and only if their row spaces are

isotopic if and only if their column spaces are isotopic.

Proposition 5.0.17

Two D-equivalent matrices have the same row and column [Link]

J-equivalent matrices have the same fuzzy rank.

56
 Example

   
0 0 0 0 0 0
   
A =  1 0.6 0 ; B =  1 0.6 0.4
   
   
0.6 0 0 0.6 0.4 0

ρc (A) = 2 and ρc (B) = 3. Hence A and B are not D-equivalent.

However R(A) and R(B) are isomorphic.

Proposition 5.0.18

For fuzzy matrices, X and Y , XDY if and only if XJY

Proof

Suppose XJY ,then X = AY B, Y = CXD for some matrices A,

B, C, D.
Let K be the subset of the fuzzy algebra F = [0, 1] consisting of all
entries of X, Y, A, B, C, D.
Let S be the semigroup of all fuzzy matrices whose entries lie in K.
Then S is finite, since under max-min composition, sum of entries of
matrices AB, that is s(AB) ⊆ s(A) ∪ s(B) and XJY in S. Therefore
XDY in S and XDY is in the semigroup of all fuzzy matrices.

57
 CONCLUSION
Fuzzy matrices is one of the deepest and fascinating topic in math-
ematics. Through this project we covered the basic informations
about fuzzy matrices. We were understand that one of the most
important ways to study a fuzzy matrix is to consider its raw space.
The applications of fuzzy matrix are in retrieval system, medical di-
agnosis, database management system, decision making theory and
dynamical system. Also fuzzy matrices is one of the main research
area in mathematics.

58
 BIBLIOGRAPHY

1. [Link], in Fuzzy Matrix Theory and Application, (MJP

Publishers),Chennai, (2008).

2. Devi Prasad, in Elementary Linear Algebra, (Narosa [Link]),

(2006).

3. Devi Prasad, in An Introduction to Matrices, Sets and Groups

for Science Students, (Dover Publications, [Link] York), (1972).

59
DON BOSCO ARTS & SCIENCE
COLLEGE ANGADIKADAVU

DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

GRAPH DOMINATION
 DON BOSCO ARTS & SCIENCE COLLEGE
ANGADIKADAVU
DEPARTMENT OF MATHEMATICS
2021-2023
Project Report on

GRAPH DOMINATION

Dissertation submitted in the partial

Fulfillment of the requirement for the award of
MSc Degree in Mathematics of
Kannur University
Name : SURYA CHACKO
Roll No. : C1PSMM1908

Examiners:
1.
2.
 KANNUR UNIVERSITY

BONAFIDE CERTIFICATE

Certified that this project report “GRAPH DOMINATION ” is

the bonafide work of SURYA CHACKO who carried out the project
work under my supervision.

Signature of Head Signature of Supervisor

 DECLARATION

I, SURYA CHACKO hereby declare that the Project work entitled

GRAPH DOMINATION has been prepared by me and submitted to
Kannur University in partial fulfillment of requirement for the award
of Master of Science is a record of original work done by me under
the supervision of [Link] M V, Assistant Professor, Department
of Mathematics, Don Bosco Arts & Science College, Angadikadavu.
I, also declare that this Project work has not been submitted by
me fully for the award of any Degree, Diploma, Title or recognition
before any authority.

Place: Angadikadavu SURYA CHACKO

Date: (C1PSMM1908)
 ACKNOWLEDGEMENT

Introduction is the proper place to begin. But first I bow my head

before the Almighty who is always with me. Also I must express
my deepest gratitude to people along the way. No words can ad-
equately express the sense of gratitude, still I try to express my
heartfelt thanks through words. The outset, I am deeply indebted
to my project supervisor Mr. ANIL M V Assistant Professor, De-
partment of Mathematics, Don Bosco Arts and Science College, An-
gadikadavu, for the invaluable guidance, loving encouragement and
meticulous care towards me throughout my career. I express my deep
sense of gratitude to all the faculty members of the Department of
Mathematics, Don Bosco Arts and Science College, Angadikadavu.
I can never forget the support and encouragement rendered by the
Principal and the Staff of Don Bosco Arts & Science College, An-
gadikadavu. I could not name many who sincerely supported and
helped for the successful completion of this Project. It is my plea-
sure and duty to thank each and everyone of them who walked with
me.

SURYA CHACKO
 CONTENTS

1. INTRODUCTION ................................................................. 01

2. PRELIMINARIES ................................................................ 03

3. DOMINATION IN GRAPH ................................................. 06

4. BOUNDS FOR THE DOMINATION NUMBER ................. 15

5. VARIETIES OF DOMINATION .......................................... 23

6. INDEPENDENT DOMINATION AND

IRREDUNDANCE ............................................................... 32

7. APPLICATIONS OF DOMINATION
IN GRAPH THEORY ........................................................ 37

8. CONCLUSION ..................................................................... 39

9. BIBLIOGRAPHY ..................................................................40
 INTRODUCTION

Graph theory is a branch of Mathematics which has become quite

rich and interesting for several reasons. In last several decades,
hundreds of research articles have been published in Graph Theory.
There are several areas of Graph Theory which have received good
attention from mathematicians. Some of these areas are Coloring of
Graphs, Matching Theory, Domination Theory, Labeling of Graph
and areas related to Algebraic Graph Theory.

I found that the Theory of “Domination in Graph” deserves fur-

ther [Link], I choose this area for my project work. In this
project I introduce the basic concepts and variance of domination in
graphs. And it is reasonable to believe that “Domination in graphs”
has its origin in “chessboard domination”.

Technical systems like communication networks, power grids, traf-

fic management systems and enterprise data networks have a net
like [Link] us use a computer network for an example. Here
the computers are the components of the graph and the links be-
tween them represents the [Link] imagine, we want to monitor
the functions of each of the computers by one or a small number
of computers in such a way that every one of these computers can
control its neighbors. In Graph Theory we call these controllers a

1
dominating set. A variety of new problems appears as soon as we
impose additional properties on the dominating set.

2
 PRELIMINARIES

Definition 1
A graph G is an ordered triplet (V(G),E(G),IG ), where V(G) is a
nonempty set, E(G) is a set disjoint from V(G) and IG is an inci-
dence map that associates each elements of E(G) to an unordered
pair of element of V(G).
• Elements of V (G) are called vertices.
• Elements of E (G) are called edges.
• If for the edge e of G, IG (e) = {u,v} then we write e = u v.
• If IG (e) = {u,v} then u and v are called end vertices of e.
• If e is an edge with end vertices u and v then we say that e is
incident with u and v.
• A set of two or more edges of a graph G is called multiple edges
or parallel edges, if their points are the same.
• An edge with two ends are the same is called a loop.

Definition 2
A vertex u is a neighbor of v in G if uv is an edge in G an u̸=v.
• The set of all neighbors of v is the open neighborhood of v or
neighbor set of v and it is denoted by N(v).
• The set N[v] = N(v) ∪ {v} is called the closed neighborhood of
v in G. If the graph is explicit then open neighborhood and closed
neighborhood of v are denoted by NG (v), NG[v] respectively.

3
Definition 3
A sub graph H of G is said to be a spanning sub graph of G if
V(H) = V(G).

Definition 4
A subset S of the vertex set V of a graph G is called independent if
no two of its vertices are adjacent in G.

Definition 5
• Vertices u and v are adjacent to each other in G if and only if there
is an edge of G with u and v as its end.
• Two edges e and f are adjacent to each other in G if and only if
they have common end vertices.

Definition 6
• We denote number of vertices of a graph G by n (G) and it is called
order of the graph.
• The number of edges in a graph G is denoted by m(G) and it is
called the size of the graph.

Definition 7
A sub graph H of G is said to be induced sub graph of G if each
edge of G having its ends in V(H) is also an edge of H. The induced
sub graph G with vertex set S⊆V(G) is called the sub graph of G

4
induced by S and is denoted by G[S].

Definition 8
Let G be a graph and v∈V(G), then the number of edges incident
with v in the graph G is called the degree of the vertex and it is
denoted by d(v).
• The minimum degree of a graph G is denoted by δ(G).
• The maximum degree of a graph G is denoted by ∆(G).

Definition 9
A maximal independent set of G is an independent set that is not a
proper subset of another independent set of G.

Theorem 1 (Euler theorem)

The sum of the degrees of vertices of a graph is equal to twice the
number of edges.

5
 Chapter 1
DOMINATION IN GRAPH

Definition 1.1
A vertex v in a graph G is said to dominate itself and each of its
[Link] say in other words that v dominates the vertices of
its closed neighborhood N[v].

Figure 1.1 : Two dominating sets in a graph G

6
Definition 1.2
Let G be a graph. A set S⊂V is called a dominating set of G if every
vertex u ∈ V|S has a neighbor v∈S. Equivalently, every vertex of G
is either in S or in the neighbor set N(S)=U v∈S N(V ) of S in G. A
vertex u is said to be dominated by a vertex v∈G if either u = v or
uv∈E (G).

Definition 1.3
A γ-set of G is a minimum dominating set of G, which is a domi-
nating set of G whose cardinality is minimum. A dominating set S
of G is minimal if S properly contains no dominating set S of G.

Definition 1.4
The domination number of G is the cardinality of a minimum dom-
inating set of [Link] is denoted by γ(G).

Example 1.1
For the Petersen graph P,γ(p)=[Link] Figure 1.2, {v1 , v8 , v9 } is a γ-set
of P while the set {v1 ,v2 ,v3 ,v4 ,v5 } is a minimal dominating set of P.

7
 Figure 1.2 Petersen graph P for which γ(p)

Example 1.2
Determine the minimum number of queens that can be placed on the
chessboard (see figure 1.3) such that every square is either occupied
by one of the queens in a single move.

8
 Figure 1.3 Queen Domination

Two queens on a chessboard are attacking if the square occupied by

one of the queens can be reached by the other queen in a single move.
Otherwise they are non-attacking [Link] minimum number of
non-attacking queens that dominate all the squares of a chessboard
is the minimum cardinality of an independent dominating set in G.
The answer of this example is 5. Figure 1.3 gives one set of domi-
nating locations for the five queens.

Theorem 1.1
Let S be a dominating set of a graph [Link] S is a minimal domi-
nating set if and only if for each vertex u of S, one of the following
two conditions hold:
i. u is an isolated vertex of G[S],the sub graph induced by S in G.
ii. There exist a vertex v∈V|S such that u is the only neighbor of v
in S.

9
 Proof
Suppose that S is a minimal dominating set of [Link] for each ver-
tex u of S, S∈{u} is not a dominating set of [Link] there exist
V|(S|u) such that v is dominated by no vertex of S|{u}.
If v = u, then u is an isolated vertex of G[S],and hence condition (i)
holds.
If v̸=u, as S is a dominating set of G, then condition (ii) holds.
Conversely, assume that S is not a minimal dominating set of [Link]
there up that in example 1.1 the set S={v1 , v2 , v3 , v4 , v5 } is a min-
imal dominating set of the Petersen graph P in which no vertex is
an isolated vertex of P[S] and for each i={1, 2, 3, 4, 5}, vi is the only
vertex of S that is adjacent to vi + 5.

Definition 1.5

Let S be a dominating set of a graph G, and u ∈ [Link] private

neighborhood of u relative to S in G is the set of vertices which are
in the closed neighborhood of u, but not in the closed neighborhood
of any vertex in S |{u}.
Thus, the private neighborhood PN (u,s) of u with respect to S is
given by PN (u,s)=N[u]|Uv ∈ S|{u}N[v].Note that u∈ PN (u,s) if and
only if u is an isolated vertex of G[S] in G.
Theorem 1.1 can be restated as follows:

10
 Theorem 1.2
A dominating set S of a graph G is a minimal dominating set of G
if and only if PN (u,s)̸= ϕ for every u ∈ s.

Corollary 1.1
Let G be a graph having no isolated vertices. If S is a minimal dom-
inating set of G then V|S is a dominating set of G.

Proof
As S is a minimal dominating set, by theorem 1.2, PN (u,s) ̸= ϕ for
every u∈S. This means that for every u∈S, there exists v∈V|S such
that uv∈E(G), and consequently, V|S is a dominating set of G.

Corollary 1.2
Let G be a graph of order n≥2. If δ(G)≥1, then γ(G)≤ n2 .

Proof

As δ(G) ≥ 1, G has no isolated vertices. If S is a minimal dominat-

ing set of G, by corollary 1.1 both S and V | S are dominating sets
of G. Certainly, atleast one of them is of cardinality at most n2 .
Since, γ(G) ≤ min{| S |, | V − S |} ≤ n2 .

Corollary 1.3
If G is a connected graph of order n ≥ 2, γ(G) ≤ n2 .

11
 Proof
As G is a connected graph and n≥2, G has no isolated [Link],
by applying corollary 1.2 we get γ(G) ≤ n2 .

Definition 1.6
There are a number of variations of domination, we consider the best
known of these. In this variation, we could resist domination so that
a vertex u is only permitted to dominate a vertex v if v is a neighbour
of u. In order to distinguish this kind of domination from ordinary
domination we refer to this kind of domination as open domination,
although the term total domination is used as well.

• If w ∈ N(v) then we say here that v openly dominates w. That is

a vertex v openly dominates the vertices in its open neighborhood
N(v).

Definition 1.7

A set S of vertices in a graph G is an open dominating set of G if

every vertex of G is adjacent to atleast one vertex of S. Therefore, a
graph G contains an open dominating set if and only if G can con-
tain no isolated vertices.
Furthermore, if S is an open dominating set of G, then the sub graph
induced by S contains no isolated vertices.

12
 Definition 1.8
The minimum cardinality of an open dominating set is the open
domination number γ◦(G) of G. An open dominating set of cardi-
nality γ◦(G) is a minimum open dominating set for G.

Example 1.3

For the graph G of Figure 1.4, determine a) The domination number

of G. b) The open domination number of G.

Figure 1.4

Solution
γ(G)=3
γ◦(G)=4

13
 Example 1.4

In figure 1.5, the set S={u1 ,v1 ,w, v4 } is a minimum open dominating
set of G and so,γ◦(G) = 4.

14
 Chapter 2

BOUNDS FOR THE

DOMINATION NUMBER

In this section we present lower and upper bounds for the domina-
tion number γ(G).

Observation 2.1
(a) The vertex v dominates N(v) and |N(v)|≤∆G.
(b) Let v be any vertex of G. Then v|N(v) is a dominating set of G.
These two observations yield the following lower and upper bounds
for γ(G).

Theorem 2.1
n
For any graph G, 1+∆(G) ≤ γ(G) ≤ n − ∆(G).

15
 Illustration of Theorem 2.1

G
Figure 2.1

The graph G in Figure.2.1 be a graph with a dominating set S =

{v3 , v4 }.
Here , the domination number γ(G) = 2 , n = 5 and ∆(G) = 2.

n 5
   
1+∆(G) ≤ γ(G) ≤ n − ∆(G) ⇒ 1+2 ≤ 2 ≤ 5−2

5
⇒ 3 < 2 < 3

Bounds for the size m in terms of order n and

domination number γ(G)

Theorem 2.2
Let G be a graph of order n , size m and domination number γ .
Then

16
 m ≤ [ 12 (n − γ)(n − γ + 2)]..............................(a)

Proof

If γ = 1
1
2 (n − γ)(n − γ + 2) = 12 (n − 1)(n − 1 + 2)

= 12 (n − 1)(n + 1)

= 12 (n2 − 1)

n(n−1)
While the maximum value for m = 2 ( when G = Kn )
Since m = 12 n(n − 1) ≤ 12 (n2 − 1), and the result is true.

If γ = 2 ,

1
2 (n − γ)(n − γ + 2) = 12 (n − 2)(n − 2 + 2)

= 12 n(n − 2)

Now when γ = 2, by the theorem 2.1, ∆ ≤ n − 2

(since γ ≤ n − ∆ , then, 2 ≤ n − ∆) and by Euler’s Theorem
P
( di = 2m)
We get , 2m ≤ ∆
Then, 2m ≤ ∆ ≤ n − 2 gives 2m ≤ n − 2

m ≤ 12 (n − 2)

1
m ≤ 2 n(n − 2)

( 12 (n − 2)) ≤ 1
2 n(n − 2)

17
and then the result is true.
Thus the result is true for γ = 1 and γ = 2.
We now assume that γ ≥ 3.
We apply induction on n.
Let G be a graph of order n and size m and γ ≥ 3.
If v is a vertex of maximum degree ∆ of G , again by theorem 2.1,
| N (v) |= ∆ ≤ n − γ and hence ∆ = n − γ − r, where 0 ≤ r.
Given that,
| N (v) |= n − γ − r (see fig 2.2)

Let S = V \ N [v].
Then | S |=| V | − | N (v) | −1

= n − (n − γ − r) − 1

| S |= γ + r − 1..............................................(b)

Fig 2.2 the set S in the proof of 2.2

18
Let m1 be the size of G [ S ], m2 be the number of edges between S
and N(v ), and m3 be the size of G [ N [v ]].
Clearly, m = m1 + m2 + m3 . If D is a γ–set of G [ S ], then D ∪ {v}
is a dominating set of G.
Hence,
γ(G) = γ ≤ | D | +1 .............................................(c)

By the induction hypothesis, this implies, by virtue of (b) and (c),

that

m1 ≤ [ 12 (| S | − | D |) (| S | − | D | +2) ]

≤ [ 21 ( γ + r − 1 ) − ( γ − 1 ) ] [ ( γ + r − 1 ) − ( γ − 1) + 2 ]

= [ 21 [ γ + r − 1 − γ + 1] [ γ + r − 1 − γ + 1 + 2]]

= [ 12 r ( r + 2 ) ]

If u ∈ N (v) , then ( S \ N (v) ) ∪ {u, v} is dominating set of G.

Therefore , γ ≤ | S \ N (v) | +2

=| S | − | S ∩ N (u) | +2

≤ ( γ + r − 1 ) − | S ∩ N (u) | +2 [by (b)]

i.e , | S ∩ N (u) | ≤ γ + r − 1 + 2 − γ

=γ+1

This in turn implies that for each vertex u ∈ N (v) ,

| S ∩ N (u) | ≤ r + 1.

19
Consequently,

m2 = the number of edges between N(v) and S

≤ | N (v) | ( r + 1 )

= ∆(r + 1) ......................................................(e)

Now the sum of the degrees of vertices of N [v] ≤ ( ∆ + 1 ).

As there are m2 edges between N (v) and S.

The sum of the degrees of the vertices of N [v] in G [ N [v]]

= ( the sum of the degrees of the vertices of N (v) in G ) − m2

≤ ( ∆ + 1 ) − m2

Thus,
1
m3 ≤ 2 [ ∆( ∆ + 1 ) − m2 ]..............................................(f)

From (d), (e), and (f), we get

m = m1 + m2 + m3
1
≤ 2 r( r + 2 ) + m2 + 21 [ ∆( ∆ + 1 ) − m2 ]
m2
= 12 r( r + 2 ) + 12 [ ∆( ∆ + 1 ) ] + m2 − 2

= 12 r( r + 2 ) + 12 [ ∆( ∆ + 1)˙ ] + m2
2

= 12 r( r + 2 ) + 12 [ ∆( ∆ + 1 ) + m2 ]

1
≤ 2 r( r + 2 ) + 21 [ ∆( ∆ + 1 ) + ∆(r + 1 )]

1
≤ 2 ( n − γ − ∆ ) (n − γ − ∆ + 2) + 21 [ ∆(∆ + 1) + ∆(r + 1) ]

20
 (As ∆ = n − γ − r )

= 21 [ (n − γ)(n − γ + 2) + (n − γ)(−∆) − ∆(n − γ − ∆ + 2)] +

∆
2 [(∆ + 1) + (r + 1)]

= 21 [ (n−γ)(n−γ+2)]− ∆2 [(n−γ)+(n−γ−∆+2)]+ ∆2 [(∆+1)+(r+1)]

= 21 (n − γ)(n − γ + 2) − ∆2 [(n − γ) + (n − γ − ∆ + 2) − (∆ + 1)(r + 1)]

= 12 (n − γ)(n − γ + 2) − ∆2 [(n − γ) + (n − γ + 2) − ∆ − (∆ + 1)(r + 1)]

( since ∆ = n − γ − r)

= 21 (n − γ)(n − γ + 2) + ∆2 [(∆ + r) + (∆ + r + 2) − ∆ − ∆ − 1 − r − 1]

[ n − γ = ∆ + r]

= 12 ( n − γ ) (n − γ + 2 ) − ∆
2 (γ)

1
≤ 2 ( n − γ ) ( n − γ + 2)

Illustration of theorem 2.2

21
Let G be a graph with domination number γ(G), order n = 5 , size
m = 5.

Then ,
m ≤ [ 21 (n − γ)(n − γ + 2)] ⇒ 5 ≤ [ 12 ( 5 − 2) (5 − 2 + 2)]

⇒ 5 ≤ 7.5

22
 Chapter 3

VARIETIES OF DOMINATION
Definition 3.1
In this topic, we will consider a variety of conditions that can be im-
posed either on the dominated set V - S , or on V , or on the method
by which the vertices in V - S are dominated . These include the
following.

i) Multiple domination in which we insist that each vertex in V - S

be dominated by at least k vertices in S for a fixed positive integer k.

ii) Locating domination in which we insist that each vertex in

V - S has a unique set of vertices in S which dominate it.

iii) Strong domination in which we insist that each vertex v in

V - S be dominated by at least one vertex in S whose degree is
greater than or equal to the degree of v.

iv) Weak domination specifies that each vertex v in V - S domi-

nated by at least one vertex in S whose degree is less than or equal
to the degree of v.

v) Global domination in which we insist that the domination set

23
S also dominates the vertices V - S in the complement of G.

vi) Directed domination in digraphs in which we insist that for

each vertex v in V - S , there is a directed edge from u to v for at
least one vertex u in S.

MULTIPLE DOMINATION

Let S be a dominating set in a graph G = ( V , E ).

If we view the dominating set or a set that either monitors or con-
trols the vertices in V - S , then the removal or failure , of an edge
may result in a set which is no longer dominating.
If this is an undesirable situation , then it may be necessary to in-
crease the level of domination of each vertex so that , even if an edge
fails , the set S will still be a dominating set.
The idea of dominating each vertex in V-S multiple times originated
with Fink and Jacobson.

Theorem 3.1

If S is a γ-set of a graph G , then at least one vertex in V - S is

dominated by no more than two vertices in S.

Proof
Let S be a minimum dominating set in G.
Assume that every vertex in V - S is dominated by three or more
vertices.

24
Let u ∈ V - S and let v and w be two vertices in S which dominate
u.
It follows from our assumption that every vertex in V - S is domi-
nated by at least one vertex in S - { v , w }.
Therefore, the set V - S = S - {v,w } ∪ {u} is a dominating set.
But since |V - S |<| S |, we contradicts the assumption that S is a
minimum dominating set.

Theorem 3.2

If G is a graph with ∆(G) ≥ k ≥ 2 ,then γk (G) ≥ γ(G) + K − 2.

Proof
Let S be a minimum k - dominating set in G.
Let u ∈ V − S
And let v1 , v2 , ....., vk be distinct vertices in S which dominate u.
Notice that , since ∆(G) ≥ k ≥ 2.
We know that V − S ̸= ϕ because there is always a k - dominating
set, each vertex in V − S is dominated by at least one vertex in
S −{v1 , v2 , ....., vk }.
Therefore, since u dominates each vertex in {v2 , .....vk }.
We know that the set S = S − {v2 , ......, vk } − {u} is a dominating
set in G .
Therefore, γ(G) ≤ | V −S | = γK (G) − (K −1)+1 = γk (G)−k +2.

25
 Theorem 3.3

For any graph G ,

kn
γk (G) ≥ (∆(G)+k)

Proof

Let S be a minimum k - dominating set.

Let t denote the number of edges between S and V − S.
Since the degree of each vertex in S is at most ∆ ,
t ≤ ∆(G)γk (G)

But since each vertex in V S is adjacent with at least k vertices in

S.
We know that t ≥ k[n − γk (G)].
Combining these two inequalities produces
kn
γk (G) ≥ (∆(G)+k)

LOCATING DOMINATION

In the main, conditions on a dominating set S put restrictions either

on the sub graph induced by S , or on the number of times each
vertex v in V (G) must be dominated.

For example, each vertex v must be dominated at least k times, at

most k times,exactly k times , at least deg (v) times or as a certain
parity of times.

26
When considering locating sets in graphs one can think of selecting
a minimum set S of vertices of a graph G to achieve the function “
location through triangulation ” .

For a given k- tuple of vertices (v1 , v2 , ......, vk ) assign to each vertex

v ∈ V(G) the k-tuple of its distance to these vertices.
f(v) = (d(v, v1 ), d(v, v2 ), ....., d(v, vk ))

Letting S = v1 , v2 , ......, vk ), the k - tuple ƒ (v) is called S - location

of v.

A set S is a locating set for G if no two vertices have the same S -

location and the location number L(G) is the minimum cardinality
of a locating set.

Definition 3.2

i) The open k − neighborhood of a vertex v ∈ V, is the set

Nk (v) = {u : u ∈ v and d(u, v) ≤ k}

ii) The set Nk [v] = Nk (v) ∪ {v} is caled the closed k − neighborhood
of v.

iii) Every vertex w ∈ Nk [v] is said to be k − adjacent to v. Thus

we define the k - degree degk (v) of a vertex as | NK (V ).

The minimum k − degree ð equals min {degk (v) : v ∈ V } , while

the maximum k − degree ∆k (G) equals max {degk (v) : v ∈ V }.

iv) Finally , for a set S of vertices , we define Nk (S) to be the union

27
of the open k - neighborhoods of vertices in S , while Nk [S] is the
union of the closed k- neighborhoods of vertices in S.

v) A set is a distance − k dominating set if Nk [S] = V .

vi) The distance - k domination number γk (G) equals the minimum

cardinality of a distance - k dominating set in G.

STRONG AND WEAK DOMINATION

Given two adjacent vertices u and v we say that u strongly domi-

nates v if deg (u) ≥ deg (v).
Similarly , we say that v weakly dominates if deg (u) ≤ deg (v).
A set S ≤ V(G) is a strong dominating set of G if every vertex in
V−S is strongly dominated by at least one vertex in S .
Similarly, S is a weak - dominating set if every vertex in V − S is
weakly dominated by at least one vertex in S.
The strong ( weak ) domination number γs (G) ( respectively γw (G))
is the minimum cardinality of a strong ( weak ) dominating set of
G.

GLOBAL AND FACTOR DOMINATION

The notion of a dominating set S in a graph G = ( V , E ) can be

extended in a natural way to a set which is a dominating set in both
G and the complement G = (V, E)of G.

This concept was introduced independently by Sampathkumar , who

used the term global domination , and by Brigham and Dutton , who

28
used the more general term of factor domination.

A graph H = ( V, E ) is said to have a t - factoring into factors

F(H) = {G1 , G2 , ....., Gt if each graph Gi = (Vi , Ei ) has the same ver-
tex set , Vi = V , and the edge sets {E1 , E2 , ...., Et } form a partition
of E.
Given a t- factoring F of H, a subset Df ≤ V is a factor dominating
set if Df is a dominating set of Gi , for 1 ≤ i ≤ t.

The factor domination number γft (F (H)) is the minimum cardinal-

ity of a factor dominating set of F (H).
We will write γft when the graph H and the factoring F (H) are
understood , and γi will denote γ(Gi ).

DOMINATION IN DIRECTED GRAPHS

A directed graph ( also called a digraph ) D = ( V, A ) consists of

a finite set of vertices and a set A of directed edges , called arcs ,
where A ≤ V × V .
An arc (u, v) is said to be directed from u to v in which case u is
said to be a predecessor of v , v is successor of u and u dominates
v.

The outset of a vertex u is the set O(u) = {v : (u, v) ∈ A} , while

the inset is the set I(u) = {w : (w, u) ∈ A}. We also define
O[u] = O(u) ∪ {u} and I[u] = I(u) ∪ {u}.
The outdegree of a vertex u is od(u) =| O(u) | and the indegree of
u is in(u) =| I(u) |.

29
A set S ≤ V is independent if no two vertices of S are joined by an
arc .
The independence number β ◦ (D) is the maximum cardinality of an
independent set in D.

A set S ≤ V is called absorbent if for every vertex v ∈ V − S ,

there exists a vertex u ∈ S which is a successor of v , that is , v →
u is an arc in A.
A set S ≤ V is a dominating set of D if every vertex v ∈ V − S is
dominated by at least one vertex u ∈ S.
The domination number γ(D) of a digraph D is the minimum cardinality
of a dominating set in D.
A set S ≤ V is a kernel if it is both independent and absorbent.

Example 3.1

We observe that not every digraph has a kernel.

Fig 3.1 : Directed graph with and without kernals

30
The shaded vertices in the digraph D1 in the figure above form a
kernel of D1 ; while the graph D2 , which is obtained from D1 by
reversing the directions of three arcs ,has no kernel .
Thus we observe that not every digraph has a kernel.

31
 Chapter 4

INDEPENDENT DOMINATION
AND IRREDUNDANCE

Definition 4.1

A subset S of the vertex set of a graph G is an independent set of G

if S is both an independent and a dominating set. The independent
domination number i(G) of G is the minimum cardinality of an in-
dependent dominating set of G .

It is clear that γ(G) ≤ i(G) for any graph G .

For the path 5, γ(5) = i(5) = 2, (see fig.3.1(a)) while the graph G
of fig.3.1(b), γ(G = 2) and i(G) =2.

In fact, {v2 , v5 } is a γ-set for G , while {v1 , v3 , v5 } is a minimum

dominating set of G.

(a)

32
 (b)

Figure 3.1
(a) γ(G) = 2 = i(G)
(b) γ(G) = 2 while i(G) = 3

Theorem 4.1

Every maximal independent set of a graph G is a minimal dominat-

ing set.

Proof

Let S be a maximal independent set of G . Then S must be a domi-

nating set of G.
If not there exist a vertex v ∈ V | S that is not dominated by S , and
so S ∪ {u} is an independent set of G , violating the maximality of
S. Further, S must be a minimal dominating set of G . If not, there
exist a vertex u of S such that T = S − {u} is also a dominating set
of G .
This means, as u ∈
/ T , u has a neighbor in T and hence S is not
independent , a contradiction.

33
Definition 4.2

A set S ⊂ V (G) is called irredundant if every vertex V of S has

atleast one private neighbor.

This definition means that either v is an isolated vertex of G [ S] or

else v has a private neighbor in V | S ; that is , there exist at least
one vertex w ∈ V | S that is adjacent only to v in S .

Fig 4.1

S is an irredundant set but not a dominating set. Hence an irredun-

dant set S need not be dominating. If S is both irredundant and
dominating, then it is minimal dominating and vice versa.

34
Theorem 4.2

A set S ⊂ V is a minimal dominating set of G if and only if S is

both dominating and irredundant.

Proof

Assume that S is both a dominating and an irredundant set of G .

′
If S was not a minimal dominating set, there exist v ∈ S such that
′ ′
S | {v } is also a dominating set. But as S is irredundant, v has a
′ ′
private neighbor w (may be equal to w ).
′ ′
Since w has no neighbor in S | {v } is not a dominating set of G .
Thus, S is a minimal dominating set of G .
The proof of the converse is similar.

We define below a few more well-known graph parameters:

i. The minimum cardinality of a maximal irredundant set of a graph

G is known as the irredundance number and is denoted by ir(G).
ii. The maximum cardinality of an irredundant set is known as the
upper irredundance number and is denoted by IR(G).
iii. The maximum cardinality of a minimal dominating set is known
as the upper domination number and is denoted by Γ(G).

35
 Theorem 4.3

For any graph G the following inequality chain holds:

ir(G) ≤ γ(G) ≤ i(G) ≤ β ◦ (G) ≤ Γ(G) ≤ IR(G).

Proof

The inequality, ir(G) ≤ γ(G) is a consequence of the fact that every

minimal dominating set of vertices of G is an irredundant set.

We have already observed that γ(G) ≤ i(G).

From the above two inequalities we get , ir(G) ≤ γ(G) ≤ i(G).
Since an independent dominating set is independent, so
i(G) ≤ β ◦ (G).

Moreover, every maximum independent set is a dominating set, so

β ◦ (G) ≤ Γ(G).
Since every minimal dominating set is an irredundant set, it follows
that Γ(G) ≤ IR(G).
Then by combining all the above inequalities, we get
ir(G) ≤ γ(G) ≤ i(G) ≤ β ◦ (G) ≤ Γ(G) ≤ IR(G).

36
 APPLICATIONS OF
DOMINATION IN GRAPH
THEORY
Domination in graphs has application to several fields. Domination
arises in the facility problems, where the number of facilities (eg:
hospitals, fire stations...) is fixed and one attempts to minimize the
distance that a person needs to travel to get to the closet facility.

Facility location problems: The dominating sets in graphs are

natural models for facility location problems in operational research.
Facility location problems are concerned with the location of one or
more facilities in a way that optimizes a certain objectives such as
minimizing transportation cost, providing equitable service to cus-
tomers and capturing the largest market share.

Radio stations: We have a collection of small villages in a remote

part of the world. We locate the radio stations in some of these
villages so that the messages can be broadcast to all the villages in
that region. The radio station has a limited broadcasting range. We
have to place several stations to reach all the villagers. It is very
costly, we use domination graphs to locate as few as possible so that
the villagers got benefited.

37
 Modelling Biological Networks: Using graph theory as a
modelling tool in biologicl networks allow the utilization of the most
graphical invariants in such a way that is possible to identify sec-
ondary RNA motifs numerically. Those graphical invariants are
variations of the domination number of a graph. The results of the
research carried out in show that the variations of the domination
number can be used for correctly distinguished among the trees that
represent native structures and those that are not likely candidates
to represent RNA.

38
 CONCLUSION
In this thesis my research mainly focused on domination in graphs.
Domination has many applications in the areas such as facility lo-
cation problems, planning of defence strategies, surveillance related
problems etc.

In the four chapters of the thesis, first chapter is based on the concept
of Domination, Open Domination and some examples. The main re-
sult of the chapter helps us to find Domination number by using
various methods. In the second chapter we saw and discussed about
the bounds for the Domination number with the help of suitable ex-
amples. In chapter three we discussed about the various varities of
Domination. We also introduced a new kind of Domination in the
fourth chapter namely Independent Domination.

39
 BIBLIOGRAPHY

[1] [Link] , [Link] :A Textbook of Graph Theory”Second

edition.

[2]Gary Chartrand , Ping Zhang : Introduction to Graph Theory” TATA

McGRAW-HILL EDITION.

[3] Jennifer [Link] : Domination in graphs(2010)”,Graduate theses and-

Dissertations.

[4] Douglas [Link] : Introduction to Graph Theory”, Second edition.