See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/311589083

INTRODUCTION TO METRIC SPACES

Chapter · November 2016

CITATIONS READS

0 9,379

2 authors:

Rajesh Singh Jeetendra Aggarwal

University of Delhi Shivaji College, University of Delhi
16 PUBLICATIONS   4 CITATIONS    12 PUBLICATIONS   26 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

GRAPHOIDAL LENGTH OF A GRAPH View project

GRAPHOIDAL DOMINATION View project

All content following this page was uploaded by Rajesh Singh on 13 December 2016.

The user has requested enhancement of the downloaded file.

 INTRODUCTION TO METRIC SPACES

Subject : Analysis
Course : Metric Spaces
Author Name : Dr. Jeetendra Aggarwal and Rajesh Singh
College/Department : Department of Mathematics
Shivaji College, University of Delhi
and
Department of Mathematics
University of Delhi
 INTRODUCTION TO METRIC SPACES 2

Table of Contents

Chapter: Introduction Metric Spaces

1. Learning Outcomes
2. Prerequisites
3. Introduction
4. Example of Metric Spaces
5. Metric Subspaces And Metric Superspaces
6. Isometries
7. Assorted Examples Of Metric Spaces
8. Norms on Vector Spaces
9. Solved Problems
10. Summary
11. Problems
12. References
 INTRODUCTION TO METRIC SPACES 3

1. LEARNING OUTCOMES
This chapter will introduce the reader to the concept of metrics (a class of functions which
is regarded as generalization of the notion of distance) and metric spaces. A lot emphasis
has been given to motivate the ideas under discussion to help the reader develop skill in
using his imagination to visualize the abstract nature of the subject. Variety of examples
along with real life applications have been provided to understand and appreciate the
beauty of metric spaces. Moreover the concepts of metric subspace, metric superspace,
isometry (i.e., distance preserving functions between metric spaces) and norms on linear
spaces are also discussed in detail.

2. PREREQUISITES
It is assumed that the reader has done a course which includes introductory real
analysis, that is, the reader has familiarity with concepts like convergence of sequence of
real numbers, continuity of real valued functions etc. But it is nowhere assumed that the
reader has mastered these topics and hence all the concepts are well explained. Next we
list few inequalities that are required in the chapter.

Inequalities

1. Cauchy-Schwarz Inequality

Let 𝑥𝑖 , 𝑦𝑖 ∈ ℝ for 𝑖 = 1,2, … , 𝑛, then following inequality holds:

𝑛 𝑛 𝑛

∑|𝑥𝑖 𝑦𝑖 | ≤ √∑|𝑥𝑖 |2 √∑|𝑦𝑖 |2 .

𝑖=1 𝑖=1 𝑖=1

2. Minkowski’s Inequality

Let 𝑥𝑖 , 𝑦𝑖 ∈ ℝ for 𝑖 = 1,2, … , 𝑛 and 𝑝 ≥ 1 be any real number. Then

1 1 1
𝑛 𝑝 𝑛 𝑝 𝑛 𝑝

(∑|𝑥𝑖 + 𝑦𝑖 |𝑝 ) ≤ (∑|𝑥𝑖 |𝑝 ) + (∑|𝑦𝑖 |𝑝 ) .

𝑖=1 𝑖=1 𝑖=1

3. Minkowski’s Inequality for Infinite Sums

Let 𝑝 ≥ 1 be any real number and {𝑥𝑛 }𝑛≥1 , {𝑦𝑛 }𝑛≥1 be real sequences such that
∞ ∞

∑|𝑥𝑛 |𝑝 < ∞ 𝑎𝑛𝑑 ∑|𝑦𝑛 |𝑝 < ∞ .

𝑛=1 𝑛=1

Then ∑∞
𝑛=1|𝑥𝑛 + 𝑦𝑛 | is convergent. Moreover,
𝑝

1 1 1
∞ 𝑝 ∞ 𝑝 ∞ 𝑝

(∑|𝑥𝑛 + 𝑦𝑛 |𝑝 ) ≤ (∑|𝑥𝑛 |𝑝 ) + (∑|𝑦𝑛 |𝑝 ) .

𝑛=1 𝑛=1 𝑛=1

Theorem A. For any 𝑤, 𝑥, 𝑦, 𝑧 ∈ ℝ ,

2
[√(𝑤 2 + 𝑦 2 ) + √(𝑥 2 + 𝑧 2 )] ≥ [𝑤 + 𝑥]2 + [𝑦 + 𝑧]2 .

Proof: Consider
(𝑤𝑧 − 𝑥𝑦)2 ≥ 0
⇒ 𝑤 2 𝑧 2 + 𝑥 2 𝑦 2 − 2𝑤𝑧𝑥𝑦 ≥ 0
 INTRODUCTION TO METRIC SPACES 4

⇒ 𝑤 2 𝑧 2 + 𝑥 2 𝑦 2 ≥ 2𝑤𝑧𝑥𝑦
⇒ 𝑤 2 𝑥 2 + 𝑦 2 𝑧 2 + 𝑤 2 𝑧 2 + 𝑥 2 𝑦 2 ≥ 𝑤 2 𝑥 2 + 𝑦 2 𝑧 2 + 2𝑤𝑧𝑥𝑦 [𝑎𝑑𝑑𝑖𝑛𝑔 𝑤 2 𝑥 2 + 𝑦 2 𝑧 2 𝑏𝑜𝑡ℎ 𝑠𝑖𝑑𝑒𝑠 ]
⇒ 𝑤 2 (𝑥 2 + 𝑧 2 ) + 𝑦 2 (𝑥 2 + 𝑧 2 ) ≥ (𝑤𝑥 + 𝑦𝑧)2
⇒ (𝑤 2 + 𝑦 2 )(𝑥 2 + 𝑧 2 ) ≥ (𝑤𝑥 + 𝑦𝑧)2

⇒ √(𝑤 2 + 𝑦 2 )(𝑥 2 + 𝑧 2 ) ≥ (𝑤𝑥 + 𝑦𝑧) [𝑻𝒉𝒊𝒔 𝒊𝒔 𝑪𝒂𝒖𝒄𝒉𝒚 𝑺𝒄𝒉𝒂𝒘𝒓𝒛 𝑰𝒏𝒆𝒒𝒖𝒂𝒍𝒊𝒕𝒚]

⇒ 2√(𝑤 2 + 𝑦 2 )(𝑥 2 + 𝑧 2 ) ≥ 2(𝑤𝑥 + 𝑦𝑧)

⇒ (𝑤 2 + 𝑦 2 ) + (𝑥 2 + 𝑧 2 ) + 2√(𝑤 2 + 𝑦 2 )(𝑥 2 + 𝑧 2 ) ≥ (𝑤 2 + 𝑦 2 ) + (𝑥 2 + 𝑧 2 ) + 2(𝑤𝑥 + 𝑦𝑧)

2
⇒ [√(𝑤 2 + 𝑦 2 ) + √(𝑥 2 + 𝑧 2 )] ≥ [𝑤 + 𝑥]2 + [𝑦 + 𝑧]2

Hence the Inequality. ∎

3. INTRODUCTION

A metric space is a non-empty set equipped with structure determined by a well-defined

notion of distance. The term ‘metric’ is derived from the word metor (measure). Natural
and immediate questions that comes to mind are what do we mean by measure, what can
be measured and how it can be measured? In the search of answers to these questions,
let us consider the following example:

Suppose a person wants to go from New Delhi to

Mumbai. The adjoining figure gives possible routes from
New Delhi to Mumbai. Depending on the situation, he
may travel by taking any of the given possible option.
We note that there are two different ways to interpret
his journey.

(i) Navigational distance (in km) from New Delhi

to Mumbai.
(ii) Navigational time (in hrs) to reach from New
Delhi to Mumbai

Suppose he travels via NH48, then the distance travelled is 1402 km and time taken is
21h 31min. So in this example, Time and Distance represent two different modes of
measurement.

Here, we shall discuss and learn about a very special class of functions that ‘measure
difference’ which mathematicians were able to identify in the beginning of the 20th century.
In the mathematical literature, this special class is represented as “distance”. In the plane,
distance between two points is measured along the straight line joining them. Our
objective in this chapter is to illustrate through examples the different ways of measuring
difference (distance) between objects besides straight line measurements, so that
students can grasp the abstract nature of the subject.
 INTRODUCTION TO METRIC SPACES 5

To begin with, let us observe the fundamental properties of

straight line distance measured between two points in ℝ2 . From
high school geometry, we know that straight line distance
between points A and B is √(𝑥1 − 𝑦1 )2 + (𝑥2 − 𝑦2 )2 .

Properties of straight line distance

1. Measurement between distinct points is a positive real number.
2. Two points in a space are identical if and only if measurement between them is
zero.
3. Measurement is symmetric in nature i.e., distance measured along 𝐴 to 𝐵 is same
as measured along 𝐵 to 𝐴.
4. Measurement between two points is less than or equal
to the total distance taken when we travel via some
other point.

From the first two properties, we observe that

straight line distance is non-negative real 𝑨𝑪 ≤ 𝑨𝑩 + 𝑩𝑪
number.

How to generalize all these ideas under one notion so that the properties remain intact?
The solution is provided by real valued functions which measures difference. Such
functions are known as metric in the mathematical literature. Further since the prototype
for such functions is straight line distance, these functions are often regarded as distance
functions.

These functions were first considered in 1905, by the French mathematician Maurice
Frechet who thought of generalizing the notion of distances and extending them to arbitrary
sets. In his doctoral dissertation “Less Espaces Abstrait”, he introduced the concept of a
metric on a set.

Metric Space
Let 𝑋 be any set and let 𝑑: 𝑋 × 𝑋 → ℝ be a real valued function satisfying the following
properties:
P1. 𝑑(𝑥, 𝑦) ≥ 0 for all 𝑥, 𝑦 ∈ 𝑋;
P2. 𝑑(𝑥, 𝑦) = 0 ⇔ 𝑥 = 𝑦
P3. 𝑑(𝑥, 𝑦) = 𝑑(𝑦, 𝑥) for all 𝑥, 𝑦 ∈ 𝑋
P4. 𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) for all 𝑥, 𝑦, 𝑧 ∈ 𝑋

The function 𝑑 is called a metric on 𝑋 (sometimes the distance function on 𝑋). The
ordered pair (𝑋, 𝑑) is called a metric space. Thus a metric space consists of a non-empty
set equipped with a concept of distance (metric). If there is no ambiguity on the metric
considered, then we simply denote the metric space (𝑋, 𝑑) by 𝑋. We refer the elements in
𝑋 as points and 𝑑(𝑥, 𝑦) as the distance between the points 𝑥 and 𝑦.

Trivially, an empty function is the only metric on the empty set. Also, owing to condition
second, the only metric on a singleton set is the zero function.
 INTRODUCTION TO METRIC SPACES 6

4. EXAMPLES OF METRIC SPACES

Example 4.1 The Real Line ℝ
Let ℝ be the set of all real numbers and 𝑢: ℝ × ℝ → ℝ be a function defined as
𝑢(𝑥, 𝑦) = |𝑥 − 𝑦| ∀ 𝑥, 𝑦 ∈ ℝ.
Then we shall prove that 𝑢 is a metric on ℝ

First observe that by definition, (𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ ℝ . Therefore P1 holds.

For any 𝑥, 𝑦 in ℝ,
𝑢(𝑥, 𝑦) = 0 ⇔ |𝑥 − 𝑦| = 0 ⇔ 𝑥 = 𝑦.
Therefore P2 holds.
Again, for any 𝑥, 𝑦 in ℝ,
𝑢(𝑥, 𝑦) = |𝑥 − 𝑦| = |𝑦 − 𝑥| = 𝑢(𝑦, 𝑥).
Therefore P3 holds.

To see the triangle inequality (P4), suppose 𝑥, 𝑦, 𝑧 ∈ ℝ be any three points.

Consider
𝑢(𝑥, 𝑦) = |𝑥 − 𝑦|
= |(𝑥 − 𝑧) + (𝑧 − 𝑦)|
≤ |𝑥 − 𝑧| + |𝑧 − 𝑦|
= 𝑢(𝑥, 𝑧) + 𝑢(𝑧, 𝑦).
It follows that
𝑢(𝑥, 𝑦) ≤ 𝑢(𝑥, 𝑧) + 𝑢(𝑧, 𝑦) ∀ 𝑥, 𝑦, 𝑧 ∈ ℝ.
Thus all the four axioms are satisfied. Hence 𝑢 is a metric on ℝ and the ordered pair (ℝ, 𝑢)
is a metric space. The metric 𝑢 is called the usual or standard metric or Euclidean
metric on ℝ. █

Example 4.2 The Euclidean Metric on ℂ (Extension of Euclidean metric on ℝ)

Let ℂ be the set of all complex number and 𝑑 ∶ ℂ × ℂ → ℝ be a function defined as
𝑑(𝑧, 𝑧 ′ ) = |𝑧 − 𝑧 ′ | ∀ 𝑧, 𝑧 ′ ∈ ℂ .
Then 𝑑 is a metric on ℂ, called the usual metric or Euclidean Metric on ℂ. Of course,
𝑑 is an extension to ℂ × ℂ of the Euclidean metric 𝑢 on ℝ i.e.,
𝑢 = 𝑑|ℝ . █

Example 4.3 The Euclidean Plane ℝ2

Let 𝑋 = ℝ2 be the set of all ordered pairs of real numbers and
𝑑 ∶ ℝ2 × ℝ2 → ℝ be a function defined as
𝑑(𝑥, 𝑦) = √(𝑥1 − 𝑦1 )2 + (𝑥2 − 𝑦2 )2 ∀ 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ) ∈ ℝ2 .

We shall show that 𝑑 is a metric on ℝ2 . By definition,

𝑑(𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ ℝ2 .
 INTRODUCTION TO METRIC SPACES 7

For any 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ) ∈ ℝ2 ,

𝑑(𝑥, 𝑦) = 0 ⇔ √(𝑥1 − 𝑦1 )2 + (𝑥2 − 𝑦2 )2 = 0

⇔ 𝑥1 − 𝑦1 = 0 𝑎𝑛𝑑 𝑥2 − 𝑦2 = 0
⇔ 𝑥1 = 𝑦1 𝑎𝑛𝑑 𝑥2 = 𝑦2
⇔ 𝑥 = 𝑦.
For all 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ) ∈ ℝ ,
2

𝑑(𝑥, 𝑦) = √(𝑥1 − 𝑦1 )2 + (𝑥2 − 𝑦2 )2

= √(𝑦1 − 𝑥1 )2 + (𝑦2 − 𝑥2 )2
= 𝑑(𝑦, 𝑥).
Suppose 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ), 𝑧 = (𝑧1 , 𝑧2 ) ∈ ℝ2 be any three points.
Consider,

𝑑(𝑥, 𝑦) = √(𝑥1 − 𝑦1 )2 + (𝑥2 − 𝑦2 )2

= √[(𝑥1 − 𝑧1 ) + (𝑧1 − 𝑦1 )]2 + [(𝑥2 − 𝑧2 ) + (𝑧2 − 𝑦2 )]2

= √[𝑎 + 𝑏]2 + [𝑐 + 𝑑]2

where = 𝑥1 − 𝑧1 , 𝑏 = 𝑧1 − 𝑦1 , 𝑐 = 𝑥2 − 𝑧2 𝑎𝑛𝑑 𝑑 = 𝑧2 − 𝑦2 .
Applying Theorem A , we get

𝑑(𝑥, 𝑦) ≤ √𝑎2 + 𝑐 2 + √𝑏 2 + 𝑑 2

≤ √(𝑥1 − 𝑧1 )2 + (𝑥2 − 𝑧2 )2 + √(𝑧1 − 𝑦1 )2 + (𝑧2 − 𝑦2 )2

≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦).

Thus all the four axioms are satisfied. It follows that 𝑑 is a metric on ℝ2 and the ordered
pair (ℝ2 , 𝑑) is a metric space. The metric 𝑑 is called the Euclidean metric on ℝ2 , and the
metric space (ℝ2 , 𝑑) is called the 2-dimensional Euclidean Space ℝ𝟐 . ∎

Example 4.4 Taxi Cab Metric on ℝ2

Let ℝ2 be the set of all ordered pairs of real numbers and 𝑑 ∶ ℝ2 × ℝ2 → ℝ be a function
defined as
𝑑(𝑥, 𝑦) = |𝑥1 − 𝑦1 | + |𝑥2 − 𝑦2 | ∀ 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ) ∈ ℝ2 .
We shall show that 𝑑 is a metric on ℝ2 .

By definition, 𝑑 is a non-negative function and hence P1 holds.

For P2, consider any 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ) ∈ ℝ2, then

𝑑(𝑥, 𝑦) = 0 ⇔ |𝑥1 − 𝑦1 | + |𝑥2 − 𝑦2 | = 0
⇔ |𝑥1 − 𝑦1 | = 0 𝑎𝑛𝑑 |𝑥2 − 𝑦2 | = 0
⇔ 𝑥1 = 𝑦1 𝑎𝑛𝑑 𝑥2 = 𝑦2
⇔ (𝑥1 , 𝑥2 ) = (𝑦1 , 𝑦2 ) 𝑖. 𝑒., 𝑥=𝑦 .

Now again for P3, consider any 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ) ∈ ℝ2 ,

𝑑(𝑥, 𝑦) = |𝑥1 − 𝑦1 | + |𝑥2 − 𝑦2 | = |𝑦1 − 𝑥1 | + |𝑦2 − 𝑥2 | = 𝑑(𝑦, 𝑥) .

Thus P3 is satisfied.
 INTRODUCTION TO METRIC SPACES 8

To see triangle inequality (P4), let 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ), 𝑧 = (𝑧1 , 𝑧2 ) ∈ ℝ2 be any points in
ℝ3 . Then
𝑑(𝑥, 𝑦) = |𝑥1 − 𝑦1 | + |𝑥2 − 𝑦2 |
= |(𝑥1 − 𝑧1 ) + (𝑧1 − 𝑦1 )| + |(𝑥2 − 𝑧2 ) + (𝑧2 − 𝑦2 )|
≤ |𝑥1 − 𝑧1 | + |𝑧1 − 𝑦1 | + |𝑥2 − 𝑧2 | + |𝑧2 − 𝑦2 |
= |𝑥1 − 𝑧1 | + |𝑥2 − 𝑧2 | + |𝑧1 − 𝑦1 | + |𝑧2 − 𝑦2 |
= 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) .
Hence all the four axioms of a metric is satisfied by 𝑑, therefore 𝑑 is a metric on ℝ2 . █

Figure 2. Road Map of a City

Example 4.5 Maximum Metric on ℝ2

Let ℝ2 be the set of all ordered pairs of real numbers and 𝑑 ∶ ℝ2 × ℝ2 → ℝ be a function
defined as
𝑑(𝑥, 𝑦) = max {|𝑥1 − 𝑦1 |, |𝑥2 − 𝑦2 |} ∀ 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ) ∈ ℝ2 .
We shall show that 𝑑 is a metric on ℝ2 .

By definition, 𝑑 is a non-negative function and hence P1 holds.

For P2, consider any 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ) ∈ ℝ2,

𝑑(𝑥, 𝑦) = 0 ⇔ max {|𝑥1 − 𝑦1 |, |𝑥2 − 𝑦2 |} = 0
⇔ |𝑥1 − 𝑦1 | = 0 𝑎𝑛𝑑 |𝑥2 − 𝑦2 | = 0
⇔ 𝑥1 = 𝑦1 𝑎𝑛𝑑 𝑥2 = 𝑦2
⇔ (𝑥1 , 𝑥2 ) = (𝑦1 , 𝑦2 ) 𝑖. 𝑒., 𝑥=𝑦 .
For any 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ) ∈ ℝ2 ,
𝑑(𝑥, 𝑦) = max {|𝑥1 − 𝑦1 |, |𝑥2 − 𝑦2 |}
= max {|𝑦1 − 𝑥1 |, |𝑦2 − 𝑥2 |}
= 𝑑(𝑦, 𝑥) .
Thus P3 is satisfied.
To see triangle inequality (P4), let 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ), 𝑧 = (𝑧1 , 𝑧2 ) ∈ ℝ2 be any points in
ℝ3 . Consider
|𝑥1 − 𝑦1 | = |(𝑥1 − 𝑧1 ) + (𝑧1 − 𝑦1 )|
≤ |𝑥1 − 𝑧1 | + |𝑧1 − 𝑦1 |
 INTRODUCTION TO METRIC SPACES 9

≤ max {|𝑥1 − 𝑧1 |, |𝑥2 − 𝑧2 | } + max{|𝑧1 − 𝑦1 |, |𝑧2 − 𝑦2 |}

= 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)
𝑖. 𝑒., |𝑥1 − 𝑦1 | ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) − − − − − − − (𝑨)
Similarly,
|𝑥2 − 𝑦2 | ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) . − − − − − − − (𝑩)
From (𝐴) and (𝐵) it follows that
max {|𝑥1 − 𝑦1 |, |𝑥2 − 𝑦2 |} ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)
𝑖. 𝑒. , 𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) .
Hence the triangle inequality holds and therefore 𝑑 is a metric on ℝ2 . █

Example 4.6 Let 𝑋 = ℝ2 and 𝑑 ∶ ℝ2 × ℝ2 → ℝ be defined by

|𝑥 − 𝑦| 𝑖𝑓 𝑥 𝑎𝑛𝑑 𝑦 𝑎𝑟𝑒 𝑖𝑛 𝑡ℎ𝑒

𝑑(𝑥, 𝑦) = { 𝑠𝑎𝑚𝑒 𝑟𝑎𝑦 𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑜𝑟𝑖𝑔𝑖𝑛
|𝑥| + |𝑦| 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
where 𝑥 = (𝑥1 , 𝑥2 ) and 𝑦 = (𝑦1 , 𝑦2 ) ∈ ℝ2 . Show that 𝑑 is a metric
on ℝ2 . (Here |𝑥 − 𝑦| = 𝑢(𝑥, 𝑦) and |𝑥| = 𝑢(𝑥, 0) and 𝑢 is Euclidean
metric on ℝ2 . )

Proof: Clearly, 𝑑(𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ ℝ2 .

For any 𝑥, 𝑦 ∈ ℝ2
|𝑥 − 𝑦| 𝑖𝑓 𝑥 𝑎𝑛𝑑 𝑦 𝑎𝑟𝑒 𝑖𝑛 𝑡ℎ𝑒 𝑠𝑎𝑚𝑒 𝑟𝑎𝑦 𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑜𝑟𝑖𝑔𝑖𝑛
𝑑(𝑥, 𝑦) = {
|𝑥| + |𝑦| 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑢(𝑥, 𝑦) 𝑖𝑓 𝑥 𝑎𝑛𝑑 𝑦 𝑎𝑟𝑒 𝑖𝑛 𝑡ℎ𝑒 𝑠𝑎𝑚𝑒 𝑟𝑎𝑦 𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑜𝑟𝑖𝑔𝑖𝑛
={
𝑢(𝑥, 0) + 𝑢(𝑦, 0) 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑢(𝑦, 𝑥) 𝑖𝑓 𝑥 𝑎𝑛𝑑 𝑦 𝑎𝑟𝑒 𝑖𝑛 𝑡ℎ𝑒 𝑠𝑎𝑚𝑒 𝑟𝑎𝑦 𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑜𝑟𝑖𝑔𝑖𝑛
={
𝑢(𝑦, 0) + 𝑢(𝑥, 0) 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
|𝑦 − 𝑥| 𝑖𝑓 𝑥 𝑎𝑛𝑑 𝑦 𝑎𝑟𝑒 𝑖𝑛 𝑡ℎ𝑒 𝑠𝑎𝑚𝑒 𝑟𝑎𝑦 𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑜𝑟𝑖𝑔𝑖𝑛
={
|𝑦| + |𝑥| 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
= 𝑑(𝑦, 𝑥) .
By definition of 𝑑, observe that
𝑑(𝑥, 𝑦) ≥ |𝑥 − 𝑦| ∀ 𝑥, 𝑦 ∈ ℝ2 . − − − − − − − (𝐴) [∵ |𝑥 − 𝑦| ≤ |𝑥| + |𝑦|]
Thus for any 𝑥, 𝑦 ∈ ℝ2
𝑑(𝑥, 𝑦) = 0 ⇒ |𝑥 − 𝑦| = 0 ⇒𝑥 =𝑦.
Also, 𝑥 = 𝑦 implies that 𝑥 and 𝑦 are in the same ray from the origin and therefore
𝑑(𝑥, 𝑦) = |𝑥 − 𝑦| = 0 .
Finally to prove triangle inequality, consider any 𝑥, 𝑦, 𝑧 ∈ ℝ2 .
Case I 𝑥 and 𝑦 are in the same ray from the origin
Then
𝑑(𝑥, 𝑦) = |𝑥 − 𝑦|
= 𝑢(𝑥, 𝑦)
≤ 𝑢(𝑥, 𝑧) + 𝑢(𝑧, 𝑦)
= |𝑥 − 𝑧| + |𝑧 − 𝑦|
≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) [𝑈𝑠𝑖𝑛𝑔 (𝐴)]
 INTRODUCTION TO METRIC SPACES 10

Case II 𝑥 and 𝑦 are in the different ray from the origin.

Subcase I 𝑧 and 𝑥 are in different ray from the origin

Then
𝑑(𝑥, 𝑦) = |𝑥| + |𝑦|
= 𝑢(𝑥, 0) + 𝑢(0, 𝑦)
≤ 𝑢(𝑥, 0) + [𝑢(0, 𝑧) + 𝑢(𝑧, 𝑦)]
= |𝑥| + [|𝑧| + |𝑦 − 𝑧|]
= [|𝑥| + |𝑧|] + |𝑦 − 𝑧|
≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑦, 𝑧) [∵ |𝑦 − 𝑧| ≤ 𝑑(𝑦, 𝑧) … . (𝐴)]
= 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)

Subcase II 𝑧 and 𝑥 are in same ray from the origin.

Then 𝑧 and 𝑦 are in different ray from the origin. Therefore

𝑑(𝑥, 𝑦) = |𝑥| + |𝑦|
= 𝑢(𝑥, 0) + 𝑢(0, 𝑦)
≤ 𝑢(𝑥, 𝑧) + 𝑢(𝑧, 0) + 𝑢(0, 𝑦)
= |𝑥 − 𝑧| + |𝑧| + |𝑦|
= 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) .

Thus in all the cases triangle inequality is satisfied and hence 𝑑 is a metric on ℝ2 . ∎

Railway Metric

The metric given in Example 4 is

called the 𝑹𝒂𝒊𝒍𝒘𝒂𝒚 𝒎𝒆𝒕𝒓𝒊𝒄 as it can be
used to describe the following
situation (hypothetical).
Consider a proposed metro network
of India for 2030 where all the major
towns lie on some metro track
originating from Delhi (see adjoining
Figure). Thus on this network, one
can travel directly between any two
towns which lie on the same metro
track to Delhi. Otherwise first one has
to go Delhi and change to another
line.
 INTRODUCTION TO METRIC SPACES 11

Example 4.7 Metric on a Circle in ℝ𝟐

Consider any circle
𝒞 = {(𝑥, 𝑦) ∈ ℝ2 ∶ 𝑥 2 + 𝑦 2 = 𝑟 2 }
= {(𝑟 cos 𝜃 , 𝑟 sin 𝜃) ∶ 0 ≤ 𝜃 < 2𝜋} .
Then the function 𝑑: 𝒞 × 𝒞 → ℝ given by
𝑟|𝜃2 − 𝜃1 | 𝑖𝑓 0 ≤ |𝜃2 − 𝜃1 | ≤ 𝜋 𝑥 = (𝑟 cos 𝜃1 , 𝑟 sin 𝜃1 )
𝑑(𝑥, 𝑦) = { ∀ }∈𝒞
𝑟 (2𝜋 − |𝜃2 − 𝜃1 |) 𝑖𝑓 𝜋 < |𝜃2 − 𝜃1 | < 2𝜋 𝑦 = (𝑟 cos 𝜃2 , 𝑟 sin 𝜃2 )
is a metric on 𝒞.

Note that here 𝑑(𝑥, 𝑦) denotes the shortest distance along the circle from 𝑥 to 𝑦 i.e., length
of the minor arc between 𝑥 and 𝑦. ∎

Example 4.8 Navigation of Flights (Metric on a Sphere in ℝ𝟑 )

Have you ever wondered why an individual flight is flying on a particular route? Long
distance flight paths are designed in the most efficient way to get from point A to point B
on the other side of the world. The shortest distance between two points in Euclidean
space is the length of the straight line between them. But as we can travel only along the
surface of the earth (sphere) i.e., circular path, we have to design some way to find the
shortest path between them.

Figure 3. Flights following great circle routes.

In the above figure, an Air Canada flight from Toronto to Hong Kong looks like it is taking
a very long route, but it’s actually the shortest distance between the two cities.

On the sphere, distances between two points is measured along the surface of the sphere
(as opposed to a straight line through the sphere’s interior).
Through any two points on a sphere that are not directly opposite to each other,
there is a unique great circle (i.e. a circle around the surface of the earth whose center
coincides with the center of the earth). The length of the minor arc between the two points
on the great circle is the great-circle distance between the points.
 INTRODUCTION TO METRIC SPACES 12

Between two points that are directly opposite each other, there are infinitely many
great circles, but all great-circle arcs between antipodal points have the same length. i.e.
half the circumference of the circle.
In the adjoining figure, 𝑃 and 𝑄 are points
that are not directly opposite each other
and 𝑢, 𝑣 are points that are directly
opposite each other.
In the figure, it can be easily seen that
there is a unique great-circle passing
through 𝑃 and 𝑄. Also, two distinct great
circles passing through 𝑢 and 𝑣 are
Figure 4 illustrated.

Mostly long distance flights use great-circle routes to travel between two locations on the
globe. So the next time when you want to go east and the moving map on your flight
shows that you are heading north, remember that you are actually taking the shortest
path to your destination. And if you are lucky enough, you might even get to see the north
pole!

If 𝑆 is a sphere and 𝑃, 𝑄 belong to 𝑆, we can define 𝑑𝑠𝑝ℎ𝑒𝑟𝑒 (𝑃, 𝑄) to be the shortest distance
along a great circle joining 𝑃 and 𝑄 i.e.,

𝑑𝑠𝑝ℎ𝑒𝑟𝑒 (𝑃, 𝑄) = 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑡ℎ𝑒 𝑚𝑖𝑛𝑜𝑟 𝑎𝑟𝑐

= 𝑟. 𝜃
𝑝1 𝑞1 + 𝑝2 𝑞2 + 𝑝3 𝑞3
= 𝑟. cos −1 ( )
𝑟2

where 𝑟 is the radius of the great circle

and 𝜃 is the angle which the 𝑎𝑟𝑐 𝑃𝑄
subtends at the center of the great circle.

It is a good exercise to verify that 𝑑𝑠𝑝ℎ𝑒𝑟𝑒 is a metric on 𝑆. The metric 𝑑𝑠𝑝ℎ𝑒𝑟𝑒 is popularly
known as geodesic metric. This is the most useful metric to determine distance on the
Earth’s surface.
Example 4.9 Metric on the Set of Intervals of ℝ
Let ℐ denote the collection of closed intervals of ℝ of the type [𝑎, 𝑏] i.e.,
ℐ = {[𝑎, 𝑏] ∶ 𝑎, 𝑏 ∈ ℝ, 𝑎 ≤ 𝑏} .
The function 𝑑 ∶ ℐ × ℐ → ℝ given by
𝑑(𝐼, 𝐽) = max {|𝑐 − 𝑎|, |𝑑 − 𝑏|} ∀ 𝐼 = [𝑎, 𝑏] , 𝐽 = [𝑐, 𝑑] ∈ ℐ
is a metric on ℐ.
On the lines similar to Example 4.5, it is straightforward to check that P1, P2, P3 holds.
Now we need to check triangle inequality (P4), let 𝐼 = [𝑎, 𝑏], 𝐽 = [𝑐, 𝑑], 𝐾 = [𝑒, 𝑓] ∈ ℐ be any
three closed intervals of ℝ. Consider
𝑑(𝐼, 𝐽) = max {|𝑐 − 𝑎|, |𝑑 − 𝑏|}
= max {|(𝑐 − 𝑒) + (𝑒 − 𝑎)|, |(𝑑 − 𝑓) + (𝑓 − 𝑏)|}
≤ max {|𝑐 − 𝑒| + |𝑒 − 𝑎|, |𝑑 − 𝑓| + |𝑓 − 𝑏|} − − − − − − − − − (𝐴)
 INTRODUCTION TO METRIC SPACES 13

Now
|𝑐 − 𝑒| ≤ max {|𝑐 − 𝑒|, |𝑑 − 𝑓|} = 𝑑(𝐾, 𝐽) − − − − − (∗)
|𝑒 − 𝑎| ≤ max {|𝑒 − 𝑎|, |𝑓 − 𝑏|} = 𝑑(𝐼, 𝐾) − − − − − (∗∗)
Adding (∗) and (∗∗), we get
|𝑐 − 𝑒| + |𝑒 − 𝑎| ≤ 𝑑(𝐼, 𝐾) + 𝑑(𝐾, 𝐽) . − − − − − (𝐵)
Similarly,
|𝑑 − 𝑓| + |𝑓 − 𝑏| ≤ 𝑑(𝐼, 𝐾) + 𝑑(𝐾, 𝐽) . − − − − − (𝐶)
From (𝐵) and (𝐶),
max {|𝑐 − 𝑒| + |𝑒 − 𝑎|, |𝑑 − 𝑓| + |𝑓 − 𝑏|} ≤ 𝑑(𝐼, 𝐾) + 𝑑(𝐾, 𝐽) − − − − − (𝐷)
Now from (𝐴) and (𝐷), we conclude that
𝑑(𝐼, 𝐽) ≤ 𝑑(𝐼, 𝐾) + 𝑑(𝐾, 𝐽) .
Thus P4 holds and hence 𝑑 satisfies all the four axioms of a metric. █

In above example we restricted ourselves to only closed intervals of the type

[𝒂, 𝒃] and it is necessary to do so. Excluding intervals of the type (−∞, 𝒂), (−∞, 𝒂],
(𝒂, ∞), [𝒂, ∞) and (−∞, ∞) ensures that 𝒅 takes values in ℝ. Also, if we have
included all intervals of the type (𝒂, 𝒃), (𝒂, 𝒃] and [𝒂, 𝒃), then the function 𝒅
would have failed to satisfy P2. In fact
𝒅([𝒂, 𝒃], [𝒂, 𝒃)) = 𝟎 𝒃𝒖𝒕 [𝒂, 𝒃] ≠ [𝒂, 𝒃) .

Example 4.10 The space of all functions from [𝟎, 𝟏] to [𝟎, 𝟏]

Let ℱ be the set of all functions from [0,1] to [0,1] and 𝑑 be the real valued function on
ℱ × ℱ given by
𝑑(𝑓, 𝑔) = sup{|𝑓(𝑥) − 𝑔(𝑥)| ∶ 𝑥 ∈ [0, 1]} ∀ 𝑓, 𝑔 ∈ ℱ .
Then (ℱ, 𝑑) is a metric space.
To verify the fact that (ℱ, 𝑑) is a metric space, we shall first verify that 𝑑 is indeed a function
i.e., 𝑑 is well defined. For any 𝑓, 𝑔 ∈ ℱ,

𝑓(𝑥), 𝑔(𝑥) ∈ [0,1]

⇒ |𝑓(𝑥) − 𝑔(𝑥)| ≤ 1 ∀ 𝑥 ∈ [0,1]
⇒ {|𝑓(𝑥) − 𝑔(𝑥)| ∶ 𝑥 ∈ [0, 1]} 𝑖𝑠 𝑏𝑜𝑢𝑛𝑑𝑒𝑑 𝑎𝑏𝑜𝑣𝑒 [∵ 1 𝑖𝑠 𝑎𝑛 𝑢𝑝𝑝𝑒𝑟 𝑏𝑜𝑢𝑛𝑑]
⇒ sup{|𝑓(𝑥) − 𝑔(𝑥)| ∶ 𝑥 ∈ [0, 1]} 𝑒𝑥𝑖𝑠𝑡𝑠 .

It follows that 𝑑(𝑓, 𝑔) < ∞ ∀ 𝑓, 𝑔 ∈ ℱ. Hence 𝑑 is well defined.

Next to show that d is a metric on ℱ, we shall verify triangle inequality (other properties
can be easily verified). Consider any 𝑓, 𝑔, ℎ ∈ ℱ. Then for any 𝑥 ∈ [0, 1], we observe that

|𝑓(𝑥) − ℎ(𝑥)| ≤ sup {|𝑓(𝑥) − ℎ(𝑥)|} 𝑎𝑛𝑑 |ℎ(𝑥) − 𝑔(𝑥)| ≤ sup {|ℎ(𝑥) − 𝑔(𝑥)|} .
x∈[0,1] x∈[0,1]

Then
|𝑓(𝑥) − 𝑔(𝑥)| ≤ |𝑓(𝑥) − ℎ(𝑥)| + |ℎ(𝑥) − 𝑔(𝑥)|
 INTRODUCTION TO METRIC SPACES 14

≤ sup {|𝑓(𝑥) − ℎ(𝑥)|} + sup {|ℎ(𝑥) − 𝑔(𝑥)|}

x∈[0,1] x∈[0,1]

= 𝑑(𝑓, ℎ) + 𝑑(ℎ, 𝑔).

𝑖. 𝑒., |𝑓(𝑥) − 𝑔(𝑥)| ≤ 𝑑(𝑓, ℎ) + 𝑑(ℎ, 𝑔) 𝑓𝑜𝑟 𝑒𝑎𝑐ℎ 𝑥 ∈ [0,1]

Now, taking supremum over all 𝑥 ∈ [0, 1], we get

sup {|𝑓(𝑥) − 𝑔(𝑥)|} ≤ 𝑑(𝑓, ℎ) + 𝑑(ℎ, 𝑔)
x∈[0,1]

𝑖. 𝑒. , 𝑑(𝑓, 𝑔) ≤ 𝑑(𝑓, ℎ) + 𝑑(ℎ, 𝑔).

Hence triangle inequality follows, therefore 𝑑 is a metric on ℱ. ∎

Example 4.11 The Space of bounded functions

Let 𝑆 be any non-empty set and 𝔅(𝑆) be the set of all real (complex) valued bounded
functions defined on 𝑆 i.e.,
𝔅(𝑆) = {𝑓 | 𝑓 ∶ 𝑆 → ℝ 𝑜𝑟 ℂ 𝑎𝑛𝑑 sup|𝑓(𝑥)| < ∞} .
x∈S

Define a function 𝑑 ∶ 𝔅(𝑆) × 𝔅(𝑆) → ℝ as

𝑑(𝑓, 𝑔) = sup {|𝑓(𝑥) − 𝑔(𝑥)|} ∀ 𝑓, 𝑔 ∈ 𝔅(𝑆).
x∈S

Then 𝑑 is a well defined function.

For this we just need to show that 𝑑(𝑓, 𝑔) ∈ ℝ, whenever 𝑓, 𝑔 ∈ 𝔅(𝑆).

Now 𝑓, 𝑔 ∈ 𝔅(𝑆) implies that there exist positive real numbers 𝑀, 𝑁 such that

sup{|𝑓(𝑥)|} ≤ 𝑀 𝑎𝑛𝑑 sup{|𝑔(𝑥)|} ≤ 𝑁.

x∈S x∈S

For any 𝑥 ∈ 𝑆,
|𝑓(𝑥) − 𝑔(𝑥)| ≤ |𝑓(𝑥)| + |𝑔(𝑥)| ≤ 𝑀 + 𝑁 .

Taking supremum over all 𝑥 ∈ 𝑆, we get

sup{|𝑓(𝑥) − 𝑔(𝑥)|} ≤ 𝑀 + 𝑁 < ∞.
x∈S

Thus 𝑑(𝑓, 𝑔) < ∞ ∀ 𝑓, 𝑔 ∈ 𝔅(𝑆). Hence 𝑑 is well defined.

On the lines similar to that in Example 4.10, it can be shown that 𝑑 is a metric on 𝔅(𝑆). ∎

Example 4.12 The space of continuous functions on [𝒂, 𝒃]

Let 𝑋 be the set of all real valued continuous functions defined on the interval [𝑎, 𝑏]. Define
a function 𝑑 ∶ 𝑋 × 𝑋 → ℝ as
𝑑(𝑓, 𝑔) = sup {|𝑓(𝑥) − 𝑔(𝑥)|} ∀ 𝑓, 𝑔 ∈ 𝑋.
𝑥∈[𝑎,𝑏]

Since 𝑓 and 𝑔 are continous on [𝑎, 𝑏], therefore 𝑓 − 𝑔 is a continous function on [𝑎, 𝑏].
Consequently, |𝑓 − 𝑔| is a continous function on [𝑎, 𝑏]. Since on a closed and bounded
interval every continuous function is bounded, therefore |𝑓 − 𝑔| is bounded on [𝑎, 𝑏].
Hence sup {|𝑓(𝑥) − 𝑔(𝑥)|} exists. Thus 𝑑 is a well defined function. Geometrically, 𝑑(𝑓, 𝑔)
𝑥∈[𝑎,𝑏]

measures the largest vertical distance between the graph of 𝑓 and the graph of 𝑔 (see
Figure 5).
 INTRODUCTION TO METRIC SPACES 15

Figure 5. 𝑑(𝑓, 𝑔)represents the length of the dotted line segment.

It can be easily verified that d is a metric on 𝑋 (𝒔𝒊𝒎𝒊𝒍𝒂𝒓 𝒕𝒐 𝑬𝒙𝒂𝒎𝒑𝒍𝒆 𝟒. 𝟏𝟎). The metric space
(𝑋, 𝑑) is denoted by 𝒞[𝑎, 𝑏]. Since every continuous function on a closed and bounded
interval is bounded, therefore we have 𝒞[𝑎, 𝑏] ⊆ 𝔅[𝑎, 𝑏]. In fact the metric 𝑑 can be seen as
the one induced by the metric in Example 4.11. ∎

Note: Whatever we have discussed in this example is also true for the complex valued
functions defined on the closed interval [𝑎, 𝑏].

Example 4.13 Another metric on the set of all continuous functions

Let 𝑋 be the set of all continuous functions defined on the interval [𝑎, 𝑏]. We equip the set
𝑋 with another metric defined as
𝑎
𝑑(𝑓, 𝑔) = ∫ |𝑓(𝑥) − 𝑔(𝑥)| 𝑑𝑥 ∀ 𝑓, 𝑔 ∈ 𝑋 .
𝑏

As discussed in Example 4.12, continuity of 𝑓, 𝑔 implies continuity of |𝑓 − 𝑔|. Since integral

of a continuous function over a closed and bounded interval is always finite, therefore
𝑑(𝑓, 𝑔) exists for any 𝑓, 𝑔 ∈ 𝑋. Thus 𝑑 is a well defined function.

For any 𝑓, 𝑔 ∈ 𝑋

|𝑓(𝑥) − 𝑔(𝑥)| ≥ 0 ∀ 𝑥 ∈ [𝑎, 𝑏]

𝑎
⇒ ∫ |𝑓(𝑥) − 𝑔(𝑥)| 𝑑𝑥 ≥ 0
𝑏

⇒ 𝑑(𝑓, 𝑔) ≥ 0.

Hence P1 holds.
For P2, first observe that, 𝑑(𝑓, 𝑔) = 0 whenever 𝑓 = 𝑔.
On the other hand,
𝑑(𝑓, 𝑔) = 0
𝑎
⇒ ∫ |𝑓(𝑥) − 𝑔(𝑥)| 𝑑𝑥 = 0
𝑏

⇒ |𝑓(𝑥) − 𝑔(𝑥)| = 0 ∀ 𝑥 ∈ [𝑎, 𝑏] [∵ |𝑓 − 𝑔| 𝑖𝑠 𝑐𝑜𝑛𝑡𝑖𝑛𝑢𝑜𝑢𝑠 𝑎𝑛𝑑 |𝑓 − 𝑔| ≥ 0]

⇒ 𝑓=𝑔
Thus 𝑑(𝑓, 𝑔) = 0 if and only if 𝑓 = 𝑔.
It is easy to see that 𝑑(𝑓, 𝑔) = 𝑑(𝑔, 𝑓) ∀ 𝑓, 𝑔 ∈ 𝑋.
Now for triangle inequality, consider 𝑓, 𝑔, ℎ ∈ 𝑋 be any three elements, then
𝑎
𝑑(𝑓, 𝑔) = ∫ |𝑓(𝑥) − 𝑔(𝑥)| 𝑑𝑥
𝑏
 INTRODUCTION TO METRIC SPACES 16

𝑎
= ∫ |(𝑓(𝑥) − ℎ(𝑥)) + (ℎ(𝑥) − 𝑔(𝑥))| 𝑑𝑥
𝑏
𝑎 𝑎 𝑎
≤ ∫ [|𝑓(𝑥) − ℎ(𝑥)| + |ℎ(𝑥) − 𝑔(𝑥)|] 𝑑𝑥 [ ∵ 𝑓 ≤ 𝑔 ⇒ ∫ 𝑓 𝑑𝑥 ≤ ∫ 𝑔 𝑑𝑥]
𝑏 𝑏 𝑏
𝑎 𝑎
= ∫ |𝑓(𝑥) − ℎ(𝑥)| 𝑑𝑥 + ∫ |ℎ(𝑥) − 𝑔(𝑥)| 𝑑𝑥
𝑏 𝑏

= 𝑑(𝑓, ℎ) + 𝑑(ℎ, 𝑔).

Thus all the axioms for a metric is satisfied. Hence 𝑑 is indeed a metric on 𝑋. ∎
Geometrically, the measure of the distance between 𝑓 and 𝑔 i.e., 𝑑(𝑓, 𝑔) represents the
area between their graphs.

Figure 6. 𝑑(𝑓, 𝑔) represents the area of the shaded portion.

Note: There exists non-negative non-zero functions whose integral is zero. Thus the set
of all functions cannot be made a metric space with the above metric. Consider the function
𝑓 ∶ [0,2] → ℝ given by
0 𝑖𝑓 𝑥 ≠ 1
𝑓(𝑥) = { .
1 𝑖𝑓 𝑥 = 1
2
Then it is can be seen that ∫0 𝑓(𝑥) 𝑑𝑥 = 0.

Example 4.14 Discrete Metric

Let 𝑋 be a non-empty set and 𝑑 ∶ 𝑋 × 𝑋 → ℝ be a function defined as

1 𝑖𝑓 𝑥 ≠ 𝑦
𝑑(𝑥, 𝑦) = { .
0 𝑖𝑓 𝑥 = 𝑦
Then 𝑑 is a metric on 𝑋.
Clearly, 𝑑(𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑋.

By definition of 𝑑, 𝑑(𝑥, 𝑦) = 0 ⇔ 𝑥 = 𝑦 .

Let 𝑥, 𝑦 ∈ 𝑋 be any two points.

𝐼𝑓 𝑥 = 𝑦 , 𝑡ℎ𝑒𝑛 𝑑(𝑥, 𝑦) = 0 = 𝑑(𝑦, 𝑥) 𝑎𝑛𝑑
𝐼𝑓 𝑥 ≠ 𝑦 , 𝑡ℎ𝑒𝑛 𝑑(𝑥, 𝑦) = 1 = 𝑑(𝑦, 𝑥) .

Thus in either case 𝑑(𝑥, 𝑦) = 𝑑(𝑦, 𝑥).

Let 𝑥, 𝑦, 𝑧 ∈ 𝑋 be any three points.

Case 1. 𝑥 = 𝑦
In this case 𝑑(𝑥, 𝑦) = 0 and therefore
𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦). [𝐵𝑦 𝑑𝑒𝑓. 𝑜𝑓 𝑑 , 𝑑(𝑥, 𝑧) ≥ 0 𝑎𝑛𝑑 𝑑(𝑧, 𝑦) ≥ 0]

Case 2. 𝑥 ≠ 𝑦
In this case, 𝑑(𝑥, 𝑦) = 1. Since 𝑥 ≠ 𝑦, we have three possibilities :
i. 𝑧=𝑥, 𝑧≠𝑦 ⇒ 𝑑(𝑥, 𝑧) = 0 𝑎𝑛𝑑 𝑑(𝑧, 𝑦) = 1
 INTRODUCTION TO METRIC SPACES 17

ii. 𝑧≠𝑥, 𝑧=𝑦 ⇒ 𝑑(𝑥, 𝑧) = 1 𝑎𝑛𝑑 𝑑(𝑧, 𝑦) = 0

iii. 𝑧≠𝑥, 𝑧≠𝑦 ⇒ 𝑑(𝑥, 𝑧) = 1 𝑎𝑛𝑑 𝑑(𝑧, 𝑦) = 1

In each case we observe that 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) ≥ 1. Therefore 1 = 𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦).

Hence in both Case 1 and Case 2, triangle inequality is satisfied. Thus 𝑑 is a metric on 𝑋
and is called the discrete metric on 𝑋. ∎

Given any set 𝑿 (≠ ∅), there always exists a metric on 𝑿, viz., discrete metric.

Example 4.15 The Euclidean Plane ℝ𝒏

Let ℝ𝑛 = {(𝑥1 , . . . , 𝑥𝑛 ) ∶ 𝑥𝑖 ∈ ℝ, 𝑖 = 1,2, . . . , 𝑛} be the set of all 𝑛-tuples of real numbers and
𝑑 ∶ ℝ𝑛 × ℝ𝑛 → ℝ be a function defined as
1
𝑛 2

𝑑(𝑥, 𝑦) = (∑(𝑥𝑖 − 𝑦𝑖 )2 ) ∀ 𝑥 = (𝑥1 , … , 𝑥𝑛 ), 𝑦 = (𝑦1 , … , 𝑦𝑛 ) ∈ ℝ𝑛 .

𝑖=1

Then d is a metric on ℝ . 𝑛

By definition, 𝑑 is a non-negative real valued function.

For any 𝑥 = (𝑥1 , . . . , 𝑥𝑛 ), 𝑦 = (𝑦1 , . . . , 𝑦𝑛 ) 𝑖𝑛 ℝ𝑛 ,
𝑑(𝑥, 𝑦) = 0
1
𝑛 2

⇔ (∑(𝑥𝑖 − 𝑦𝑖 )2 ) = 0
𝑖=1

⇔ 𝑥𝑖 − 𝑦𝑖 = 0 , ∀ 𝑖 = 1,2, … , 𝑛
⇔ 𝑥𝑖 = 𝑦𝑖 , ∀ 𝑖 = 1,2, … , 𝑛
⇔ 𝑥 = 𝑦.

For any 𝑥 = (𝑥1 , . . . , 𝑥𝑛 ), 𝑦 = (𝑦1 , . . . , 𝑦𝑛 ) ∈ ℝ𝑛 ,

1
𝑛 2

𝑑(𝑥, 𝑦) = (∑(𝑥𝑖 − 𝑦𝑖 )2 )
𝑖=1
1
𝑛 2

= (∑(𝑦𝑖 − 𝑥𝑖 )2 ) = 𝑑(𝑦, 𝑥).

𝑖=1

Let 𝑥 = (𝑥1 , . . . , 𝑥𝑛 ), 𝑦 = (𝑦1 , . . . , 𝑦𝑛 ), 𝑧 = (𝑧1 , . . . , 𝑧𝑛 ) ∈ ℝ𝑛 be any three points. Then

1 1
𝑛 2 𝑛 2

𝑑(𝑥, 𝑦) = (∑(𝑥𝑖 − 𝑦𝑖 )2 ) = (∑{(𝑥𝑖 − 𝑧𝑖 ) + (𝑧𝑖 − 𝑦𝑖 )}2 )

𝑖=1 𝑖=1

Let 𝑎𝑖 = 𝑥𝑖 − 𝑧𝑖 , 𝑏𝑖 = 𝑧𝑖 − 𝑦𝑖 for 𝑖 = 1,2, . . . , 𝑛. This implies

1
𝑛 2

𝑑(𝑥, 𝑦) = (∑(𝑎𝑖 + 𝑏𝑖 )2 ) − − − − − − − −(𝑨)

𝑖=1
 INTRODUCTION TO METRIC SPACES 18

Now from Cauchy-Schwarz inequality we have,

1 1
𝑛 𝑛 2 𝑛 2

∑ 𝑎𝑖 𝑏𝑖 ≤ (∑ 𝑎𝑖2 ) (∑ 𝑏𝑖2 )
𝑖=1 𝑖=1 𝑖=1
1 1
𝑛 𝑛 2 𝑛 2

⇒ 2 ∑ 𝑎𝑖 𝑏𝑖 ≤ 2 (∑ 𝑎𝑖2 ) (∑ 𝑏𝑖2 )
𝑖=1 𝑖=1 𝑖=1

Now adding ∑𝑛𝑖=1 𝑎𝑖2 + ∑𝑛𝑖=1 𝑏𝑖2 on both sides we get,

1 1
𝑛 𝑛 𝑛 𝑛 𝑛 𝑛 2 𝑛 2

∑ 𝑎𝑖2 + ∑ 𝑏𝑖2 + 2 ∑ 𝑎𝑖 𝑏𝑖 ≤ ∑ 𝑎𝑖2 + ∑ 𝑏𝑖2 + 2 (∑ 𝑎𝑖2 ) (∑ 𝑏𝑖2 )

𝑖=1 𝑖=1 𝑖=1 𝑖=1 𝑖=1 𝑖=1 𝑖=1
1 1
𝑛 𝑛 𝑛 𝑛 2 𝑛 2

⇒ ∑(𝑎𝑖2 + 𝑏𝑖2 + 2𝑎𝑖 𝑏𝑖 ) ≤ ∑ 𝑎𝑖2 + ∑ 𝑏𝑖2 + 2 (∑ 𝑎𝑖2 ) (∑ 𝑏𝑖2 )

𝑖=1 𝑖=1 𝑖=1 𝑖=1 𝑖=1
1 1 2
𝑛 𝑛 2 𝑛 2

⇒ ∑(𝑎𝑖 + 𝑏𝑖 )2 ≤ [(∑ 𝑎𝑖2 ) + (∑ 𝑏𝑖2 ) ]

𝑖=1 𝑖=1 𝑖=1

1 1 1
𝑛 2 𝑛 2 𝑛 2

⇒ (∑(𝑎𝑖 + 𝑏𝑖 )2 ) ≤ (∑ 𝑎𝑖2 ) + (∑ 𝑏𝑖2 ) − − − (𝑩)

𝑖=1 𝑖=1 𝑖=1

Using (𝑩) in equation (𝑨), we get

1 1
𝑛 2 𝑛 2

𝑑(𝑥, 𝑦) ≤ (∑ 𝑎𝑖2 ) + (∑ 𝑏𝑖2 )

𝑖=1 𝑖=1
1 1
𝑛 2 𝑛 2

= (∑(𝑥𝑖 − 𝑧𝑖 )2 ) + (∑(𝑧𝑖 − 𝑦𝑖 )2 )
𝑖=1 𝑖=1

= 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦).

It follows that
𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) ∀ 𝑥, 𝑦, 𝑧 ∈ ℝ𝑛 .
Therefore all the axioms are satisfied. Hence 𝑑 is a metric on ℝ𝑛 .

Alternate proof of triangle inequality by induction

Let 𝑥 = (𝑥1 , . . . , 𝑥𝑛 ), 𝑦 = (𝑦1 , . . . , 𝑦𝑛 ), 𝑧 = (𝑧1 , . . . , 𝑧𝑛 ) ∈ ℝ𝑛 . We need to show that

𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)

𝑛 𝑛 𝑛

𝑖. 𝑒., √∑(𝑥𝑖 − 𝑦𝑖 )2 ≤ √∑(𝑥𝑖 − 𝑧𝑖 )2 + √∑(𝑧𝑖 − 𝑦𝑖 )2 .

𝑖=1 𝑖=1 𝑖=1

If 𝑛 = 1, then

1 1 1

√∑(𝑥𝑖 − 𝑦𝑖 )2 = |𝑥1 − 𝑦1 | ≤ |𝑥1 − 𝑧1 | + |𝑧1 − 𝑦1 | = √∑(𝑥𝑖 − 𝑧𝑖 )2 + √∑(𝑧𝑖 − 𝑦𝑖 )2 .

𝑖=1 𝑖=1 𝑖=1
 INTRODUCTION TO METRIC SPACES 19

Hence for 𝑛 = 1, 𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦).

Suppose it is true for 𝑛 = 𝑘 i.e.,

𝑘 𝑘 𝑘

√∑(𝑥𝑖 − 𝑦𝑖 )2 ≤ √∑(𝑥𝑖 − 𝑧𝑖 )2 + √∑(𝑧𝑖 − 𝑦𝑖 )2 .

𝑖=1 𝑖=1 𝑖=1

Now we shall prove the claim for 𝑛 = 𝑘 + 1. Consider

𝑘+1 𝑘

∑(𝑥𝑖 − 𝑦𝑖 )2 = ∑(𝑥𝑖 − 𝑦𝑖 )2 + (𝑥𝑘+1 − 𝑦𝑘+1 )2 .

𝑖=1 𝑖=1

Using induction hypothesis , we get

2
𝑘+1 𝑘 𝑘

∑(𝑥𝑖 − 𝑦𝑖 )2 ≤ √∑(𝑥𝑖 − 𝑧𝑖 )2 + √∑(𝑧𝑖 − 𝑦𝑖 )2 + [(𝑥𝑘+1 − 𝑧𝑘+1 ) + (𝑧𝑘+1 − 𝑦𝑘+1 )]2

𝑖=1 𝑖=1 𝑖=1
[ ]

By setting = √∑𝑘𝑖=1(𝑥𝑖 − 𝑧𝑖 )2 , 𝑏 = √∑𝑘𝑖=1(𝑧𝑖 − 𝑦𝑖 )2 , 𝑐 = (𝑥𝑘+1 − 𝑧𝑘+1 ) and 𝑑 = (𝑧𝑘+1 − 𝑦𝑘+1 ),

we get
𝑘+1

∑(𝑥𝑖 − 𝑦𝑖 )2 ≤ (𝑎 + 𝑏)2 + (𝑐 + 𝑑)2

𝑖=1
2
≤ [√(𝑎2 + 𝑐 2 ) + √(𝑏 2 + 𝑑 2 )] [𝑻𝒉𝒆𝒐𝒓𝒆𝒎 𝑨]
2
𝑘 𝑘

= √∑(𝑥𝑖 − 𝑧𝑖 )2 + (𝑥𝑘+1 − 𝑧𝑘+1 )2 + √∑(𝑧𝑖 − 𝑦𝑖 )2 + (𝑧𝑘+1 − 𝑦𝑘+1 )2

𝑖=1 𝑖=1
( )
2
𝑘+1 𝑘+1

= √∑(𝑥𝑖 − 𝑧𝑖 )2 + √∑(𝑧𝑖 − 𝑦𝑖 )2
𝑖=1 𝑖=1
( )

Taking square root on both sides, we get

𝑘+1 𝑘+1 𝑘+1

√∑(𝑥𝑖 − 𝑦𝑖 )2 ≤ √∑(𝑥𝑖 − 𝑧𝑖 )2 + √∑(𝑧𝑖 − 𝑦𝑖 )2 .

𝑖=1 𝑖=1 𝑖=1

Therefore for 𝑛 = 𝑘 + 1, 𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦).

Hence by the principle of mathematical induction,

𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) 𝑖𝑠 𝑡𝑟𝑢𝑒 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑛 ∈ ℕ.

The metric 𝑑 is called the Euclidean metric on ℝ𝑛 , and the metric space (ℝ𝑛 , 𝑑) is called
the n-dimensional Euclidean Space ℝ𝒏 . For 𝑛 = 2, the n-dimensional Euclidean
Space ℝ𝒏 is called the Euclidean Plane or Cartesian Plane ℝ𝟐 (Example 4.3). ∎
 INTRODUCTION TO METRIC SPACES 20

Example 4.16 The n-dimensional Product Space ℝ𝒏

Let ℝ𝑛 = {(𝑥1 , . . . , 𝑥𝑛 ) ∶ 𝑥𝑖 ∈ ℝ, 𝑖 = 1,2, . . . , 𝑛} be the set of all 𝑛-tuples of real numbers and
𝑑∞ ∶ ℝ𝑛 × ℝ𝑛 → ℝ be a function defined as
𝑑∞ (𝑥, 𝑦) = max {|𝑥𝑖 − 𝑦𝑖 | ∶ 1 ≤ 𝑖 ≤ 𝑛 } ∀ 𝑥 = (𝑥1 , … , 𝑥𝑛 ), 𝑦 = (𝑦1 , … , 𝑦𝑛 ) ∈ ℝ𝑛 .
Then 𝑑∞ is a metric on ℝ𝒏 .
By definition, 𝑑∞ is a real valued non-negative function.
For any 𝑥 = (𝑥1 , . . . , 𝑥𝑛 ), 𝑦 = (𝑦1 , . . . , 𝑦𝑛 ) ∈ ℝ𝑛 ,
𝑑∞ (𝑥, 𝑦) = 0
⇔ max {|𝑥𝑖 − 𝑦𝑖 | ∶ 1 ≤ 𝑖 ≤ 𝑛 } = 0
⇔ 𝑥𝑖 − 𝑦𝑖 = 0 , 𝑖 = 1,2, … , 𝑛
⇔ 𝑥𝑖 = 𝑦𝑖 , 𝑖 = 1,2, … , 𝑛
⇔ 𝑥 = 𝑦.
For any 𝑥 = (𝑥1 , . . . , 𝑥𝑛 ), 𝑦 = (𝑦1 , . . . , 𝑦𝑛 ) ∈ ℝ𝑛 ,
𝑑∞ (𝑥, 𝑦) = max {|𝑥𝑖 − 𝑦𝑖 | ∶ 1 ≤ 𝑖 ≤ 𝑛 }
= max {|𝑦𝑖 − 𝑥𝑖 | ∶ 1 ≤ 𝑖 ≤ 𝑛 }
= 𝑑∞ (𝑦, 𝑥).
Consider any 𝑥 = (𝑥1 , . . . , 𝑥𝑛 ), 𝑦 = (𝑦1 , . . . , 𝑦𝑛 ), 𝑧 = (𝑧1 , . . . , 𝑧𝑛 ) ∈ ℝ𝑛 .
Now for each 𝑖 (1 ≤ 𝑖 ≤ 𝑛),
|𝑥𝑖 − 𝑦𝑖 | = |(𝑥𝑖 − 𝑧𝑖 ) + (𝑧𝑖 − 𝑦𝑖 )|
≤ |𝑥𝑖 − 𝑧𝑖 | + |𝑧𝑖 − 𝑦𝑖 |
≤ max {|𝑥𝑖 − 𝑧𝑖 | ∶ 1 ≤ 𝑖 ≤ 𝑛 } + max {|𝑧𝑖 − 𝑦𝑖 | ∶ 1 ≤ 𝑖 ≤ 𝑛 }
= 𝑑∞ (𝑥, 𝑧) + 𝑑∞ (𝑧, 𝑦)
Taking maximum over 𝑖, we have
max {|𝑥𝑖 − 𝑦𝑖 | ∶ 1 ≤ 𝑖 ≤ 𝑛 } ≤ 𝑑∞ (𝑥, 𝑧) + 𝑑∞ (𝑧, 𝑦)
⇒ 𝑑∞ (𝑥, 𝑦) ≤ 𝑑∞ (𝑥, 𝑧) + 𝑑∞ (𝑧, 𝑦).

It follows that 𝑑∞ (𝑥, 𝑦) ≤ 𝑑∞ (𝑥, 𝑧) + 𝑑∞ (𝑧, 𝑦) ∀ 𝑥, 𝑦, 𝑧 ∈ ℝ𝑛 .

Thus all the four axioms are satisfied. Hence 𝑑∞ is a metric on ℝ𝑛 and the ordered pair
(ℝ𝑛 , 𝑑∞ ) is a metric space. The metric 𝑑∞ is called the Product metric on ℝ𝑛 , and the
metric space (ℝ𝑛 , 𝑑∞ ) is called the n-dimensional Product Space ℝ𝒏 . For 𝑛 = 2, the metric
in Example 4.5 is a special case of 𝑑∞ metric. █

Example 4.17 Generalized Taxi Cab Metric for ℝ𝒏

Let ℝ𝑛 = {(𝑥1 , . . . , 𝑥𝑛 ) ∶ 𝑥𝑖 ∈ ℝ, 𝑖 = 1,2, . . . , 𝑛} be the set of all n-tuples of real numbers and
𝑑 ∶ ℝ𝑛 × ℝ𝑛 → ℝ be a function defined as
𝑛

𝑑(𝑥, 𝑦) = ∑|𝑥𝑖 − 𝑦𝑖 | ∀ 𝑥 = (𝑥1 , … , 𝑥𝑛 ), 𝑦 = (𝑦1 , … , 𝑦𝑛 ) ∈ ℝ𝑛 .

𝑖=1

Then 𝑑 is a metric on ℝ . 𝒏

By definition, 𝑑 is a non-negative real valued function.

 INTRODUCTION TO METRIC SPACES 21

For any 𝑥 = (𝑥1 , . . . , 𝑥𝑛 ), 𝑦 = (𝑦1 , . . . , 𝑦𝑛 ) ∈ ℝ𝑛 ,

𝑑(𝑥, 𝑦) = 0
𝑛

⇔ ∑|𝑥𝑖 − 𝑦𝑖 | = 0
𝑖=1

⇔ |𝑥𝑖 − 𝑦𝑖 | = 0 , 𝑖 = 1,2, … , 𝑛
⇔ 𝑥𝑖 − 𝑦𝑖 = 0 , 𝑖 = 1,2, … , 𝑛
⇔ 𝑥𝑖 = 𝑦𝑖 , 𝑖 = 1,2, … , 𝑛
⇔ 𝑥 = 𝑦.
For any 𝑥 = (𝑥1 , . . . , 𝑥𝑛 ), 𝑦 = (𝑦1 , . . . , 𝑦𝑛 ) ∈ ℝ𝑛 ,
𝑛 𝑛

𝑑(𝑥, 𝑦) = ∑|𝑥𝑖 − 𝑦𝑖 | = ∑|𝑦𝑖 − 𝑥𝑖 | = 𝑑(𝑦, 𝑥).

𝑖=1 𝑖=1

Consider any 𝑥 = (𝑥1 , . . . , 𝑥𝑛 ), 𝑦 = (𝑦1 , . . . , 𝑦𝑛 ), 𝑧 = (𝑧1 , . . . , 𝑧𝑛 ) ∈ ℝ𝑛 . Now

𝑛

𝑑(𝑥, 𝑦) = ∑|𝑥𝑖 − 𝑦𝑖 |
𝑖=1
𝑛

= ∑|(𝑥𝑖 − 𝑧𝑖 ) + (𝑧𝑖 − 𝑦𝑖 )|
𝑖=1
𝑛

≤ ∑(|𝑥𝑖 − 𝑧𝑖 | + |𝑧𝑖 − 𝑦𝑖 |)
𝑖=1
𝑛 𝑛

= ∑|𝑥𝑖 − 𝑧𝑖 | + ∑|𝑧𝑖 − 𝑦𝑖 |
𝑖=1 𝑖=1

= 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)

It follows that 𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) ∀ 𝑥, 𝑦, 𝑧 ∈ ℝ𝑛 .

Thus all the four axioms are satisfied. Hence 𝑑 is a metric on ℝ𝑛 and the ordered pair
(ℝ𝑛 , 𝑑) is a metric space. Taxi cab metric on ℝ2 in Example 4.4 is a special case of the
above metric 𝑑. █

Metric Derived from other Metrics

Example 4.18 Let (𝑋, 𝑑) be a metric space and 𝑒: 𝑋 × 𝑋 → ℝ be a function defined as

𝑑(𝑥, 𝑦)
𝑒(𝑥, 𝑦) = ∀ 𝑥, 𝑦 ∈ 𝑋 .
1 + 𝑑(𝑥, 𝑦)
Then 𝑒 is a metric on 𝑋.
Proof: Since (𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑋 , therefore 𝑒(𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑋.
For any 𝑥, 𝑦 ∈ 𝑋,
𝑑(𝑥, 𝑦) 𝑑(𝑦, 𝑥)
𝑒(𝑥, 𝑦) = = = 𝑒(𝑦, 𝑥)
1 + 𝑑(𝑥, 𝑦) 1 + 𝑑(𝑦, 𝑥)
Also,
𝑑(𝑥, 𝑦)
𝑒(𝑥, 𝑦) = 0 ⇔ = 0 ⇔ 𝑑(𝑥, 𝑦) = 0 ⇔ 𝑥 = 𝑦
1 + 𝑑(𝑥, 𝑦)
 INTRODUCTION TO METRIC SPACES 22

To prove triangle inequality, consider any 𝑥, 𝑦, 𝑧 ∈ 𝑋

𝑑(𝑥, 𝑦)
𝑒(𝑥, 𝑦) =
1 + 𝑑(𝑥, 𝑦)
1
=1−
1 + 𝑑(𝑥, 𝑦)
𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)
1 ⇒ 1 + 𝑑(𝑥, 𝑦) ≤ 1 + 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)
≤1−
1 + 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) 1 1
⇒ ≥
[ 1 + 𝑑(𝑥, 𝑦) 1 + 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) ]
𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)
=
1 + 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)
𝑑(𝑥, 𝑧) 𝑑(𝑧, 𝑦)
= +
1 + 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) 1 + 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)
𝑑(𝑥, 𝑧) 𝑑(𝑧, 𝑦)
≤ +
1 + 𝑑(𝑥, 𝑧) 1 + 𝑑(𝑧, 𝑦)
= 𝑒(𝑥, 𝑧) + 𝑒(𝑧, 𝑦) .

Hence 𝑒 is a metric on 𝑋. ∎

Example 4.19 Let (𝑋, 𝑑) be a metric space. The function 𝑑 ∗ ∶ 𝑋 × 𝑋 → ℝ given by

𝑑 ∗ (𝑥, 𝑦) = min {1, 𝑑(𝑥, 𝑦)} ∀ 𝑥, 𝑦 ∈ 𝑋

is also a metric on 𝑋.

Proof: Since (𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑋 , therefore 𝑑 ∗ (𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑋.

For any 𝑥, 𝑦 ∈ 𝑋,
𝑑 ∗ (𝑥, 𝑦) = 0 ⇔ min {1, 𝑑(𝑥, 𝑦)} = 0
⇔ 𝑑(𝑥, 𝑦) = 0
⇔𝑥 =𝑦.

Also,
𝑑 ∗ (𝑥, 𝑦) = min {1, 𝑑(𝑥, 𝑦)} = min {1, 𝑑(𝑦, 𝑥)} = 𝑑 ∗ (𝑦, 𝑥)

Now to verify triangle inequality, let 𝑥, 𝑦, 𝑧 ∈ 𝑋 be any points. Then

𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) .

Since min {1, 𝑑(𝑥, 𝑦)} ≤ 𝑑(𝑥, 𝑦), it follows that

min {1, 𝑑(𝑥, 𝑦)} ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) − − − − − −(A)

Also,
min {1, 𝑑(𝑥, 𝑦)} ≤ 1 − − − − − −(𝐵)
From (𝐴) and (𝐵),
min {1, 𝑑(𝑥, 𝑦)} ≤ min {1, 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)} − − − − − −(𝐶)

Claim: min {1, 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)} ≤ 𝑑 ∗ (𝑥, 𝑧) + 𝑑 ∗ (𝑧, 𝑦).

Case I 𝑑(𝑥, 𝑧) ≥ 1 or 𝑑(𝑧, 𝑦) ≥ 1

We discuss the case for 𝑑(𝑥, 𝑧) ≥ 1. Then

 INTRODUCTION TO METRIC SPACES 23

min {1, 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)} = 1

= min {1, 𝑑(𝑥, 𝑧)}
≤ min {1, 𝑑(𝑥, 𝑧)} + min {1, 𝑑(𝑧, 𝑦)}
= 𝑑 ∗ (𝑥, 𝑧) + 𝑑 ∗ (𝑧, 𝑦).

Similar calculation can be done for 𝑑(𝑧, 𝑦) ≥ 1.

Case II 𝑑(𝑥, 𝑧) < 1 and 𝑑(𝑧, 𝑦) < 1

Then
𝑑 ∗ (𝑥, 𝑧) = min {1, 𝑑(𝑥, 𝑧)} = 𝑑(𝑥, 𝑧) 𝑎𝑛𝑑 𝑑 ∗ (𝑧, 𝑦) = min {1, 𝑑(𝑧, 𝑦)} = 𝑑(𝑧, 𝑦).
Now since
min {1, 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)} ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) ,

it follows that
min {1, 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)} ≤ 𝑑 ∗ (𝑥, 𝑧) + 𝑑 ∗ (𝑧, 𝑦) .

Therefore from both the cases we conclude that

min {1, 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)} ≤ 𝑑 ∗ (𝑥, 𝑧) + 𝑑 ∗ (𝑧, 𝑦) − − − − − − − −(𝐷)

From (𝐶) and (𝐷), we get

𝑑 ∗ (𝑥, 𝑦) = min {1, 𝑑(𝑥, 𝑦)} ≤ 𝑑 ∗ (𝑥, 𝑧) + 𝑑 ∗ (𝑧, 𝑦) .

Thus 𝑑 ∗ is a metric on 𝑋. █

Observe that the metrics defined in Example 4.18 and Example 4.19 are
bounded metrics (𝒆(𝒙, 𝒚) ≤ 𝟏 𝒂𝒏𝒅 𝒅∗ (𝒙, 𝒚) ≤ 𝟏 ∀ 𝒙, 𝒚 ∈ 𝑿 ). Thus given any
metric on any given set 𝑿, we can always define a bounded metric on 𝑿.

Example 4.20 Dilation of a Metric

Let (𝑋, 𝑑) be a metric space and 𝛼 > 0 be any real number. Then 𝑒: 𝑋 × 𝑋 → ℝ defined as
𝑒(𝑥, 𝑦) = 𝛼 𝑑(𝑥, 𝑦)
is a metric on 𝑋.

Proof: Since 𝛼 > 0 and 𝑑(𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑋, it follows that 𝑒(𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑋.

For any 𝑥, 𝑦 ∈ 𝑋,
𝑒(𝑥, 𝑦) = 0
⇔ 𝛼 𝑑(𝑥, 𝑦) = 0
⇔ 𝑑(𝑥, 𝑦) = 0 [∵ 𝛼 > 0]
⇔ 𝑥=𝑦

Also, 𝑒(𝑥, 𝑦) = 𝛼 𝑑(𝑥, 𝑦) = 𝛼 𝑑(𝑦, 𝑥) = 𝑒(𝑦, 𝑥) .

For triangle inequality consider any 𝑥, 𝑦, 𝑧 ∈ 𝑋

𝑒(𝑥, 𝑦) = 𝛼 𝑑(𝑥, 𝑦)
≤ 𝛼 [𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)]
= 𝛼 𝑑(𝑥, 𝑧) + 𝛼 𝑑(𝑧, 𝑦)
= 𝑒(𝑥, 𝑧) + 𝑒(𝑧, 𝑦) .

Hence 𝑒 is a metric on 𝑋. ∎
 INTRODUCTION TO METRIC SPACES 24

Example 4.21 Metric derived using an injective real valued function

Let 𝑍 be a set, (𝑋, 𝑑) be a metric space and 𝑓: 𝑍 → 𝑋 be an injective function. Then the
function 𝑒 ∶ 𝑍 × 𝑍 → ℝ defined as
𝑒(𝑥, 𝑦) = 𝑑(𝑓(𝑥), 𝑓(𝑦))
is a metric on 𝑍.

Proof: Since 𝑑(𝑓(𝑥), 𝑓(𝑦)) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑍, therefore 𝑒(𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑍.

For any 𝑥, 𝑦 ∈ 𝑍,
𝑒(𝑥, 𝑦) = 0 ⇔ 𝑑(𝑓(𝑥), 𝑓(𝑦)) = 0 ⇔ 𝑓(𝑥) = 𝑓(𝑦) ⇔ 𝑥 = 𝑦 .

Also,
𝑒(𝑥, 𝑦) = 𝑑(𝑓(𝑥) , 𝑓(𝑦)) = 𝑑(𝑓(𝑦), 𝑓(𝑥)) = 𝑒(𝑦, 𝑥) .

For triangle inequality, consider any 𝑥, 𝑦, 𝑧 ∈ 𝑍. Then

𝑒(𝑥, 𝑦) = 𝑑(𝑓(𝑥), 𝑓(𝑦))

≤ 𝑑(𝑓(𝑥), 𝑓(𝑧)) + 𝑑(𝑓(𝑧), 𝑓(𝑦))
= 𝑒(𝑥, 𝑧) + 𝑒(𝑧, 𝑦) .

Hence 𝑒 is a metric on 𝑍. ∎

Note that we cannot drop the condition of injectiveness of the given

function 𝒇 in Example 4.21. The function 𝒆 may fail to be a metric.

For example consider the many one function 𝒇: ℝ → ℝ given by

𝒙 ↦ 𝒙𝟐 .
Then the function 𝒆 ∶ ℝ × ℝ → ℝ given by
𝒆(𝒙, 𝒚) = |𝒙𝟐 − 𝒚𝟐 | ∀ 𝒙, 𝒚 ∈ ℝ
is not a metric on ℝ. It fails to satisfy P2 . In fact
𝒆(𝟏, −𝟏) = 𝟎 𝒃𝒖𝒕 𝟏 ≠ −𝟏 .

Example 4.22 Let 𝑋 ⊆ ℝ be any subset and 𝑓 ∶ 𝑋 → ℝ be an injective (one-one) function.

Define a function 𝑒 ∶ 𝑋 × 𝑋 → ℝ as
𝑒(𝑥, 𝑦) = |𝑓(𝑥) − 𝑓(𝑦)| ∀ 𝑥, 𝑦 ∈ 𝑋.
Then 𝑒 is a metric on 𝑋.
Proof: To show that 𝑒 is a metric on 𝑋, we shall just show that 𝑒(𝑥, 𝑦) = 0 ⇔ 𝑥 = 𝑦. Other
axioms can be proved easily using the basic properties of real numbers and the modulus
function.
Now for any 𝑥, 𝑦 ∈ 𝑋, consider
𝑒(𝑥, 𝑦) = 0
⇔ |𝑓(𝑥) − 𝑓(𝑦)| = 0
⇔ 𝑓(𝑥) = 𝑓(𝑦)
⇔ 𝑥=𝑦 [∵ 𝑓 𝑖𝑠 𝑎𝑛 𝑖𝑛𝑗𝑒𝑐𝑡𝑖𝑣𝑒 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛. ]
Thus 𝑒(𝑥, 𝑦) = 0 ⇔ 𝑥 = 𝑦.
 INTRODUCTION TO METRIC SPACES 25

Alternatively, we can deduce that 𝑒 is a metric on 𝑋 from Example 4.21 by simply putting
𝑍 = 𝑋 and taking 𝑑 to be usual metric on 𝑋 i.e., 𝑑(𝑥, 𝑦) = |𝑥 − 𝑦| ∀𝑥, 𝑦 ∈ 𝑋. █

Example 4.23 The function 𝑒 ∶ ℝ × ℝ → ℝ is a metric on ℝ in each of the following cases:

For any 𝑥, 𝑦 ∈ ℝ,
1. 𝑒(𝑥, 𝑦) = |𝑥 3 − 𝑦 3 |
2. 𝑒(𝑥, 𝑦) = |𝑒 𝑥 − 𝑒 𝑦 |

Example 4.24 Inverse Metric on ℝ+

Since the function 𝑓 ∶ ℝ+ → ℝ+ given by 𝑥 ↦ 𝑥 −1 is an injective function, therefore the
function 𝑑 ∶ ℝ+ × ℝ+ → ℝ given by
𝑑(𝑥, 𝑦) = |𝑥 −1 − 𝑦 −1 | ∀ 𝑥, 𝑦 ∈ ℝ+
is a metric on ℝ+ . The metric 𝑑 is called the inverse metric on ℝ+ . [Hint : Take 𝑍 = ℝ+ ,
𝑋 = ℝ+ , 𝑑 usual metric on ℝ+ and 𝑓: ℝ+ → ℝ+ as 𝑓(𝑥) = 𝑥 −1 in Example 4.21]

Example 4.25 Inverse Metric on ℕ

̃ = ℕ ∪ {∞}

Consider the set ℕ̃ = ℕ ∪ {∞} (here symbol "∞" represents a point outside ℕ).
Define 𝑑 ∶ ℕ × ℕ → ℝ as
̃ ̃
𝑑(𝑚, 𝑛) = |𝑚−1 − 𝑛−1 |

𝑑(𝑛, ∞) = 𝑑(∞, 𝑛) = 𝑛−1 ∀ 𝑚, 𝑛 ∈ ℕ .

𝑑(∞, ∞) = 0 }
Then 𝑑 is a metric on ℕ
̃.

Proof: Define 𝑓 ∶ ℕ
̃ → ℝ as
𝑛−1 𝑖𝑓 𝑛 ∈ ℕ
𝑓(𝑛) = { .
0 𝑖𝑓 𝑛 = ∞
Then the function 𝑑 ∶ ℕ ̃ → ℝ can be rewritten as
̃×ℕ

𝑑(𝑚, 𝑛) = |𝑓(𝑚) − 𝑓(𝑛)| ̃ .

∀ 𝑚, 𝑛 ∈ ℕ
̃ , it is enough to prove that 𝑓 is
In view of Example 4.21, to show that 𝑑 is a metric on ℕ
an injective function.
Claim: 𝑓 is an injective function.
Let 𝑥, 𝑦 ∈ ℕ
̃ be any two distinct points i.e., 𝑥 ≠ 𝑦.

Case I 𝑥, 𝑦 ∈ ℕ
Since 𝑥 ≠ 𝑦, therefore 𝑥 −1 ≠ 𝑦 −1 and hence 𝑓(𝑥) ≠ 𝑓(𝑦).
Case II 𝑥 ∈ ℕ and 𝑦 = ∞
Then 𝑓(𝑥) = 𝑥 −1 ≠ 0 and 𝑓(𝑦) = 𝑓(∞) = 0. Therefore 𝑓(𝑥) ≠ 𝑓(𝑦).
Case III 𝑥 = ∞ and 𝑦 ∈ ℕ
Then 𝑓(𝑥) = 𝑓(∞) = 0 and 𝑓(𝑦) = 𝑦 −1 ≠ 0. Therefore 𝑓(𝑥) ≠ 𝑓(𝑦).

̃ . It follows that 𝑓 is
Thus in all the above cases we get 𝑓(𝑥) ≠ 𝑓(𝑦), whenever 𝑥 ≠ 𝑦 , 𝑥, 𝑦 ∈ ℕ
an injective function on ℕ
̃ . Hence 𝑑 is a metric on ℕ
̃ (see 𝑬𝒙𝒂𝒎𝒑𝒍𝒆 𝟒. 𝟐𝟏). █
 INTRODUCTION TO METRIC SPACES 26

Example 4.26 Sum of two metrics is again a metric

Let 𝑋 be any set and 𝑑1 , 𝑑2 be any two metrics on 𝑋. Then the function 𝑒: 𝑋 × 𝑋 → ℝ
defined as
𝑒(𝑥, 𝑦) = 𝑑1 (𝑥, 𝑦) + 𝑑2 (𝑥, 𝑦)
is a metric on 𝑋.

Proof: Since 𝑑1 (𝑥, 𝑦) ≥ 0 𝑎𝑛𝑑 𝑑2 (𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑋, therefore 𝑒(𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑋.

For any 𝑥, 𝑦 ∈ 𝑋,
𝑒(𝑥, 𝑦) = 0
⇔ 𝑑1 (𝑥, 𝑦) + 𝑑2 (𝑥, 𝑦) = 0
⇔ 𝑑1 (𝑥, 𝑦) = 0, 𝑑2 (𝑥, 𝑦) = 0 [∵ 𝑑1 , 𝑑2 ≥ 0]
⇔ 𝑥=𝑦
Also,
𝑒(𝑥, 𝑦) = 𝑑1 (𝑥, 𝑦) + 𝑑2 (𝑥, 𝑦) = 𝑑1 (𝑦, 𝑥) + 𝑑2 (𝑦, 𝑥) = 𝑒(𝑦, 𝑥) .

For triangle inequality, consider any 𝑥, 𝑦, 𝑧 ∈ 𝑋. Then

𝑒(𝑥, 𝑦) = 𝑑1 (𝑥, 𝑦) + 𝑑2 (𝑥, 𝑦)

≤ 𝑑1 (𝑥, 𝑧) + 𝑑1 (𝑧, 𝑦) + 𝑑2 (𝑥, 𝑧) + 𝑑2 (𝑧, 𝑦)
= [𝑑1 (𝑥, 𝑧) + 𝑑2 (𝑥, 𝑧)] + [𝑑1 (𝑧, 𝑦) + 𝑑2 (𝑧, 𝑦)]
= 𝑒(𝑥, 𝑧) + 𝑒(𝑧, 𝑦) .
Hence 𝑒 is a metric on 𝑋. ∎

Example 4.27 Suppose (𝑋, 𝑑) is a metric space and 𝑓: 𝑋 → ℝ be a function. Show that
the function (𝑎, 𝑏) ↦ 𝑑(𝑎, 𝑏) + |𝑓(𝑎) − 𝑓(𝑏)| is a metric on 𝑋.

Proof: Define 𝑒: 𝑋 × 𝑋 → ℝ as
𝑒(𝑎, 𝑏) = 𝑑(𝑎, 𝑏) + |𝑓(𝑎) − 𝑓(𝑏)| ∀ 𝑎, 𝑏 ∈ 𝑋 .

Since 𝑑(𝑎, 𝑏) ≥ 0 𝑎𝑛𝑑 |𝑓(𝑎) − 𝑓(𝑏)| ≥ 0 ∀ 𝑎, 𝑏 ∈ 𝑋, therefore 𝑒(𝑎, 𝑏) ≥ 0 ∀ 𝑎, 𝑏 ∈ 𝑋.

For any 𝑎, 𝑏 ∈ 𝑋,
𝑒(𝑎, 𝑏) = 0 ⇒ 𝑑(𝑎, 𝑏) + |𝑓(𝑎) − 𝑓(𝑏)| = 0
⇒ 𝑑(𝑎, 𝑏) = 0 𝑎𝑛𝑑 |𝑓(𝑎) − 𝑓(𝑏)| = 0

Now 𝑑(𝑎, 𝑏) = 0 implies 𝑎 = 𝑏. [Note that |𝑓(𝑎) − 𝑓(𝑏)| = 0 need not imply 𝑎 = 𝑏]
On the other hand,
𝑎 = 𝑏 ⇒ 𝑑(𝑎, 𝑏) = 0 𝑎𝑛𝑑 𝑓(𝑎) = 𝑓(𝑏)
⇒ 𝑑(𝑎, 𝑏) + |𝑓(𝑎) − 𝑓(𝑏)| = 0
⇒ 𝑒(𝑎, 𝑏) = 0 .
For any 𝑎, 𝑏 ∈ 𝑋,

𝑒(𝑎, 𝑏) = 𝑑(𝑎, 𝑏) + |𝑓(𝑎) − 𝑓(𝑏)|

= 𝑑(𝑏, 𝑎) + |𝑓(𝑏) − 𝑓(𝑎)|
= 𝑒(𝑏, 𝑎) .

Finally to prove triangle inequality, consider any 𝑎, 𝑏, 𝑐 ∈ 𝑋,

𝑒(𝑎, 𝑏) = 𝑑(𝑎, 𝑏) + |𝑓(𝑎) − 𝑓(𝑏)|

≤ 𝑑(𝑎, 𝑐) + 𝑑(𝑐, 𝑏) + |𝑓(𝑎) − 𝑓(𝑐)| + |𝑓(𝑐) − 𝑓(𝑏)|
= 𝑑(𝑎, 𝑐) + |𝑓(𝑎) − 𝑓(𝑐)| + 𝑑(𝑐, 𝑏) + |𝑓(𝑐) − 𝑓(𝑏)|
= 𝑒(𝑎, 𝑐) + 𝑒(𝑐, 𝑏) .

Hence the triangle inequality. Thus 𝑒 is a metric on 𝑋. █

 INTRODUCTION TO METRIC SPACES 27

Looking at the function 𝒆 on 𝑿 × 𝑿 in the above example, one might

be tempted to think that to deduce that 𝒆 is a metric on 𝑿, we can
use the fact sum of two metrics is again a metric. But it is
worthwhile to note here that we cannot deduce such conclusion
directly as 𝒇 is not given injective and thus the function
(𝒂, 𝒃) ↦ |𝒇(𝒂) − 𝒇(𝒃)|
may fail to be metric.

Example 4.28 Let 𝑋 be any set and 𝑑1 , 𝑑2 be any two metrics on 𝑋. Then the function
𝑒: 𝑋 × 𝑋 → ℝ defined as
2 2
𝑒(𝑥, 𝑦) = √(𝑑1 (𝑥, 𝑦)) + (𝑑2 (𝑥, 𝑦))
is a metric on 𝑋.

Proof: Since 𝑑1 (𝑥, 𝑦) ≥ 0 𝑎𝑛𝑑 𝑑2 (𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑋, therefore 𝑒(𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑋.

For any 𝑥, 𝑦 ∈ 𝑋,
𝑒(𝑥, 𝑦) = 0
2 2
⇔ √(𝑑1 (𝑥, 𝑦)) + (𝑑2 (𝑥, 𝑦))

⇔ 𝑑1 (𝑥, 𝑦) = 0 𝑎𝑛𝑑 𝑑2 (𝑥, 𝑦) = 0

⇔ 𝑥=𝑦 .
Also,
2 2
𝑒(𝑥, 𝑦) = √(𝑑1 (𝑥, 𝑦)) + (𝑑2 (𝑥, 𝑦))

2 2
= √(𝑑1 (𝑦, 𝑥)) + (𝑑2 (𝑦, 𝑥))

= 𝑒(𝑦, 𝑥).

To prove triangle inequality we will use the following inequality

2
[√(𝑤 2 + 𝑦 2 ) + √(𝑥 2 + 𝑧 2 )] ≥ [𝑤 + 𝑥]2 + [𝑦 + 𝑧]2 ∀ 𝑤, 𝑥, 𝑦, 𝑧 ∈ ℝ [𝑇ℎ𝑒𝑜𝑟𝑒𝑚 𝐴] − − − −(∗)

Now consider any 𝑥, 𝑦, 𝑧 ∈ 𝑋

2 2
𝑒(𝑥, 𝑦) = √(𝑑1 (𝑥, 𝑦)) + (𝑑2 (𝑥, 𝑦))

2 2
≤ √(𝑑1 (𝑥, 𝑧) + 𝑑1 (𝑧, 𝑦)) + (𝑑2 (𝑥, 𝑧) + 𝑑2 (𝑧, 𝑦)) [𝐵𝑦 𝑡𝑟𝑖𝑎𝑛𝑔𝑙𝑒 𝑖𝑛𝑒𝑞𝑢𝑎𝑙𝑖𝑡𝑦 𝑓𝑜𝑟 𝑑1 𝑎𝑛𝑑 𝑑2 ]

2 2 2 2
≤ √(𝑑1 (𝑥, 𝑧)) + (𝑑2 (𝑧, 𝑦)) + √(𝑑1 (𝑥, 𝑧)) + (𝑑2 (𝑧, 𝑦)) . [𝑈𝑠𝑖𝑛𝑔 (∗)]

Hence 𝑒 is a metric on 𝑋. ∎

Example 4.29 Let 𝑋 be any set and 𝑑1 , 𝑑2 be any two metrics on 𝑋. Then the function
𝑒: 𝑋 × 𝑋 → ℝ defined as
𝑒(𝑥, 𝑦) = max {𝑑1 (𝑥, 𝑦), 𝑑2 (𝑥, 𝑦)}
is a metric on 𝑋.
 INTRODUCTION TO METRIC SPACES 28

Proof: Since 𝑑1 (𝑥, 𝑦) ≥ 0 𝑎𝑛𝑑 𝑑2 (𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑋, therefore 𝑒(𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑋.

For any 𝑥, 𝑦 ∈ 𝑋,
𝑒(𝑥, 𝑦) = 0
⇔ max{𝑑1 (𝑥, 𝑦), 𝑑2 (𝑥, 𝑦)}
⇔ 𝑑1 (𝑥, 𝑦) = 0 𝑎𝑛𝑑 𝑑2 (𝑥, 𝑦) = 0
⇔ 𝑥=𝑦 .

Also,
𝑒(𝑥, 𝑦) = max{𝑑1 (𝑥, 𝑦), 𝑑2 (𝑥, 𝑦)}
= max{𝑑1 (𝑦, 𝑥), 𝑑2 (𝑦, 𝑥)}
= 𝑒(𝑦, 𝑥) .

For triangle inequality, consider any 𝑥, 𝑦, 𝑧 ∈ 𝑋. Then

𝑑1 (𝑥, 𝑦) ≤ 𝑑1 (𝑥, 𝑧) + 𝑑1 (𝑧, 𝑦) ≤ 𝑒(𝑥, 𝑧) + 𝑒(𝑧, 𝑦)

𝑎𝑛𝑑 𝑑2 (𝑥, 𝑦) ≤ 𝑑2 (𝑥, 𝑧) + 𝑑2 (𝑧, 𝑦) ≤ 𝑒(𝑥, 𝑧) + 𝑒(𝑧, 𝑦)

Therefore it follows that

max {𝑑1 (𝑥, 𝑦), 𝑑2 (𝑥, 𝑦)} ≤ 𝑒(𝑥, 𝑧) + 𝑒(𝑧, 𝑦)
⇒ 𝑒(𝑥, 𝑦) ≤ 𝑒(𝑥, 𝑧) + 𝑒(𝑧, 𝑦) .
Hence 𝑒 is a metric on 𝑋. ∎

If 𝑿 is any non-empty set and 𝒅𝟏 , 𝒅𝟐 are two metrics on 𝑿. Then the function 𝒆 ∶
𝑿 × 𝑿 → ℝ defined as
𝒆(𝒙, 𝒚) = 𝐦𝐢𝐧 {𝒅𝟏 (𝒙, 𝒚), 𝒅𝟐 (𝒙, 𝒚)}
need not be a metric on 𝑿.

Consider 𝑋 = ℝ and 𝑑1 , 𝑑2 be two metrics on ℝ given by

𝑑1 (𝑥, 𝑦) = | 𝑥 − 𝑦 |
} ∀ 𝑥, 𝑦 ∈ ℝ .
𝑑2 (𝑥, 𝑦) = |𝑥 3 − 𝑦 3 |

Let 𝑒 ∶ 𝑋 × 𝑋 → ℝ be the function defined as

𝑒(𝑥, 𝑦) = min {𝑑1 (𝑥, 𝑦), 𝑑2 (𝑥, 𝑦)} ∀ 𝑥, 𝑦 ∈ ℝ .
Claim: 𝑒 is not a metric on ℝ.
We will show that 𝑒 does not satisfy triangle inequality. Let
1 1
𝑥= , 𝑦=2 , 𝑧= .
4 2
Then
7 1 3
𝑑1 (𝑥, 𝑦) = ≈ 1.75 , 𝑑1 (𝑥, 𝑧) = ≈ 0.25 𝑎𝑛𝑑 𝑑1 (𝑧, 𝑦) = ≈ 1.5 ;
4 4 2
511 7 63
𝑑2 (𝑥, 𝑦) = ≈ 7.98 , 𝑑2 (𝑥, 𝑧) = ≈ 0.11 𝑎𝑛𝑑 𝑑2 (𝑧, 𝑦) = ≈ 7.88
64 64 8
7 7 3
⇒ 𝑒(𝑥, 𝑦) = ≈ 1.75 , 𝑒(𝑥, 𝑧) = ≈ 0.11 𝑎𝑛𝑑 𝑒(𝑧, 𝑦) = ≈ 1.5 .
4 64 2
Thus 𝑒(𝑥, 𝑦) > 𝑒(𝑥, 𝑧) + 𝑒(𝑧, 𝑦) and consequently, P4 is not satisfied. Hence 𝑒 is not a
metric on ℝ. █
 INTRODUCTION TO METRIC SPACES 29

5. METRIC SUBSPACES AND METRIC SUPERSPACES

Definition 5.1 (𝑺𝒖𝒃𝒔𝒑𝒂𝒄𝒆 𝒐𝒇 𝒂 𝑴𝒆𝒕𝒓𝒊𝒄 𝑺𝒑𝒂𝒄𝒆)

Let (𝑋, 𝑑) be a metric space and 𝑌 ⊆ 𝑋 be any nonempty set. Let 𝑑𝑌 be the restriction of 𝑑
to the set 𝑌 × 𝑌 i.e, 𝑑𝑌 ∶ 𝑌 × 𝑌 → ℝ given as

𝑑𝑌 (𝑥, 𝑦) = 𝑑(𝑥, 𝑦) ∀ 𝑥, 𝑦 ∈ 𝑌 𝑖. 𝑒., 𝑑𝑌 = 𝑑|𝑌×𝑌 .

Since 𝑑 is a metric on 𝑋, therefore the mapping 𝑑𝑌 is a metric on 𝑌 and is called the

relative metric induced on 𝒀 by 𝒅. The space (𝑌, 𝑑𝑌 ) is called the metric subspace of
the metric space (𝑋, 𝑑).

Example 5.2 Consider the real line (ℝ, 𝑢) with usual metric 𝑢 given by
𝑢(𝑥, 𝑦) = |𝑥 − 𝑦| ∀ 𝑥, 𝑦 ∈ ℝ

and the complex plane (ℂ, 𝑑) with usual metric 𝑑 given by

𝑑(𝑧, 𝑧 ′ ) = |𝑧 − 𝑧 ′ | ∀ 𝑧, 𝑧 ′ ∈ ℂ

From definition of 𝑢 and 𝑑, it is clear that

𝑢(𝑥, 𝑦) = 𝑑(𝑥, 𝑦) ∀ 𝑥, 𝑦 ∈ ℝ 𝑖. 𝑒. , 𝑢 = 𝑑|ℝ

Therefore (ℝ, 𝑢) is a metric subspace of the complex plane (ℂ, 𝑑).

Example 5.3 Any subset of ℝ (𝑓𝑜𝑟 𝑒𝑔. ℚ , ℝ ∖ ℚ, ℕ, ℤ, [0,1], [0,1), (0,1), (0,1) ∪ (2, 3) 𝑒𝑡𝑐. ) is
a metric subspace of the real line (ℝ, u).

Definition 5.4 (𝑺𝒖𝒑𝒆𝒓𝒔𝒑𝒂𝒄𝒆 𝒐𝒇 𝒂 𝑴𝒆𝒕𝒓𝒊𝒄 𝑺𝒑𝒂𝒄𝒆)

If 𝑌 is a metric space and 𝑋 is a subset of 𝑌, we can induce metric on 𝑋 by restricting the

metric of 𝑌 on 𝑋. The question arises, can we do the reverse thing?
Suppose (𝑋, 𝑑) is a metric space and 𝑌 is a proper superset of 𝑋. Can we define a metric
on Y that is an extension of 𝑑? The answer is yes and it can be done in several ways, but
we shall elaborate only one method.

Consider a metric space (𝑋, 𝑑) and 𝑌 ⊋ 𝑋 . Since 𝑌 ∖ 𝑋 is non-empty, take any metric 𝑒 on
𝑌 ∖ 𝑋. Choose and fix two points 𝑎 ∈ 𝑋 and 𝑏 ∈ 𝑌\𝑋.

Now define 𝐷 ∶ 𝑌 × 𝑌 → ℝ as

𝑑(𝑥, 𝑦) 𝑖𝑓 𝑥, 𝑦 ∈ 𝑋

𝑒(𝑥, 𝑦) 𝑖𝑓 𝑥, 𝑦 ∈ 𝑌 ∖ 𝑋
𝐷(𝑥, 𝑦) = ,
𝑑(𝑥, 𝑎) + 1 + 𝑒(𝑏, 𝑦) 𝑖𝑓 𝑥 ∈ 𝑋 𝑎𝑛𝑑 𝑦 ∈ 𝑌 ∖ 𝑋
{𝑒(𝑥, 𝑏) + 1 + 𝑑(𝑎, 𝑦) 𝑖𝑓 𝑥 ∈ 𝑌 ∖ 𝑋 𝑎𝑛𝑑 𝑦 ∈ 𝑋

Then 𝐷 is a metric on 𝑌 such that

𝑑 = 𝐷|𝑋×𝑋 .

First observe that 𝐷(𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑌. Thus P1 holds.

For P2, consider any 𝑥, 𝑦 ∈ 𝑌.

Case I 𝑥, 𝑦 ∈ 𝑋
Then
𝐷(𝑥, 𝑦) = 0 ⇔ 𝑑(𝑥, 𝑦) = 0 [∵ 𝑥, 𝑦 ∈ 𝑋 ⇒ 𝐷(𝑥, 𝑦) = 𝑑(𝑥, 𝑦)]
⇔ 𝑥=𝑦
 INTRODUCTION TO METRIC SPACES 30

Case II 𝑥, 𝑦 ∈ 𝑌 ∖ 𝑋
Then
𝐷(𝑥, 𝑦) = 0 ⇔ 𝑒(𝑥, 𝑦) = 0 [∵ 𝑥, 𝑦 ∈ 𝑌 ∖ 𝑋 ⇒ 𝐷(𝑥, 𝑦) = 𝑒(𝑥, 𝑦)]
⇔ 𝑥=𝑦

Case III 𝑥 ∈ 𝑋 and 𝑦 ∈ 𝑌 ∖ 𝑋

In this case 𝑥 ≠ 𝑦 and 𝐷(𝑥, 𝑦) ≠ 0.

Case IV 𝑥 ∈ 𝑌 ∖ 𝑋 and 𝑦 ∈ 𝑋
In this case 𝑥 ≠ 𝑦 and 𝐷(𝑥, 𝑦) ≠ 0.

From all the above cases it follows that

𝐷(𝑥, 𝑦) = 0 ⇔ 𝑥 = 𝑦 ∀ 𝑥, 𝑦 ∈ 𝑌.
Thus P2 holds.

To see that P3 holds, consider any 𝑥, 𝑦 ∈ 𝑌.

Case I 𝑥, 𝑦 ∈ 𝑋.
Then
𝐷(𝑥, 𝑦) = 𝑑(𝑥, 𝑦) = 𝑑(𝑦, 𝑥) = 𝐷(𝑦, 𝑥) .
Case II 𝑥, 𝑦 ∈ 𝑌 ∖ 𝑋.
Then
𝐷(𝑥, 𝑦) = 𝑒(𝑥, 𝑦) = 𝑒(𝑦, 𝑥) = 𝐷(𝑦, 𝑥) .

Case III 𝑥 ∈ 𝑋 and 𝑦 ∈ 𝑌 ∖ 𝑋

Then
𝐷(𝑥, 𝑦) = 𝑑(𝑥, 𝑎) + 1 + 𝑒(𝑏, 𝑦)
= 𝑒(𝑏, 𝑦) + 1 + 𝑑(𝑥, 𝑎)
= 𝑒(𝑦, 𝑏) + 1 + 𝑑(𝑎, 𝑥)
= 𝐷(𝑦, 𝑥) .

Case IV 𝑥 ∈ 𝑌 ∖ 𝑋 and 𝑦 ∈ 𝑋
Similar to Case III.

Hence P3 is verified.
Now for triangle Inequality consider any 𝑥, 𝑦, 𝑧 ∈ 𝑌. We have

Case I 𝑥, 𝑦 ∈ 𝑋

If 𝑧 ∈ 𝑋, then
𝐷(𝑥, 𝑦) = 𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) = 𝐷(𝑥, 𝑧) + 𝐷(𝑧, 𝑦).
If 𝑧 ∈ 𝑌 ∖ 𝑋, then
𝐷(𝑥, 𝑦) = 𝑑(𝑥, 𝑦)
≤ 𝑑(𝑥, 𝑎) + 𝑑(𝑎, 𝑦) [∵ 𝑎 ∈ 𝑋]
≤ [𝑑(𝑥, 𝑎) + 1 + 𝑒(𝑏, 𝑧)] + [𝑒(𝑧, 𝑏) + 1 + 𝑑(𝑎, 𝑦)]
= 𝐷(𝑥, 𝑧) + 𝐷(𝑧, 𝑦)

Case II 𝑥, 𝑦 ∈ 𝑌 ∖ 𝑋

If 𝑧 ∈ 𝑋, then
𝐷(𝑥, 𝑦) = 𝑒(𝑥, 𝑦)
≤ 𝑒(𝑥, 𝑏) + 𝑒(𝑏, 𝑦) [∵ 𝑎 ∈ 𝑋]
≤ [𝑒(𝑥, 𝑏) + 1 + 𝑑(𝑎, 𝑧)] + [𝑑(𝑧, 𝑎) + 1 + 𝑒(𝑏, 𝑦)]
= 𝐷(𝑥, 𝑧) + 𝐷(𝑧, 𝑦)
 INTRODUCTION TO METRIC SPACES 31

If 𝑧 ∈ 𝑌 ∖ 𝑋, then
𝐷(𝑥, 𝑦) = 𝑒(𝑥, 𝑦) ≤ 𝑒(𝑥, 𝑧) + 𝑒(𝑧, 𝑦) = 𝐷(𝑥, 𝑧) + 𝐷(𝑧, 𝑦).

Case III 𝑥 ∈ 𝑋 and 𝑦 ∈ 𝑌 ∖ 𝑋

If 𝑧 ∈ 𝑋, then
𝐷(𝑥, 𝑦) = 𝑑(𝑥, 𝑎) + 1 + 𝑒(𝑏, 𝑦)
≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑎) + 1 + 𝑒(𝑏, 𝑦)
≤ 𝑑(𝑥, 𝑧) + {𝑑(𝑧, 𝑎) + 1 + 𝑒(𝑏, 𝑦)}
= 𝐷(𝑥, 𝑧) + 𝐷(𝑧, 𝑦)
If 𝑧 ∈ 𝑌 ∖ 𝑋, then
𝐷(𝑥, 𝑦) = 𝑑(𝑥, 𝑎) + 1 + 𝑒(𝑏, 𝑦)
≤ 𝑑(𝑥, 𝑎) + 1 + 𝑒(𝑏, 𝑧) + 𝑒(𝑧, 𝑦)
≤ {𝑑(𝑥, 𝑎) + 1 + 𝑒(𝑏, 𝑧)} + 𝑒(𝑧, 𝑦)
= 𝐷(𝑥, 𝑧) + 𝐷(𝑧, 𝑦) .
Case IV 𝑥 ∈ 𝑌 ∖ 𝑋 𝑎𝑛𝑑 𝑦 ∈ 𝑋
Similar to case III. ∎

How the metric 𝑫 is obtained? (An Intuitive Idea)

Since 𝑌 = 𝑋 ∪ (𝑌 ∖ 𝑋), therefore

𝑌 × 𝑌 = (𝑋 × 𝑋) ∪ (𝑋 × (𝑌 ∖ 𝑋)) ∪ ((𝑌 ∖ 𝑋) × (𝑌 ∖ 𝑋)) ∪ ((𝑌 ∖ 𝑋) × 𝑋) .

Since we want the metric 𝐷 to be an extension of 𝑑, therefore on 𝑋 × 𝑋 the metric 𝐷

should coincide with 𝑑. Hence we set
𝐷(𝑥, 𝑦) = 𝑑(𝑥, 𝑦) ∀ 𝑥, 𝑦 ∈ 𝑋 .

Since triangle inequality has to be satisfied we cannot define 𝐷 arbitrarily on the

remaining portion. So what’s next ? Since 𝐷 restricted to 𝑌 ∖ 𝑋 must be a metric and
((𝑌 ∖ 𝑋) × (𝑌 ∖ 𝑋)) ∩ (𝑋 × 𝑋) = ∅, we can take any arbitrary metric 𝑒 on 𝑌 ∖ 𝑋 and set
𝐷(𝑥, 𝑦) = 𝑒(𝑥, 𝑦) ∀ 𝑥, 𝑦 ∈ 𝑌 ∖ 𝑋 .

Now we need to define 𝐷 on (𝑋 × (𝑌 ∖ 𝑋)) ∪ ((𝑌 ∖ 𝑋) × 𝑋) using metrics 𝑑 and 𝑒. But to use
metrics 𝑑 and 𝑒 for any point (𝑥, 𝑦) ∈ 𝑋 × (𝑌 ∖ 𝑋), we have to connect 𝑥 with some point
of 𝑋 and 𝑦 with some point of 𝑌 ∖ 𝑋. Since it has to be done for all (𝑥, 𝑦) ∈ 𝑋 × (𝑌 ∖ 𝑋) as
well as for all (𝑥, 𝑦) ∈ (𝑌 ∖ 𝑋) × 𝑋, so we fix two points 𝑎 ∈ 𝑋

and 𝑏 ∈ 𝑌 ∖ 𝑋 which will do our required job.

Now if we define
𝑑(𝑥, 𝑎) + 𝑒(𝑏, 𝑦) ∀ (𝑥, 𝑦) ∈ 𝑋 × (𝑌 ∖ 𝑋)
𝐷(𝑥, 𝑦) = {
𝑒(𝑥, 𝑎) + 𝑑(𝑏, 𝑦) ∀ (𝑥, 𝑦) ∈ (𝑌 ∖ 𝑋) × 𝑋

then 𝐷 is non-negative, symmetric and satisfies triangle inequality, but in this case
𝐷(𝑎, 𝑏) = 0, with 𝑎 ≠ 𝑏. To avoid this failure, we add 1 in both the cases i.e., we define
𝑑(𝑥, 𝑎) + 𝑒(𝑏, 𝑦) + 1 ∀ (𝑥, 𝑦) ∈ 𝑋 × (𝑌 ∖ 𝑋)
𝐷(𝑥, 𝑦) = {
𝑒(𝑥, 𝑎) + 𝑑(𝑏, 𝑦) + 1 ∀ (𝑥, 𝑦) ∈ (𝑌 ∖ 𝑋) × 𝑋
It is important to note that instead of 1, we could have used any positive real number
𝛼. Also, we have an arbitrary choice of metric 𝑒 on 𝑌 ∖ 𝑋. Thus given a metric space
(𝑿, 𝒅) and a proper superset 𝒀 of 𝑿, we can have uncountable number of metrics
on 𝒀 extending the metric 𝒅 on 𝑿.
 INTRODUCTION TO METRIC SPACES 32

6. ISOMETRIES
Definition 6.1 Suppose (𝑋, 𝑑) and (𝑌, 𝑒) are metric spaces and 𝑓 ∶ 𝑋 → 𝑌 be any function.
Then 𝑓 is said to be an isometry or an isometric map if
𝑒(𝑓(𝑎), 𝑓(𝑏)) = 𝑑(𝑎, 𝑏) ∀ 𝑎, 𝑏 ∈ 𝑋 .
In view of the definition, we observe that an
isometry is a distance preserving function.
Furthermore, we observe that an isometry is an
injective function

𝑓(𝑎) = 𝑓(𝑏) ⇔ 𝑒(𝑓(𝑎), 𝑓(𝑏)) = 0

⇔ 𝑑(𝑎, 𝑏) = 0
⇔ 𝑎=𝑏 .

If 𝑓 is an isometry, then we say that the metric subspace (𝑓(𝑋), 𝑒) of (𝑌, 𝑒) is an isomteric
copy of the space (𝑋, 𝑑). Thus metric space (𝑌, 𝑒) is isometric copy of (𝑋, 𝑑) if and only if 𝑓
is a surjective isometry.

Theorem 6.2 Inverse of a surjective isometry is an isometry.

Proof: Let (𝑋, 𝑑) and (𝑌, 𝑒) be two metric spaces and 𝑓: 𝑋 → 𝑌 be a surjective isometry.

Claim : 𝑓 −1 ∶ 𝑌 → 𝑋 is an isometry.
Consider any 𝑎, 𝑏 ∈ 𝑌, then there exists 𝑥, 𝑦 ∈ 𝑋 such that 𝑓(𝑥) = 𝑎 and 𝑓(𝑦) = 𝑏. Now

𝑑(𝑓 −1 (𝑎), 𝑓 −1 (𝑏)) = 𝑑(𝑓 −1 (𝑓(𝑥)), 𝑓 −1 (𝑓(𝑦)) )

= 𝑑(𝑥, 𝑦)
= 𝑒(𝑓(𝑥), 𝑓(𝑦)) [∵ 𝑓 𝑖𝑠 𝑎𝑛 𝑖𝑠𝑜𝑚𝑒𝑡𝑟𝑦]
= 𝑒(𝑎, 𝑏) .

Thus 𝑑(𝑓 −1 (𝑎), 𝑓 −1 (𝑏)) = 𝑒(𝑎, 𝑏) ∀ 𝑎, 𝑏 ∈ 𝑌. Hence it follows that 𝑓 −1 is an isometry. ∎

From above theorem it follows that if (𝑌, 𝑒) is an isometric copy of (𝑋, 𝑑), then (𝑋, 𝑑) is also
an isometric copy of (𝑌, 𝑒). Thus we can simply say that (𝑋, 𝑑) is isometric to (𝑌, 𝑒).

Example 6.3 Consider the real line (ℝ, 𝑢) with 𝑢 as usual metric and the euclidean space
(ℝ3 , 𝑑). Then the inclusion map 𝜇1 ∶ ℝ → ℝ3 given by
𝜇1 (𝑥) = (𝑥, 0, 0) ∀ 𝑥 ∈ ℝ
is an isometry. In fact, for any 𝑥, 𝑦 ∈ ℝ

𝑑(𝜇1 (𝑥), 𝜇1 (𝑦)) = √(𝑥 − 𝑦)2 + (0 − 0)2 + (0 − 0)2

= √(𝑥 − 𝑦)2
= |𝑥 − 𝑦|
= 𝑢(𝑥, 𝑦) .
 INTRODUCTION TO METRIC SPACES 33

Thus 𝜇1 is an isometry and {(𝑥, 0,0) ∶ 𝑥 ∈ ℝ} (popularly known as 𝑥-axis) is isometric copy of
ℝ in ℝ3 . Hence, under isometry, ℝ can be considered as a metric subspace of ℝ3 , however
as a set ℝ is not even a subset of ℝ3 .

Similarly, it can be shown that the other inclusion map 𝜇2 ∶ ℝ → ℝ3 given by

𝜇2 (𝑦) = (0, 𝑦, 0) ∀𝑦∈ℝ
is an isometry. Moreover, {(0, 𝑦, 0) ∶ 𝑥 ∈ ℝ} (𝑦-axis) is isometric copy of ℝ in ℝ3 .
Also, 𝑥𝑦-plane {(𝑥, 𝑦, 0) ∶ 𝑥, 𝑦 ∈ ℝ } in ℝ3 is an isometric copy of ℝ2 under the isometry
𝜇 ∶ ℝ2 → ℝ3 given by
𝜇((𝑥, 𝑦)) = (𝑥, 𝑦, 0) ∀ (𝑥, 𝑦) ∈ ℝ2 .
Obviously, all other lines and planes in ℝ3 are isometric copy of ℝ and ℝ2 , respectively. ∎

Example 6.4 ℝ𝟐 is isometric to ℂ

Let 𝑑 be the usual metric on ℝ2 and 𝑒 be the usual metric on ℂ. We shall show that the
function 𝑓: ℝ2 → ℂ defined as
𝑓((𝑎, 𝑏)) = 𝑎 + 𝑖𝑏 ∀ (𝑎, 𝑏) ∈ ℝ2
is an isometry.

For any (𝑎, 𝑏), (𝑐, 𝑑) ∈ ℝ2 ,

𝑒 (𝑓((𝑎, 𝑏)), 𝑓((𝑐, 𝑑))) = |𝑓((𝑎, 𝑏)) − 𝑓(𝑐, 𝑑)|

= |(𝑎 + 𝑖𝑏) − (𝑐 + 𝑖𝑑)|

= |(𝑎 − 𝑐) + 𝑖(𝑏 − 𝑑)|

= √(𝑎 − 𝑐)2 + (𝑏 − 𝑑)2

= 𝑑((𝑎, 𝑏), (𝑐, 𝑑)) .

Hence it follows that ℝ2 is isometric to ℂ. ∎

Example 6.5 Let (𝑋, 𝑑) be a metric space, 𝑍 be any set and 𝑓 ∶ 𝑍 → 𝑋 be an injective map.
Then the function 𝑒 ∶ 𝑍 × 𝑍 → ℝ2 given by

𝑒(𝑎, 𝑏) = 𝑑(𝑓(𝑎), 𝑓(𝑏)) ∀ 𝑎, 𝑏 ∈ 𝑍

is a metric on 𝑍 induced by 𝑓 and 𝑑. (Example 4.21)

From the definition of 𝑒, we observe that (𝑍, 𝑒) is an isometric copy of (𝑓(𝑍), 𝑑) with the
injective function 𝑓 as an isometry map.

Example 6.6 The injective map 𝑓: ℝ → ℝ defined as

𝑓(𝑥) = 2𝑥 ∀ 𝑥∈𝑋

is not an isometry because |𝑓(1) − 𝑓(2)| = |2 − 4| ≠ |1 − 2|.

Example 6.7 The real line (ℝ, 𝑢) and the discrete metric space (ℝ, 𝑑) are not isometric
copies of each other. In fact for any injective function 𝑓: ℝ (𝑢𝑠𝑢𝑎𝑙) → ℝ (𝑑𝑖𝑠𝑐𝑟𝑒𝑡𝑒)

𝑑(𝑓(1), 𝑓(5)) = 1 ≠ 4 = |1 − 5| = 𝑢(1,5) .

 INTRODUCTION TO METRIC SPACES 34

In view of above example we observe that isometries also depends on metric structure.

Since distances are preserved in isometric copies of any given metric space. Thus
whenever we shall be concerned with only metric properties, we will treat all isometric
spaces as identical spaces.

7. ASSORTED EXAMPLES OF METRIC SPACES

Example 7.1 𝒑-𝒂𝒅𝒊𝒄 Metric on Set of Rationals
Let 𝑝 be a prime number and 𝑥 be any non-zero rational number. Then 𝑥 = 𝑚/𝑛 for some
𝑚, 𝑛 ∈ ℤ. Consider
𝑚 𝑝𝑘 𝑟 𝑟
𝑥 = = 𝑡 = 𝑝𝑘−𝑡
𝑛 𝑝𝑠 𝑠
for some 𝑘, 𝑡, 𝑟, 𝑠 ∈ ℤ , 𝑘, 𝑡 ≥ 0 and 𝑟, 𝑠 are relatively prime to 𝑝 .
Thus each non-zero rational number 𝑥 can be expressed as
𝑝𝑘 𝑟
𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑘, 𝑟, 𝑠 ∈ ℤ 𝑤ℎ𝑒𝑟𝑒 𝑛𝑒𝑖𝑡ℎ𝑒𝑟 𝑟 𝑛𝑜𝑟 𝑠 𝑖𝑠 𝑑𝑖𝑣𝑖𝑠𝑖𝑏𝑙𝑒 𝑏𝑦 𝑝 .
𝑠
Next we show that such 𝑘 is unique. Let 𝑘1 , 𝑘2 , 𝑟1 , 𝑟2 , 𝑠1 , 𝑠2 ∈ ℤ such that 𝑟1 , 𝑟2 , 𝑠1 , 𝑠2 are
relatively prime to 𝑝 and
𝑟1 𝑟2
𝑝𝑘1 = 𝑝 𝑘2
𝑠1 𝑠2
𝑟 𝑠
2 1
⇒ 𝑝𝑘1 −𝑘2 =
𝑟1 𝑠2
⇒ 𝑘1 − 𝑘2 = 0 𝑖. 𝑒., 𝑘1 = 𝑘2
Hence the uniqueness.
Thus we have seen that for any non-zero rational 𝑥, there exists a unique 𝑘 ∈ ℤ such that
𝑟
𝑥 = 𝑝𝑘 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑟, 𝑠 ∈ ℤ 𝑠𝑢𝑐ℎ 𝑡ℎ𝑎𝑡 (𝑟, 𝑝) = 1 𝑎𝑛𝑑 (𝑠, 𝑝) = 1 .
𝑠
We define |𝑥|𝑝 to be 𝑝−𝑘 . Also, we set |0|𝑝 to be 0.

With the help of above expressions, we have a metric 𝑑 ∶ ℚ × ℚ → ℝ on the set of

rationals ℚ , called the 𝒑-𝒂𝒅𝒊𝒄 𝒎𝒆𝒕𝒓𝒊𝒄 on ℚ , given by
𝑑(𝑎, 𝑏) = |𝑎 − 𝑏|𝑝 .
We shall prove 𝑑 is indeed a metric on ℚ.
By definition, 𝑑 is a non-negative function. Now for any , 𝑏 ∈ ℚ , let
𝑟
𝑎 − 𝑏 = 𝑝𝑘 .
𝑠
Then
(– 𝑟)
𝑏 − 𝑎 = 𝑝𝑘 .
𝑠
Thus
|𝑎 − 𝑏|𝑝 = |𝑏 − 𝑎|𝑝 = 𝑝−𝑘
𝑖. 𝑒. , 𝑑(𝑎, 𝑏) = 𝑑(𝑏, 𝑎) = 𝑝−𝑘
Hence the function 𝑑 is symmetric.
 INTRODUCTION TO METRIC SPACES 35

Also, by definition,
𝑑(𝑎, 𝑏) = 0 ⟺ |𝑎 − 𝑏|𝑝 = 0 ⟺ 𝑎 − 𝑏 = 0 ⟺ 𝑎 = 𝑏 .
To see triangle inequality, consider any 𝑎, 𝑏, 𝑐 ∈ ℚ and let
𝑟1 𝑟2 𝑟3
𝑎 − 𝑏 = 𝑝𝑘 , 𝑎 − 𝑐 = 𝑝𝑚 𝑎𝑛𝑑 𝑐 − 𝑏 = 𝑝𝑛 .
𝑠1 𝑠2 𝑠3
Further let 𝑡 = min {𝑚, 𝑛}. Now
𝑟1
𝑝𝑘 = 𝑎 − 𝑏 = 𝑎 − 𝑐 + 𝑐 − 𝑏
𝑠1
𝑟2 𝑟3
= 𝑝𝑚 + 𝑝𝑛
𝑠2 𝑠3
𝑟′
= 𝑝𝑡 𝑤ℎ𝑒𝑟𝑒 𝑟 ′ , 𝑠 ′ ∈ ℤ, 𝑠 ′ = 𝑠2 𝑠3
𝑠′
𝑝𝑎 𝑟 ′′
= 𝑝𝑡 ′ 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑎, 𝑟" ∈ ℤ , 𝑎 ≥ 0 (𝑟" , 𝑝) = 1
𝑠
𝑟 ′′
= 𝑝𝑡+𝑎 ′ 𝑤ℎ𝑒𝑟𝑒 (𝑟 ′′ , 𝑝) = 1 𝑎𝑛𝑑 (𝑠 ′ , 𝑝) = 1
𝑠
Thus by uniqueness of 𝑘, we have 𝑘 = 𝑡 + 𝑎 𝑖. 𝑒. , 𝑡 ≤ 𝑘 . Consequently, we have
𝑝−𝑘 ≤ 𝑝−𝑡 = max {𝑝 −𝑚 , 𝑝−𝑛 } ≤ 𝑝−𝑚 + 𝑝−𝑛
⇒ |𝑎 − 𝑏|𝑝 ≤ |𝑎 − 𝑐|𝑝 + |𝑐 − 𝑏|𝑝 . ∎

Example 7.2 Let 𝑆 be any non-empty set and ℱ(𝑆) be the set of all finite subsets of 𝑆. For
all 𝐴, 𝐵 ∈ ℱ(𝑆), let Δ(𝐴, 𝐵) = (𝐴 ∖ 𝐵) ∪ (𝐵 ∖ 𝐴) be the symmetric difference between 𝐴 and 𝐵.
Let 𝑑: ℱ(𝑆) × ℱ(𝑆) → ℝ be the function given by

𝑑(𝐴, 𝐵) = |Δ(𝐴, 𝐵)| = |𝐴 ∖ 𝐵| + |𝐵 ∖ 𝐴| ∀ 𝐴, 𝐵 ∈ ℱ(𝑆)

where |𝐴| denotes the cardinality of set 𝐴.

We shall prove that 𝑑 is a metric on ℱ(𝑆).

Clearly, 𝑑(𝐴, 𝐵) ≥ 0 ∀ 𝐴, 𝐵 ∈ ℱ(𝑆) and hence P1 is satisfied.
For P2, consider any 𝐴, 𝐵 ∈ ℱ(𝑆),
𝑑(𝐴, 𝐵) = 0 ⇔ |Δ(𝐴, 𝐵)| = 0
⇔ |𝐴 ∖ 𝐵| + |𝐵 ∖ 𝐴| = 0
⇔ |𝐴 ∖ 𝐵| = 0 𝑎𝑛𝑑 |𝐵 ∖ 𝐴| = 0
⇔ 𝐴 ∖ 𝐵 = ∅ 𝑎𝑛𝑑 𝐵 ∖ 𝐴 = ∅
⇔ 𝐴 ⊆ 𝐵 𝑎𝑛𝑑 𝐵 ⊆ 𝐴
⇔𝐴=𝐵 .

Thus P2 holds.
For P3 consider any 𝐴, 𝐵 ∈ ℱ(𝑆),
𝑑(𝐴, 𝐵) = |Δ(𝐴, 𝐵)| = |𝐴 ∖ 𝐵| + |𝐵 ∖ 𝐴|
= |𝐵 ∖ 𝐴| + |𝐴 ∖ 𝐵|
= |Δ(𝐵, 𝐴)|
= 𝑑(𝐵, 𝐴) .
 INTRODUCTION TO METRIC SPACES 36

Now let 𝐴, 𝐵, 𝐶 ∈ ℱ(𝑆) be any sets. Consider

𝐴 ∖ 𝐵 = 𝐴 ∩ 𝐵𝑐
= (𝐴 ∩ (𝐶 ∪ 𝐶 𝑐 )) ∩ 𝐵𝑐
= ((𝐴 ∩ 𝐶) ∪ (𝐴 ∩ 𝐶 𝑐 )) ∩ 𝐵𝑐
= (𝐴 ∩ 𝐶 ∩ 𝐵𝑐 ) ∪ (𝐴 ∩ 𝐶 𝑐 ∩ 𝐵𝑐 )
⊆ (𝐶 ∩ 𝐵𝑐 ) ∪ (𝐴 ∩ 𝐶 𝑐 )[∵ 𝐴 ∩ 𝐶 ⊆ 𝐶 𝑎𝑛𝑑 𝐴 ∩ 𝐵𝑐 ⊆ 𝐴]
= (𝐶 ∖ 𝐵) ∪ (𝐴 ∖ 𝐶)

Thus it follows that

|𝐴 ∖ 𝐵| ≤ |𝐴 ∖ 𝐶 | + |𝐶 ∖ 𝐵| .

Similarly,
|𝐵 ∖ 𝐴| ≤ |𝐵 ∖ 𝐶| + |𝐶 ∖ 𝐴| .

Hence
𝑑(𝐴, 𝐵) = |Δ(𝐴, 𝐵)|
= |𝐴 ∖ 𝐵| + |𝐵 ∖ 𝐴|
≤ |𝐴 ∖ 𝐶| + |𝐶 ∖ 𝐵| + |𝐵 ∖ 𝐶| + |𝐶 ∖ 𝐴|
= |𝐴 ∖ 𝐶| + |𝐶 ∖ 𝐴| + |𝐵 ∖ 𝐶| + |𝐶 ∖ 𝐵|
= |Δ(𝐴, 𝐶)| + |Δ(𝐶, 𝐵)|
= 𝑑(𝐴, 𝐶) + 𝑑(𝐶, 𝐵) .

Thus P4 holds.
Since all the four axioms of a metric are satisfied by 𝑑, therefore 𝑑 is a metric on ℱ(𝑆).∎

To show 𝑨 ∖ 𝑩 ⊆ (𝑪 ∖ 𝑩) ∪ (𝑨 ∖ 𝑪), we can also take the element

containment approach. In fact,
𝒙 ∈ 𝑨\𝑩
⇒ 𝒙 ∈ 𝑨 𝒂𝒏𝒅 𝒙 ∉ 𝑩
⇒ 𝒙 ∈ 𝑨 ∩ 𝑪 𝒂𝒏𝒅 𝒙 ∉ 𝑩 𝒐𝒓 𝒙 ∈ 𝑨 ∖ 𝑪 𝒂𝒏𝒅 𝒙 ∉ 𝑩
⇒ 𝒙 ∈ 𝑪 𝒂𝒏𝒅 𝒙 ∉ 𝑩 𝒐𝒓 𝒙∈𝑨∖𝑪
⇒ 𝒙∈𝑪∖𝑩 𝒐𝒓 𝒙∈𝑨∖𝑪
⇒ 𝒙 ∈ (𝑪 ∖ 𝑩) ∪ (𝑨 ∖ 𝑪)

Example 7.3 Consider another metric 𝑑𝑝 on ℝ𝑛 defined as

1
𝑛 𝑝

𝑑𝑝 (𝑥, 𝑦) = (∑|𝑥𝑖 − 𝑦𝑖 |𝑝 ) ∀ 𝑥 = (𝑥1 , … , 𝑥𝑛 ), 𝑦 = (𝑦1 , … , 𝑦𝑛 ) ∈ ℝ𝑛 .

𝑖=1
To see that 𝑑𝑝 is indeed a metric on ℝ𝑛 , we shall just verify triangle inequality (other
axioms can be easily proved).

Consider any 𝑥 = (𝑥1 , . . . , 𝑥𝑛 ), 𝑦 = (𝑦1 , . . . , 𝑦𝑛 ), 𝑧 = (𝑧1 , . . . , 𝑧𝑛 ) ∈ ℝ𝑛 . Now

 INTRODUCTION TO METRIC SPACES 37

1
𝑛 𝑝

𝑑𝑝 (𝑥, 𝑦) = (∑|𝑥𝑖 − 𝑦𝑖 |𝑝 )
𝑖=1
1
𝑛 𝑝

= (∑|(𝑥𝑖 − 𝑧𝑖 ) + (𝑧𝑖 − 𝑦𝑖 )|𝑝 )

𝑖=1
1 1
𝑛 𝑝 𝑛 𝑝

≤ (∑|𝑥𝑖 − 𝑧𝑖 |𝑝 ) + (∑|𝑧𝑖 − 𝑦𝑖 |𝑝 ) [𝑈𝑠𝑖𝑛𝑔 𝐌𝐢𝐧𝐤𝐨𝐰𝐬𝐤𝐢′ 𝐬 𝐈𝐧𝐞𝐪𝐮𝐚𝐥𝐢𝐭𝐲].

𝑖=1 𝑖=1
≤ 𝑑𝑝 (𝑥, 𝑧) + 𝑑𝑝 (𝑧, 𝑦).

Thus triangle inequality holds. Hence 𝑑𝑝 is a metric on ℝ𝑛 and (ℝ𝑛 , 𝑑𝑝 ) is a metric space.
For 𝑝 = 2, the metric space (ℝ𝑛 , 𝑑𝑝 ) is the Euclidean space of dimension 2. █

Example 7.4 The space 𝒍𝟏 (ℝ)

Let 𝑋 = 𝒍𝟏 (ℝ) be the set of all real sequences 𝑥 = 〈𝑥𝑛 〉 such that ∑∞
𝑛=1|𝑥𝑛 | < ∞ i.e.,
∞

𝑋 = {〈𝑥𝑛 〉 ∶ 𝑥𝑛 ∈ ℝ 𝑎𝑛𝑑 ∑|𝑥𝑛 | < ∞}.

𝑛=1
Consider the function 𝑑 ∶ 𝑋 × 𝑋 → ℝ defined as
∞

𝑑(𝑥, 𝑦) = ∑|𝑥𝑛 − 𝑦𝑛 |
𝑛=1
where 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉 ∈ 𝑋.

For 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉 ∈ 𝑋, we have

∞ ∞

∑|𝑥𝑛 | < ∞ 𝑎𝑛𝑑 ∑|𝑦𝑛 | < ∞ .

𝑛=1 𝑛=1
Since the sum of two convergent series is convergent, therefore
∞

∑(|𝑥𝑛 | + |𝑦𝑛 |) < ∞.

𝑛=1
Now for all 𝑛 ∈ ℕ,
|𝑥𝑛 − 𝑦𝑛 | ≤ |𝑥𝑛 | + |𝑦𝑛 |
Thus by Comparison test for series of positive terms, it follows that
∞

∑|𝑥𝑛 − 𝑦𝑛 | < ∞ .
𝑛=1
Hence 𝑑 is a well defined function on 𝑋 × 𝑋.

By definition, 𝑑 is a non-negative real valued function.

For any 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉 ∈ 𝑋,
𝑑(𝑥, 𝑦) = 0
∞

⇔ ∑|𝑥𝑛 − 𝑦𝑛 | = 0
𝑛=1
⇔ 𝑥𝑛 − 𝑦𝑛 = 0 , ∀ 𝑛∈ℕ
⇔ 𝑥𝑛 = 𝑦𝑛 , ∀ 𝑛∈ℕ
⇔ 𝑥 = 𝑦.
For any 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉 ∈ 𝑋,
∞

𝑑(𝑥, 𝑦) = ∑|𝑥𝑛 − 𝑦𝑛 |
𝑛=1
 INTRODUCTION TO METRIC SPACES 38

= ∑|𝑦𝑛 − 𝑥𝑛 |
𝑛=1

= 𝑑(𝑦, 𝑥).

Let 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉, 𝑧 = 〈𝑧𝑛 〉 ∈ 𝑋 be any three points. Then

∞

𝑑(𝑥, 𝑦) = ∑|𝑥𝑛 − 𝑦𝑛 |
𝑛=1
∞

= ∑|(𝑥𝑛 − 𝑧𝑛 ) + (𝑧𝑛 − 𝑦𝑛 )|
𝑛=1
∞ ∞

≤ ∑|𝑥𝑛 − 𝑧𝑛 | + ∑|𝑧𝑛 − 𝑦𝑛 |
𝑛=1 𝑛=1

= 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦).

It follows that
𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) ∀ 𝑥, 𝑦, 𝑧 ∈ 𝑋 .
Therefore all the axioms are satisfied. Hence 𝑑 is a metric on 𝑋. █

Example 7.5 The space 𝒍𝟐 (ℝ) of all square summable sequences

Let 𝑋 = 𝒍𝟐 (ℝ) be the set of all square summable real sequences 𝑥 = 〈𝑥𝑛 〉 i.e.,
1
∞ 2

𝑋 = {〈𝑥𝑛 〉 ∶ 𝑥𝑛 ∈ ℝ 𝑎𝑛𝑑 (∑|𝑥𝑛 |2 ) < ∞}.

𝑛=1

Consider the function 𝑑 ∶ 𝑋 × 𝑋 → ℝ defined as

1
∞ 2

𝑑(𝑥, 𝑦) = (∑|𝑥𝑛 − 𝑦𝑛 |2 )
𝑛=1
where 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉 ∈ 𝑋. Then 𝑑 is a metric on 𝑋.

For 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉 ∈ 𝑋, we have

1 1
∞ 2 ∞ 2

(∑|𝑥𝑛 |2 ) < ∞ 𝑎𝑛𝑑 (∑|𝑦𝑛 |2 ) < ∞ .

𝑛=1 𝑛=1
Using Minkowski’s Inequality for infinite sums, we have
1 1 1
∞ 2 ∞ 2 ∞ 2

(∑|𝑥𝑛 + (−𝑦𝑛 )|2 ) ≤ (∑|𝑥𝑛 |2 ) + (∑|𝑦𝑛 |2 ) < ∞

𝑛=1 𝑛=1 𝑛=1
Therefore 𝑑(𝑥, 𝑦) is a finite real number for all 𝑥, 𝑦 ∈ 𝑋.
By definition, 𝑑 is a non-negative.
For any 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉 ∈ 𝑋,
𝑑(𝑥, 𝑦) = 0
1
∞ 2

⇔ (∑|𝑥𝑛 − 𝑦𝑛 |2 ) = 0
𝑛=1
⇔ 𝑥𝑛 − 𝑦𝑛 = 0 , ∀ 𝑛∈ℕ
⇔ 𝑥𝑛 = 𝑦𝑛 , ∀ 𝑛∈ℕ
⇔ 𝑥 = 𝑦.
 INTRODUCTION TO METRIC SPACES 39

For any 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉 ∈ 𝑋,

1
∞ 2

𝑑(𝑥, 𝑦) = (∑|𝑥𝑛 − 𝑦𝑛 |2 )
𝑛=1
1
∞ 2

= (∑|𝑦𝑛 − 𝑥𝑛 |2 ) = 𝑑(𝑦, 𝑥).

𝑛=1

Let 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉, 𝑧 = 〈𝑧𝑛 〉 ∈ 𝑋 be any three points. Then

1 1
∞ 2 ∞ 2

𝑑(𝑥, 𝑦) = (∑|𝑥𝑛 − 𝑦𝑛 |2 ) = (∑|(𝑥𝑛 − 𝑧𝑛 ) + (𝑧𝑛 − 𝑦𝑛 )|2 )

𝑛=1 𝑛=1

Let 𝑎𝑛 = 𝑥𝑛 − 𝑧𝑛 , 𝑏𝑛 = 𝑧𝑛 − 𝑦𝑛 for all 𝑛 ∈ ℕ. This implies

1
∞ 2

𝑑(𝑥, 𝑦) = (∑|𝑎𝑛 + 𝑏𝑛 |2 )
𝑛=1

Now from Minkowski’s inequality for infinite sums we have,

1 1
∞ 2 ∞ 2

𝑑(𝑥, 𝑦) ≤ (∑|𝑎𝑛 |2 ) + (∑|𝑏𝑛 |2 )

𝑛=1 𝑛=1
1 1
∞ 2 ∞ 2

= (∑|𝑥𝑛 − 𝑧𝑛 |2 ) + (∑|𝑧𝑛 − 𝑦𝑛 |2 )
𝑛=1 𝑛=1

= 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦).

It follows that
𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) ∀ 𝑥, 𝑦, 𝑧 ∈ 𝑋 .
Therefore all the axioms are satisfied. Hence 𝑑 is a metric on 𝑋.

The space 𝑙2 (ℝ) is also called the Hilbert space. █

Example 7.6 The space 𝒍𝒑 (ℝ)

Let 𝑋 be the set of all real sequences 𝑥 = 〈𝑥𝑛 〉 such that
1
∞ 𝑝

(∑|𝑥𝑛 |𝑝 ) < ∞.
𝑛=1
Consider the function 𝑑 ∶ 𝑋 × 𝑋 → ℝ defined as
1
∞ 𝑝

𝑑(𝑥, 𝑦) = (∑|𝑥𝑛 − 𝑦𝑛 |𝑝 )
𝑛=1
where 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉 ∈ 𝑋. Then 𝑑 is a metric on 𝑋.

Verification of the fact that 𝑑 is indeed a metric is similar to Example 7.5. █

Example 7.7 The space 𝒍∞ (ℝ) of all bounded sequences

Let 𝑋 = 𝑙∞ (ℝ) be the set of all bounded sequence of real numbers i.e.,
𝑋 = {〈𝑥𝑛 〉 ∶ sup {|𝑥𝑛 | ∶ 𝑛 ∈ ℕ} < ∞ }.
 INTRODUCTION TO METRIC SPACES 40

Let 𝑑 ∶ 𝑋 × 𝑋 → ℝ be a function defined as

𝑑(𝑥, 𝑦) = sup {|𝑥𝑛 − 𝑦𝑛 | ∶ 𝑛 ∈ ℕ} ∀ 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉 ∈ 𝑋.
Then 𝑑 is metric on 𝑋.

To verify this, we first need to show that 𝑑 is a well defined function. For that it is enough
to show that if 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉 ∈ 𝑋, then 𝑑(𝑥, 𝑦) ∈ ℝ.

Since 𝑥, 𝑦 ∈ 𝑋, there exist real numbers 𝑀 and 𝑀′ such that

|𝑥𝑛 | ≤ 𝑀 𝑎𝑛𝑑 |𝑦𝑛 | ≤ 𝑀′ ∀ 𝑛 ∈ ℕ .
Now for all 𝑛 ∈ ℕ,
|𝑥𝑛 − 𝑦𝑛 | ≤ |𝑥𝑛 | + |𝑦𝑛 | ≤ 𝑀 + 𝑀′
⇒ sup {|𝑥𝑛 − 𝑦𝑛 | ∶ 𝑛 ∈ ℕ} ≤ 𝑀 + 𝑀′
Thus it follows that 0 ≤ 𝑑(𝑥, 𝑦) ≤ 𝑀 + 𝑀′ and therefore 𝑑(𝑥, 𝑦) ∈ ℝ. Thus 𝑑 is a well defined
function.
For function 𝑑, properties P1, P2 and P3 can be easily verified.

For triangle inequality, consider 𝑥 = 〈𝑥𝑛 〉, 𝑦 = 〈𝑦𝑛 〉, 𝑧 = 〈𝑧𝑛 〉 ∈ 𝑋. Then

|𝑥𝑛 − 𝑦𝑛 | ≤ |𝑥𝑛 − 𝑧𝑛 | + |𝑧𝑛 − 𝑦𝑛 |
≤ sup {|𝑥𝑛 − 𝑧𝑛 | ∶ 𝑛 ∈ ℕ} + sup {|𝑧𝑛 − 𝑦𝑛 | ∶ 𝑛 ∈ ℕ}
≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦).
Taking supremum over 𝑛 ∈ ℕ, we get
sup {|𝑥𝑛 − 𝑦𝑛 | ∶ 𝑛 ∈ ℕ} ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦)
⇒ 𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) .
Thus 𝑑 is a metric on 𝑋. █

Example 7.8 (The extended complex plane and its spherical representation)
Let 𝑋 = ℂ ∪ {∞} = ℂ∞ be the extended complex plane. We wish to introduce some metric
on 𝑋 using which we can discuss the continuity of functions which assumes the value
infinity. To achieve this we represent 𝑋 as the unit sphere in ℝ3 ,
𝑋 ≡ 𝑆 = {(𝑥1 , 𝑥2 , 𝑥3 ) ∈ ℝ3 ∶ 𝑥12 + 𝑥22 + 𝑥32 = 1}.
Clearly, the complex plane ℂ can be identified with the set {(𝑥1 , 𝑥2 , 0) ∶ 𝑥1 , 𝑥2 ∈ ℝ}. With this
identification, the complex plane ℂ can be seen as the plane passing through the equator
of the unit sphere 𝑆.

Now for any point 𝑧 ∈ ℂ, clearly, the line joining the point 𝑧 and the North pole 𝑁 = (0,0,1)
on 𝑆 intersects the sphere 𝑆 at exactly one point 𝑍 ≠ 𝑁.If |𝑧| > 1, then 𝑍 lies in northern
hemispheere, and if |𝑧| < 1 then 𝑍 lies in the southern hemisphere. Moreover, if |𝑧| = 1 then
 INTRODUCTION TO METRIC SPACES 41

𝑍 = 𝑧. These fact establishes that there exists one-to-one correspondence (𝑧 → 𝑍) between

ℂ and the set 𝑆 ∖ {𝑁}. Moreover, as |𝑧| → ∞, 𝑍 approaches 𝑁. Therefore to accomplish a one-
to-one correspondence between the extended plane ℂ∞ and the sphere 𝑆, we identify the
north pole 𝑁 with ∞ of the extended plane. Thus ℂ∞ is represented as the sphere 𝑆. This
correspondence between the points of 𝑆 and ℂ∞ is called the stereographic projection.

Now let us explore this representation. Let 𝒛 = 𝒙 + 𝒊𝒚 be any point in ℂ and 𝒁 = (𝒙𝟏 , 𝒙𝟐 , 𝒙𝟑 )
be the corresponding point in ℝ3 . We wish to express 𝑥𝑖 ′𝑠 in terms of 𝑥 and 𝑦. The line in
ℝ3 through 𝑧 = (𝑥, 𝑦, 0) and 𝑁 = (0,0,1) is given by
𝐿𝑧,𝑁 = {(1 − 𝑡)𝑧 + (𝑡)𝑁 ∶ −∞ < 𝑡 < ∞}
= {((1 − 𝑡)𝑥, (1 − 𝑡)𝑦, 𝑡) ∶ −∞ < 𝑡 < ∞} . − − − − − −(𝑨)
Now since 𝑍 ∈ 𝐿𝑧,𝑁 , therefore there exists some 𝑠 ∈ (−∞, ∞) such that

𝑍 = ((1 − 𝑠)𝑥, (1 − 𝑠)𝑦, 𝑠) . − − − − − −(𝑩)

Also, since 𝑍 ∈ 𝑆, therefore we have
1 = (1 − 𝑠)2 𝑥 2 + (1 − 𝑠)2 𝑦 2 + 𝑠 2 = (1 − 𝑠)2 |𝑧|2 + 𝑠 2
⇒ 1 − 𝑠 2 = (1 − 𝑠)2 |𝑧|2
1 − 𝑠2
⇒ = |𝑧|2 [∵ 𝑍 ≠ 𝑁 ⇒ 𝑠 ≠ 1]
(1 − 𝑠)2
|𝑧|2 − 1
⇒ 𝑠= [𝑈𝑠𝑖𝑛𝑔 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑑𝑜 & 𝑑𝑖𝑣𝑖𝑑𝑒𝑛𝑑𝑜 ]
|𝑧|2 + 1
Thus using value of s in equation (𝑩), we get
2𝑥 2𝑦 |𝑧|2 − 1
𝑥1 = 2 , 𝑥2 = 2 , 𝑥3 =
|𝑧| + 1 |𝑧| + 1 |𝑧|2 + 1
𝑧 + 𝑧̅ −𝑖(𝑧 − 𝑧̅) |𝑧|2 − 1
𝑖. 𝑒., 𝑥1 = , 𝑥2 = , 𝑥3 = − − − − − (𝑪)
|𝑧|2 + 1 |𝑧|2 + 1 |𝑧|2 + 1

𝑪𝒐𝒏𝒗𝒆𝒓𝒔𝒆𝒍𝒚, if we are given a point 𝒁 = (𝒙𝟏 , 𝒙𝟐 , 𝒙𝟑 ) ≠ 𝑵, then setting 𝑡 = 𝑥3 and using it in

the equation (𝑨), we arrive at
𝑥1 + 𝑖𝑥2
𝑧= .
1 − 𝑥3
Thus we have shown that we can represent the extended complex plane ℂ∞ by the unit
sphere S. Hence using the Euclidean metric defined on ℝ3 we can define a metric on the
extended complex plane as follows:
𝑑∞ ∶ ℂ∞ × ℂ∞ → ℝ+ 𝑎𝑠
1
3 2

𝑑∞ (𝑧, 𝑧 ′ ) = 𝑑(𝑍, 𝑍 ′ ) = [∑(𝑥𝑖 − 𝑥𝑖′ )2 ] ∀ 𝑧, 𝑧 ′ ∈ ℂ∞ − − − − − (𝑫)

𝑖=1

where 𝑑 is the usual metric on ℝ3 and 𝑍 = (𝑥1 , 𝑥2 , 𝑥3 ), 𝑍 ′ = (𝑥1′ , 𝑥2′ , 𝑥3′ ) are the points in unit
sphere 𝑆 corresponding to 𝑧, 𝑧′ respectively. Now we are in a position to define our distance
function on the extended plane ℂ∞ .
Now since 𝑍, 𝑍′ lies on 𝑆, therefore from equation (𝑫) it follows that
1
𝑑∞ (𝑧, 𝑧 ′ ) = [2 − 2(𝑥1 𝑥1′ + 𝑥2 𝑥2′ + 𝑥3 𝑥3′ )]2 .
 INTRODUCTION TO METRIC SPACES 42

Then using equation (𝑪), we get

2|𝑧 − 𝑧 ′ |
𝑑∞ (𝑧, 𝑧 ′ ) = 1 ∀ 𝑧, 𝑧 ′ ∈ ℂ 𝑎𝑛𝑑
[(1 + |𝑧|2 ) + (1 + |𝑧 ′ |2 )]2
2
𝑑∞ (𝑧, ∞) = 1 ∀ 𝑧 ∈ ℂ. █
(1 + |𝑧|2 )2

8. NORMS ON VECTOR SPACES (OR LINEAR SPACES)

So far while defining metrics on any given set, we never bothered about the algebraic
properties inherited by the set considered. But most of the examples we considered till
now have some inherited algebraic properties. In fact, the most
familiar examples we considered like ℝ, ℝ2 , ℂ and space 𝒞[0,1] of
continuous function on [0,1] are all vector spaces (linear
spaces) over ℝ or ℂ.

Consider the space (ℝ2 , 𝑑), where 𝑑 is the Euclidean metric. Let
𝑢 ∈ ℝ2 . Suppose we want to measure the length of the vector
𝑢 = (𝑢1 , 𝑢2 ). From high school geometry, using Pythagoras
Theorem, length of the vector 𝑢 is √𝑢12 + 𝑢22 = 𝑑(𝑢, 0).

Also for any scalar 𝑐, length of the vector 𝑐𝑢 is |𝑐|√𝑢12 + 𝑢22 i.e., 𝑑(𝑐𝑢, 0) = |𝑐|𝑑(𝑢, 0).

Further for any vector 𝑣, length of the vector 𝑢 − 𝑣 is √(𝑢1 − 𝑣1 )2 + (𝑢2 − 𝑣2 )2 i.e.,

𝑑(𝑢 − 𝑣, 0) = 𝑑(𝑢, 𝑣).

Since we always generalize the ideas from familiar examples, thus to define a metric on
any arbitrary vector space, we must keep in mind that a metric 𝑑 defined on a vector space
𝑉 should satisfy the following conditions:

1. 𝑑(𝑢 − 𝑣, 0) = 𝑑(𝑢, 𝑣) ∀ 𝑢, 𝑣 ∈ 𝑉
2. 𝑑(𝛼𝑢, 0) = |𝛼|𝑑(𝑢, 0) for any scalar 𝛼 and for all 𝑢 ∈ 𝑉.

Now to define such a metric, we first need to find the answer to the following question.
Why the usual metrics on ℝ, ℝ2 , ℂ and 𝒞[0,1] satisfies the above two properties? We see
that to define usual metrics on ℝ, ℝ2 , ℂ and 𝒞[0,1] we use absolute value of a number in ℝ,
the magnitude of a vector in ℝ2 , the modulus of a complex number in ℂ and the maximum
value of |𝑓| for any 𝑓 in 𝒞[0,1].

These ideas leads to the introduction of an interesting concept called norms defined on a
linear space using its algebraic properties. It relates the algebraic and geometric (metric)
properties possessed by a linear space.

Norm on a Linear Space

A norm on a linear space 𝑉 over ℝ or ℂ is a real function ∥⋅∥ ∶ 𝑉 → ℝ which satisfy the
following properties:
1. ∥ 𝑥 ∥≥ 0 ∀ 𝑥 ∈ 𝑉.

2. For any 𝑥 ∈ 𝑉, ∥ 𝑥 ∥= 0 if and only if 𝑥 = 0.

3. ∥ 𝛼𝑥 ∥= |𝛼| ∥ 𝑥 ∥ ∀ 𝑥 ∈ 𝑉 and for any scalar 𝛼.

 INTRODUCTION TO METRIC SPACES 43

4. ∥ 𝑥 + 𝑦 ∥≤∥ 𝑥 ∥ +∥ 𝑦 ∥ for any 𝑥, 𝑦 ∈ 𝑉.

A normed linear space, denoted by (𝑽, ∥⋅∥) is a linear space with a norm defined on it.
For each 𝑥 ∈ 𝑉, the number ∥ 𝑥 ∥ is called the length of the vector 𝑥 w.r.t norm ∥⋅∥. If 𝑊 is
a linear subspace of 𝑉 and ∥⋅∥𝑊 is restriction of ∥⋅∥ to 𝑊, then (𝑾, ∥⋅∥𝑾 ) is called a normed
linear subspace of (𝑽, ∥⋅∥). It is easy to see that (𝑊, ∥⋅∥𝑊 ) in itself is a normed linear
space.

Theorem 8.1.1 If ∥⋅∥ is a norm on a linear space 𝑉, then 𝑑: 𝑉 × 𝑉 → ℝ given by

𝑑(𝑥, 𝑦) =∥ 𝑥 − 𝑦 ∥ ∀ 𝑥, 𝑦 ∈ 𝑉

is a metric on 𝑉.

Proof: Since ∥⋅∥ is a non-negative function, it follows that 𝑑(𝑥, 𝑦) ≥ 0 ∀ 𝑥, 𝑦 ∈ 𝑉.

Further for any 𝑥, 𝑦 ∈ 𝑉,
𝑑(𝑥, 𝑦) = 0 ⇔ ∥ 𝑥 − 𝑦 ∥= 0 ⇔ 𝑥 − 𝑦 = 0 ⇔ 𝑥 = 𝑦 .
Also,
𝑑(𝑥, 𝑦) =∥ 𝑥 − 𝑦 ∥
=∥ −1(𝑦 − 𝑥) ∥
= |−1| ∥ 𝑦 − 𝑥 ∥ [∥ 𝛼𝑥 ∥= |𝛼| ∥ 𝑥 ∥]
=∥ 𝑦 − 𝑥 ∥
= 𝑑(𝑦, 𝑥) .

Finally, to see triangle inequality, consider for any 𝑥, 𝑦, 𝑧 ∈ 𝑉

𝑑(𝑥, 𝑦) =∥ 𝑥 − 𝑦 ∥
=∥ (𝑥 − 𝑧) + (𝑧 − 𝑦) ∥
≤∥ 𝑥 − 𝑧 ∥ +∥ 𝑧 − 𝑦 ∥
= 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) .

Thus 𝑑 is a metric on 𝑉. ∎

Definition 8.1.2 Given any normed linear space (𝑉, ∥⋅∥), the metric 𝑑: 𝑉 × 𝑉 → ℝ given by

𝑑(𝑥, 𝑦) =∥ 𝑥 − 𝑦 ∥ ∀ 𝑥, 𝑦 ∈ 𝑉

is called the metric determined by the norm or a metric induced by the norm.

Theorem 8.1.3 Let 𝑉 be a linear space and 𝑑: 𝑉 × 𝑉 → ℝ be any metric. Then 𝑑 is

determined by a norm if and only if following conditions are satisfied:

1. 𝑑(𝑥 − 𝑦, 0) = 𝑑(𝑥, 𝑦) ∀ 𝑥, 𝑦 ∈ 𝑉

2. 𝑑(𝛼𝑥, 0) = |𝛼|𝑑(𝑥, 0) ∀ 𝑥 ∈ 𝑉 and for any scalar 𝛼.

Proof: Suppose 𝑑 is induced by the norm ∥⋅∥ i.e.,

𝑑(𝑥, 𝑦) =∥ 𝑥 − 𝑦 ∥ ∀ 𝑥, 𝑦 ∈ 𝑉 .

For any 𝑥, 𝑦 ∈ 𝑉,
𝑑(𝑥 − 𝑦, 0) = ∥ (𝑥 − 𝑦) − 0 ∥
=∥𝑥−𝑦∥
= 𝑑(𝑥, 𝑦) .

Also, for any 𝑥 ∈ 𝑉

𝑑(𝛼𝑥, 0) = ∥ 𝛼𝑥 − 0 ∥
= ∥ 𝛼(𝑥 − 0) ∥
 INTRODUCTION TO METRIC SPACES 44

= |𝛼| ∥ 𝑥 − 0 ∥
= |𝛼|𝑑(𝑥, 0) .

𝑪𝒐𝒏𝒗𝒆𝒓𝒔𝒆𝒍𝒚, suppose (1) and (2) hold. Define ∥⋅∥ ∶ 𝑉 → ℝ as

∥ 𝒙 ∥= 𝒅(𝒙, 𝟎) ∀ 𝒙 ∈ 𝑽 .

Since 𝑑(𝑥, 0) ≥ 0 ∀ 𝑥 ∈ 𝑉, it follows that ∥ 𝑥 ∥ ≥ 0 ∀ 𝑥 ∈ 𝑉.

Also, for any 𝑥 ∈ 𝑉

∥ 𝑥 ∥ = 0 ⇔ 𝑑(𝑥, 0) = 0 ⇔ 𝑥 = 0 .

Now for scalar 𝛼 and for any 𝑥 ∈ 𝑉,

∥ 𝛼𝑥 ∥ = 𝑑(𝛼𝑥, 0)
= |𝛼|𝑑(𝑥, 0)
= |𝛼| ∥ 𝑥 ∥ .

For triangle inequality, consider any 𝑥, 𝑦 ∈ 𝑉

∥ 𝑥 + 𝑦 ∥ = 𝑑(𝑥 + 𝑦, 0)
≤ 𝑑(𝑥 + 𝑦, 𝑦) + 𝑑(𝑦, 0) [𝑈𝑠𝑖𝑛𝑔 𝑇𝑟𝑖𝑎𝑛𝑔𝑙𝑒 𝐼𝑛𝑒𝑞𝑢𝑎𝑙𝑖𝑡𝑦 𝑓𝑜𝑟 𝑑]
= 𝑑(𝑥 + 𝑦 − 𝑦, 0) + 𝑑(𝑦, 0)
= 𝑑(𝑥, 0) + 𝑑(𝑦, 0)
=∥ 𝑥 ∥ +∥ 𝑦 ∥ .

Thus the function ∥⋅∥ ∶ 𝑉 → 𝑅 is a norm on 𝑉. ∎

Remark 8.1.4 From proof of the above theorem, if a metric 𝑑 is induced by a norm, then
that norm is given as
∥ 𝒙 ∥ = 𝒅(𝒙, 𝟎) ∀𝒙∈𝑽.

Linear Space Metric Determined by a norm Norm which induces the

given metric
𝑽 𝒅: 𝑽 × 𝑽 → ℝ
over ℝ 𝒅(𝒙, 𝒚) =∥ 𝒙 − 𝒚 ∥ ∥ 𝒙 ∥= 𝒅(𝒙, 𝟎)

ℝ |𝑥 − 𝑦| |𝑥|

ℂ |𝑧 − 𝑧 ′ | |𝑧|

ℝ2 √(𝑥1 − 𝑦1 )2 + (𝑥2 − 𝑦2 )2 √𝑥12 + 𝑥22

ℝ2 |𝑥1 − 𝑦1 | + |𝑥2 − 𝑦2 | |𝑥1 | + |𝑥2 |

ℝ𝑛 √∑𝑛𝑖=1(𝑥𝑖 − 𝑦𝑖 )2 √∑𝑛𝑖=1 𝑥𝑖2

ℝ𝑛 ∑𝑛𝑖=1|𝑥𝑖 − 𝑦𝑖 | ∑𝑛𝑖=1|𝑥𝑖 |

ℝ𝑛 max |𝑥𝑖 − 𝑦𝑖 | max |𝑥𝑖 |

1≤i≤n 1≤i≤n

ℚ |𝑥 − 𝑦| |𝑥|

ℚ |𝑥 − 𝑦|𝑝 (𝑝-adic metric) |𝑥|𝑝

𝒞[0,1] max |𝑓(𝑥) − 𝑔(𝑥)| max |𝑓(𝑥)|

x∈[0,1] x∈[0,1]

1 1
𝒞[0,1]
∫ |𝑓(𝑥) − 𝑔(𝑥)| 𝑑𝑥 ∫ |𝑓(𝑥)| 𝑑𝑥
0 0
 INTRODUCTION TO METRIC SPACES 45

Example 8.1.5 (𝑴𝒆𝒕𝒓𝒊𝒄 𝒘𝒉𝒊𝒄𝒉 𝒄𝒂𝒏𝒏𝒐𝒕 𝒃𝒆 𝒐𝒃𝒕𝒂𝒊𝒏𝒆𝒅 𝒇𝒓𝒐𝒎 𝒂 𝑵𝒐𝒓𝒎)

1. The discrete metric on a linear space 𝑉(≠ {0}) cannot be determined by a norm.
2. The metric 𝑑 on ℝ given by
|𝑥 − 𝑦|
𝑑(𝑥, 𝑦) =
1 + |𝑥 − 𝑦|
cannot be determined by a norm.
3. If 𝑑 is a metric on a vector space 𝑉(≠ {0}) which is determined by a norm, then
the metric 𝑑̃ on 𝑉 given by
0 𝑖𝑓 𝑥 = 𝑦
𝑑̃ (𝑥, 𝑦) = {
1 + 𝑑(𝑥, 𝑦) 𝑖𝑓 𝑥 ≠ 𝑦
cannot be induced by a norm.
Also, the metric 𝑒 on 𝑉 given by
𝑑(𝑥 − 𝑦)
𝑒(𝑥, 𝑦) =
1 + 𝑑(𝑥, 𝑦)
cannot be determined by a norm.

Note that all the above metric does not satisfy the conditions of Theorem 8.1.3 and
hence are metric which cannot be induced by a norm.

Proposition 8.1.6 Let 𝑉 be a linear space and 𝑑: 𝑉 × 𝑉 → ℝ be any metric. The following
statements are equivalent:

i. 𝑑(𝑢 − 𝑣, 0) = 𝑑(𝑢, 𝑣) ∀ 𝑢, 𝑣 ∈ 𝑉
ii. 𝑑(𝑢 + 𝑤, 𝑣 + 𝑤) = 𝑑(𝑢, 𝑣) ∀ 𝑢, 𝑣, 𝑤 ∈ 𝑉.

Proposition 8.1.7 Let 𝑉 be a linear space and 𝑑: 𝑉 × 𝑉 → ℝ be any metric. The following
statements are equivalent:

i. 𝑑(𝛼𝑢, 0) = |𝛼|𝑑(𝑢, 0) ∀ 𝑢 ∈ 𝑉 and for any scalar 𝛼

ii. 𝑑(𝛼𝑢, 𝛼𝑣) = |𝛼|𝑑(𝑢, 𝑣) ∀ 𝑢, 𝑣 ∈ 𝑉 and for any scalar 𝛼.

In view of above propositions, Theorem 8.1.3 can be rewritten as:

Theorem 8.1.8 Let 𝑉 be a linear space and 𝑑: 𝑉 × 𝑉 → ℝ be any metric. Then 𝑑 is

determined by a norm if and only if following conditions are satisfied:

1. 𝑑(𝑥 + 𝑧, 𝑦 + 𝑧) = 𝑑(𝑥, 𝑦) ∀ 𝑥, 𝑦, 𝑧 ∈ 𝑉

2. 𝑑(𝛼𝑥, 𝛼𝑦) = |𝛼|𝑑(𝑥, 𝑦) ∀ 𝑥 ∈ 𝑉 and for any scalar 𝛼.

9. SOLVED PROBLEMS
Problem 1. Let 𝑋 be any non-empty set and 𝑑: 𝑋 × 𝑋 → ℝ be any real valued function.
Then 𝑑 is a metric on 𝑋 if and only if for all 𝑎, 𝑏, 𝑧 ∈ 𝑋, following two conditions hold :

1. 𝑑(𝑎. 𝑏) = 0 ⇔ 𝑎 = 𝑏
2. 𝑑(𝑎, 𝑏) ≤ 𝑑(𝑧, 𝑎) + 𝑑(𝑧, 𝑏).

Proof: Let 𝑑 be a metric on 𝑋, then 𝑑 satisfies all the four axioms of a metric.
Clearly, (1) holds. Now for any , 𝑏, 𝑧 ∈ 𝑋 ,
𝑑(𝑎, 𝑏) ≤ 𝑑(𝑎, 𝑧) + 𝑑(𝑧, 𝑏) [𝑇𝑟𝑖𝑎𝑛𝑔𝑙𝑒 𝐼𝑛𝑒𝑞𝑢𝑎𝑙𝑖𝑡𝑦]
 INTRODUCTION TO METRIC SPACES 46

= 𝑑(𝑧, 𝑎) + 𝑑(𝑧, 𝑏) [𝑆𝑦𝑚𝑚𝑒𝑡𝑟𝑖𝑐 𝑃𝑟𝑜𝑝𝑒𝑟𝑡𝑦]

𝑪𝒐𝒏𝒗𝒆𝒓𝒔𝒆𝒍𝒚, suppose (1) and (2) hold.

To show 𝑑 is symmetric, consider any 𝑎, 𝑏 ∈ 𝑋, then

𝑑(𝑎, 𝑏) ≤ 𝑑(𝑏, 𝑎) + 𝑑(𝑏, 𝑏) [𝐵𝑦 (2) 𝑡𝑎𝑘𝑖𝑛𝑔 𝑧 = 𝑏]

= 𝑑(𝑏, 𝑎) [∵ 𝑏𝑦 (1) 𝑑(𝑏, 𝑏) = 0]
Also,
𝑑(𝑏, 𝑎) ≤ 𝑑(𝑎, 𝑏) + 𝑑(𝑎, 𝑎, ) = 𝑑(𝑎, 𝑏).
Thus it follows that for any 𝑎, 𝑏 ∈ 𝑋, 𝑑(𝑎, 𝑏) = 𝑑(𝑏, 𝑎).
Now to verify that 𝑑 is a non-negative function, again consider 𝑎, 𝑏 ∈ 𝑋. Then
0 = 𝑑(𝑏, 𝑏) ≤ 𝑑(𝑎, 𝑏) + 𝑑(𝑎, 𝑏) = 2𝑑(𝑎, 𝑏)
⇒ 𝑑(𝑎, 𝑏) ≥ 0 .
Thus 𝑑(𝑎, 𝑏) ≥ 0 for all 𝑎, 𝑏 ∈ 𝑋.
Now for any 𝑎, 𝑏, 𝑧 ∈ 𝑋,
𝑑(𝑎, 𝑏) ≤ 𝑑(𝑧, 𝑎) + 𝑑(𝑧, 𝑏) [𝐹𝑟𝑜𝑚 (2)]
= 𝑑(𝑎, 𝑧) + 𝑑(𝑧, 𝑏).

Thus 𝑑 satisfies all the axioms and hence is a metric on 𝑋. ∎

Problem 2. Suppose that 𝑑 is a metric on a set 𝑋. Prove that

|𝑑(𝑥, 𝑦) − 𝑑(𝑧, 𝑤)| ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑦, 𝑤) ∀ 𝑤, 𝑥, 𝑦, 𝑧 ∈ 𝑋 .

Proof: Consider for any , 𝑥, 𝑦, 𝑧 ∈ 𝑋 ,

𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑦) [𝑇𝑟𝑖𝑎𝑛𝑔𝑙𝑒 𝐼𝑛𝑒𝑞𝑢𝑎𝑙𝑖𝑡𝑦]

≤ 𝑑(𝑥, 𝑧) + [𝑑(𝑧, 𝑤) + 𝑑(𝑤, 𝑦)] [𝑇𝑟𝑖𝑎𝑛𝑔𝑙𝑒 𝐼𝑛𝑒𝑞𝑢𝑎𝑙𝑖𝑡𝑦]
= 𝑑(𝑥, 𝑧) + 𝑑(𝑧, 𝑤) + 𝑑(𝑦, 𝑤) [𝑆𝑦𝑚𝑚𝑒𝑡𝑟𝑖𝑐 𝑃𝑟𝑜𝑝𝑒𝑟𝑡𝑦]
⇒ 𝑑(𝑥, 𝑦) − 𝑑(𝑧, 𝑤) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑦, 𝑤) . − − − − − − − −(𝑨)

Again,
𝑑(𝑧, 𝑤) ≤ 𝑑(𝑧, 𝑥) + 𝑑(𝑥, 𝑤)
≤ 𝑑(𝑥, 𝑧) + [𝑑(𝑥, 𝑦) + 𝑑(𝑦, 𝑤)]
⇒ 𝑑(𝑧, 𝑤) − 𝑑(𝑥, 𝑦) ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑦, 𝑤) − − − − − − − (𝑩)

From (𝐴) and (𝐵) it follows that

|𝑑(𝑥, 𝑦) − 𝑑(𝑧, 𝑤)| ≤ 𝑑(𝑥, 𝑧) + 𝑑(𝑦, 𝑤) ∀ 𝑤, 𝑥, 𝑦, 𝑧 ∈ 𝑋 . ∎

Problem 3. Prove that the function 𝑑 ∶ ℝ2 × ℝ2 → ℝ given by

0 𝑖𝑓 𝑎=𝑏

𝑑(𝑎, 𝑏) = √(𝑏1 − 𝑎1 )2 + (𝑏2 − 𝑎2 )2 𝑖𝑓 𝑎 ≠ 𝑏 𝑎𝑛𝑑 𝑎, 𝑏 ∈ ℝ2 ∖ {(0,0)}

{1 + √(𝑏1 − 𝑎1 )2 + (𝑏2 − 𝑎2 )2 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

is a metric on ℝ2 .
 INTRODUCTION TO METRIC SPACES 47

Proof: By definition, 𝑑(𝑎, 𝑏) ≥ 0 ∀ 𝑎, 𝑏 ∈ ℝ2 and therefore P1 holds.

Since 𝑎 = 𝑏 ⇒ 𝑑(𝑎, 𝑏) = 0, therefore to prove P2, it is enough to prove that
𝑑(𝑎, 𝑏) = 0 ⇒ 𝑎 = 𝑏 𝑓𝑜𝑟 𝑎𝑛𝑦 𝑎, 𝑏 ∈ ℝ2 .
For this first observe that

𝑑(𝑎, 𝑏) ≥ √(𝑏1 − 𝑎1 )2 + (𝑏2 − 𝑎2 )2 ∀ 𝑎 = (𝑎1 , 𝑎2 ), 𝑏 = (𝑏1 , 𝑏2 ) ∈ ℝ2 − − − − − (∗)

Now
𝑑(𝑎, 𝑏) = 0
⇒ √(𝑏1 − 𝑎1 )2 + (𝑏2 − 𝑎2 )2 = 0 [𝑈𝑠𝑖𝑛𝑔 (∗) ]
⇒ 𝑏1 = 𝑎1 𝑎𝑛𝑑 𝑏2 = 𝑎2
⇒ 𝑎 = 𝑏.

Hence P2 holds.

Also, from definition, it is clear that 𝑑 is a symmetric function i.e.,

𝑑(𝑎, 𝑏) = 𝑑(𝑏, 𝑎) ∀ 𝑎, 𝑏 ∈ ℝ2 . (𝑷𝟑)
Now for triangle inequality (P4), consider any 𝑎 = (𝑎1 , 𝑎2 ), 𝑏 = (𝑏1 , 𝑏2 ), 𝑐 = (𝑐1 , 𝑐2 ) ∈ ℝ2 . There
are three possibilities:

Case I. 𝑎 = 𝑏

Then, 𝑑(𝑎, 𝑏) = 0 ≤ 𝑑(𝑎, 𝑐) + 𝑑(𝑐, 𝑏) .

Case II. 𝑎 ≠ 𝑏 and 𝑎, 𝑏 ∈ ℝ2 ∖ {(0,0)}

Then
𝑑(𝑎, 𝑏) = √(𝑏1 − 𝑎1 )2 + (𝑏2 − 𝑎2 )2
2 2
= √((𝑏1 − 𝑐1 ) + (𝑐1 − 𝑎1 )) + ((𝑏2 − 𝑐2 ) + (𝑐2 − 𝑎2 ))

≤ √(𝑏1 − 𝑐1 )2 + (𝑏2 − 𝑐2 )2 + √(𝑐1 − 𝑎1 )2 + (𝑐2 − 𝑎2 )2 [𝐴𝑝𝑝𝑙𝑦𝑖𝑛𝑔 𝑻𝒉𝒆𝒐𝒓𝒆𝒎 𝑨]

≤ 𝑑(𝑐, 𝑏) + 𝑑(𝑎, 𝑐) [𝐹𝑟𝑜𝑚 (∗)]
= 𝑑(𝑎, 𝑐) + 𝑑(𝑐, 𝑏).

Case III. Exactly one of 𝑎 and 𝑏 is (0,0).

Without loss in generality, assume that 𝑎 ≠ (0,0) and 𝑏 = (0,0).

Subcase (i) 𝑐 = (0,0) .
Then
𝑑(𝑎, 𝑏) = 𝑑(𝑎, 𝑐) [∵ 𝑏 = 𝑐]
= 𝑑(𝑎, 𝑐) + 𝑑(𝑐, 𝑏). [∵ 𝑑(𝑐, 𝑏) = 0]

Subcase (ii) 𝑐 ≠ (0,0).

Then
𝑑(𝑎, 𝑏) = 𝑑((𝑎1 , 𝑎2 ), (0,0))
 INTRODUCTION TO METRIC SPACES 48

= 1 + √𝑎12 + 𝑎22

2 2
= 1 + √(𝑐1 + (𝑎1 − 𝑐1 )) + (𝑐2 + (𝑎2 − 𝑐2 ))

≤ 1 + √𝑐12 + 𝑐22 + √(𝑎1 − 𝑐1 )2 + (𝑎2 − 𝑐2 )2 [𝐴𝑝𝑝𝑙𝑦𝑖𝑛𝑔 𝑻𝒉𝒆𝒐𝒓𝒆𝒎 𝑨]

= 𝑑(𝑐, 𝑏) + 𝑑(𝑐, 𝑎)
= 𝑑(𝑎, 𝑐) + 𝑑(𝑐, 𝑏).

Hence in all the three cases we get,

𝑑(𝑎, 𝑏) ≤ 𝑑(𝑎, 𝑐) + 𝑑(𝑐, 𝑏) ∀ 𝑎 = (𝑎1 , 𝑎2 ), 𝑏 = (𝑏1 , 𝑏2 ), 𝑐 = (𝑐1 , 𝑐2 ) ∈ ℝ2 .

It follows that 𝑑 is a metric on ℝ2 . ∎

10. SUMMARY
In this chapter, we have introduced the concept of metric space, discussed the motivation
behind studying this concept. We also discussed how this concept is being used in real life
scenario. Being closely related to the notion of distance, it has many applications in the
field of navigation. One such instance where it is being associated with the navigation of
aeroplanes is discussed. We have discussed various metrics on the set of reals, on ℝ𝑛 , on
the set of functions etc. to illustrate the abstract nature of the distance. Later in the
chapter, the notions of metric subspace and superspace were introduced and a method of
finding a metric superspace from a given metric space is discussed. In fact, we observed
that we can have uncountable extensions of a given metric to some superset of that space.
We also touched upon the concept of isometry i.e., distance preserving functions and
discussed some standard examples associated with this concept. Finally we discussed
special types of metrics determined by a norm on a vector space, which relates the
algebraic properties of the vector space with its geometric properties. We hope that the
content presented in the chapter would be of immense help to the students who are
beginners in the subject.

11. PROBLEMS
Question 1. Show that the function 𝑑: ℝ × ℝ → ℝ defined as

|𝑥| + |𝑥 − 𝑦| + |𝑦| 𝑖𝑓 𝑥 ≠ 𝑦
𝑑(𝑥, 𝑦) = { ∀ 𝑥, 𝑦 ∈ ℝ
0 𝑖𝑓 𝑥 = 𝑦
is a metric on ℝ.

Question 2. Prove that the function 𝑑 ∶ ℝ × ℝ → ℝ given by

|𝑥 − 𝑦|
𝑑(𝑥, 𝑦) = ∀ 𝑥, 𝑦 ∈ ℝ
√1 + 𝑥 2 √1 + 𝑦 2
is a metric on ℝ.
 INTRODUCTION TO METRIC SPACES 49

Question 3. Prove that the function 𝑑 ∶ ℝ2 × ℝ2 → ℝ given by

|𝑥1 − 𝑦1 | 𝑖𝑓 𝑥2 = 𝑦2
𝑑(𝑥, 𝑦) = { ∀ 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ) ∈ ℝ2
|𝑥1 | + |𝑥2 − 𝑦2 | + |𝑦1 | 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

is a metric on ℝ2 .

Question 4. Prove that the function 𝑑 ∶ ℝ2 × ℝ2 → ℝ given by

|𝑥1 − 𝑦1 | + |𝑥2 − 𝑦2 |
𝑑(𝑥, 𝑦) = ∀ 𝑥 = (𝑥1 , 𝑥2 ), 𝑦 = (𝑦1 , 𝑦2 ) ∈ ℝ2
√1 + 𝑥12 + 𝑥22 √1 + 𝑦12 + 𝑦22
is a metric on ℝ2 .

Question 5. Let (𝑋, 𝑑) be any metric space and 𝑈 and 𝑉 be disjoint subsets of 𝑋 such
that 𝑋 = 𝑈 ∪ 𝑉. Show that the function 𝐷 ∶ 𝑋 × 𝑋 → ℝ given by
1 + 𝑑(𝑥, 𝑦) 𝑖𝑓 𝑒𝑥𝑎𝑐𝑡𝑙𝑦 𝑜𝑛𝑒 𝑎𝑚𝑜𝑛𝑔 𝑥 𝑎𝑛𝑑 𝑦 𝑖𝑠 𝑖𝑛 𝑈
𝐷(𝑥, 𝑦) = {
𝑑(𝑥, 𝑦) 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
is a metric on 𝑋.

Question 6. Let 𝑋 be the set of all sequences of real numbers. Prove that the function
𝑑 ∶ 𝑋 × 𝑋 → ℝ given by
∞
1 |𝑥𝑛 − 𝑦𝑛 |
𝑑(𝑥, 𝑦) = ∑ 𝑛
∀ 𝑥 = {𝑥𝑛 }𝑛≥1 , 𝑦 = {𝑦𝑛 }𝑛≥1 ∈ 𝑋
2 1 + |𝑥𝑛 − 𝑦𝑛 |
𝑛=1

is a metric on 𝑋.

Question 7. Let 𝑋 be any set and 𝑑1 , 𝑑2 , … , 𝑑𝑛 be metrics on 𝑋. Show that the function
𝑒: 𝑋 × 𝑋 → ℝ defined as
𝑛

𝑒(𝑥, 𝑦) = ∑ 𝑑𝑖 (𝑥, 𝑦)
𝑖=1
is a metric on 𝑋.

Question 8. Let 𝑋 be any set and 𝑑1 , 𝑑2 , … , 𝑑𝑛 be metrics on 𝑋. Prove that the function
𝑒: 𝑋 × 𝑋 → ℝ defined as
𝑒(𝑥, 𝑦) = max {𝑑𝑖 (𝑥, 𝑦)}
1≤i≤n

is a metric on 𝑋.

Question 9. Let 𝑋 be any set and 𝑑1 , 𝑑2 , … , 𝑑𝑛 be metrics on 𝑋. Prove that 𝑒: 𝑋 × 𝑋 → ℝ

defined as
𝑛

𝑒(𝑥, 𝑦) = √∑[𝑑𝑖 (𝑥, 𝑦)]2

𝑖=1

is a metric on 𝑋.

Question 10. Prove that the function 𝑑 ∶ ℝ × ℝ → ℝ defined as

𝑑(𝑥, 𝑦) = |sin(𝑥 − 𝑦)| ∀ 𝑥, 𝑦 ∈ ℝ

is not a metric on ℝ.
 INTRODUCTION TO METRIC SPACES 50

12. REFERENCES
[1] Micheal O Searcoid, Metric Spaces, Springer International Edition, New Delhi, 2008.

[2] Satish Shirali and Harkrishan L. Vasudeva, Metric Spaces, Springer, 2006.

[3] George F. Simmons, Introduction to Topology and Modern Analysis, McGraw-Hill

Book Co., New York, 1963.

[4] E.T. Copson, Metric Spaces, Cambridge University Press, Cambridge, 1968.
[5] M.K Singal and Asha Rani Singal, Topics in Analysis II (Metric Spaces), R. Chand
and Co., New Delhi, 2005.

View publication stats