You are on page 1of 192

KARNATAKA STATE OPEN UNIVERSITY

Mukthagangothri, Mysuru – 570006

M.Sc. MATHEMATICS (CBCS)


(FIRST SEMESTER)

Course- MMDSC 1.2

Real Analysis I
M.Sc. MATHEMATICS (CBCS)
FIRST SEMESTER

Course: MMDSC 1.2


REAL ANALYSIS - I

i
Programme Name: M.Sc. Mathematics (CBCS) Year/Semester: I Semester
Course Code: MMDSC 1.2 Course Name: Real Analysis- I
Credit: 4 Unit Number : 1-16
COURSE DESIGN COMMITTEE
Dr. Vidyashankar S. Chairman
Vice Chancellor
Karnataka State Open University
Mukthagangothri, Mysuru-570006
Prof. Ashok Kamble Member
Dean (Academic)
Karnataka State Open University
Mukthagangothri, Mysuru-570006
Dr. Pavithra. M Course coordinator
Assistant Professor
DoS in Mathematics, KSOU, Mukthagangothri, Mysuru-06
EDITORIAL COMMITTEE
1. Dr. K. Shivashankara Chairman
BOS Chairman(PG), DoS in Mathematics, KSOU.
Associate Professor, Yuvaraja College,
University of Mysore, Mysuru-06

2. Mr. S. V. Niranjan Member & Convener


Coordinator, (DoS in Mathematics)
Assistant Professor, DoS in Physics
KSOU, Mysuru-06

3. Dr. Pavithra. M Member


Assistant Professor
DoS in Mathematics, KSOU, Mysuru-06

4. Dr. Chandru Hegde Member


Assistant Professor, DoS in Mathematics,
Mangalagangotri, Mangaluru.

COURSE WRITER
Dr. K. Shivashankara Block 1.2A to Block 1.2 D
Associate Professor ( Block I - IV)
Department of Studies in Mathematics (Unit 1 to Unit 16)
Yuvaraja College, Mysuru -06
ii
COURSE EDITOR
Prof. K. R. Vasuki
Professor
Department of Studies in Mathematics,
University of Mysore, Mysuru -06
COPYRIGHT
The Registrar
Karnataka State Open University
Mukthagangothri, Mysuru-570006
Developed by the Department of Studies in Mathematics under the guidance of Dean
(Academic), KSOU, Mysuru.
Karnataka State Open University, 2022.
All rights reserved. No part of this work may be reproduced in any form or any other means,
without permission in writing from the Karnataka State Open University.
Further information on the Karnataka State Open University Programmes may be obtained from
the University’s Office at Mukthagangothri, Mysuru – 570 006.

iii
TABLE OF CONTENTS
Page No.

BLOCK 1.2A (BLOCK- I)


The extended real number system, Euclidean space 1-15
Unit-1
Countable and uncountable sets. 16-23
Unit-2

Unit-3 Basic topology 24-40


Compact, Connected and Perfect sets. 41-53
Unit-4

BLOCK 1.2B (BLOCK- II)

Unit-1 Convergent sequences, subsequences and monotone sequences 1-15

Unit-2 Cauchy sequences, Upper and lower limits, some special sequences 16-26

Unit-3 Series 27-39

Unit-4 Power series and Algebra of series 40-53

BLOCK 1.2C (BLOCK- III)

Unit-1 Limit and Continuity 1-15

Unit-2 Continuity and compactness 16-25

Unit-3 Continuity and connectedness 26-32

Unit-4 Discontinuity and monotonic functions 33-43

BLOCK 1.2D (BLOCK- IV)

Unit-1 Differentiation 1-12

Unit-2 Mean value theorem 13-23

Unit-3 The continuity of derivative and L’Hospital’s rule 24-31

Unit-4 Derivatives of higher order and differentiation of vector valued functions. 32-39

iv
PRELUDE

Real Analysis is the branch of mathematics that deals with inequalities and limits. The present

course deals with the most basic concepts in analysis. The goal of the course is to acquaint the

reader with rigorous proofs in analysis and also to set a firm foundation for calculus. This course

shows the importance of calculus. It is here to give you a good understanding of the concept of a

limit, the derivative, and the integral. We start with a discussion of the real number system, most

importantly its completeness property and then discuss the simplest form of a limit, the limit of a

sequence. We study functions of one variable, continuity, and the derivative. Then the derivative

of higher order and differentiation of vector valued functions is discussed.

v
REAL ANALYSIS-I

BLOCK I
Unit I
The extended real number system, Euclidean space

1.1.0. Objectives
In this unit we assume the existence of real numbers and study their prop-
erties. Then we introduce extended real number system and Euclidean.

1.1.1. Introduction
Mathematical analysis deals with concepts such as convergence, continuity,
differentiation and integration. Such concepts must be based on on accurately
defined number concept. In part particular, the real analysis is based on the
real number system. So we begin our study with a discussion of the real
number system.
There are several methods to introduce real numbers. One such method
starts with integers. We shall not, however, enter into any discussion of
the axioms that govern the arithmetic of integers. We use the integers to
construct numbers like m/n where m and n are integers and n 6= 0. Such
numbers are chilled rational numbers. The rational numbers system is inad-
equate for many purpose. For instance there is no rational whose square is 2.
This leads to the introduction of so called irrational numbers. The rational
and irrational numbers together constitute the real runnier system.
These matters arc very important part ot the foundations of mathematics.
But in most phases of analysis it is only the properties of real numbers that
play pivotal role. Therefore, we concentrate on the properties of real numbers
than the methods used to construct them.

1
For convenience we use some elementary set notations and terminologies.
By a set we mean a collection of objects viewed as a single entity. The sets
are usually denoted by capital letters A, B, C, . . . . The objects are usually
denoted by small case letters a, b, c, . . . . If an object x belongs to a set A we
write ’x∈ A’and if the object does not belong to A, we write ‘x∈A’. / If A and
B are two sets and if every element of A is in B, we say that A is a subset
of B we write ’A⊆ B’. If B contains an element that is not in A, then A is
called a proper subset of B and we write ’A⊂ B’. A set with no objects is
called an empty set. It is denoted by ∅. We agree that empty set is a subset
of every set.
In what follows we denote the set of all natural numbers (positive integers)
by N the set all integers by Z, the set of all rational numbers by Q, the set
of all irrational numbers by I and the set of all real numbers by R.

1.1.2. Ordered sets


Definition 1.1.0.1 Let S be a set. An order on S is a relation, denoted by
‘<’ with the following two properties:
(i) If x ∈ S and y ∈ S then one and only one of the statements

x < y, x = y, y<x

is true.
(ii) If x, y, z ∈ S and if x < y, y < z, then x < z.

The statement x < y may be read as ”x is less than y”. Some times we
write y > x in place of x < y. ’The notation x ≤ y indicates that x < y or
x = y, without specifying which one of these two is to hold.

Definition 1.1.0.2 An ordered set is a set S in which an order is defined.


For example R is an ordered set if r < s is defined to mean that s − r is a
positive real number. Q is also an ordered set if r < s is defined to mean that
s − r is a positive rational number.

Definition 1.1.0.3 Suppose S is an ordered set and E ⊂ S. If there exists


β ∈ S such that x ≤ β for every x ∈ E, we say that E is bounded above and
call β an upper bound of E.

2
Definition 1.1.0.4 Suppose S is an ordered set and E ⊂ S. If there exists
an α ⊂ S such α ≤ x for every x ∈ E, we say that E is bounded below and
call α lower bound of E.

Definition 1.1.0.5 Suppose S is an ordered set , E ⊂ S and E is bounded


above. Suppose there exists a β ⊂ S with the following properties:
(i) β is an upper bound of E.
(ii) If γ < β t hen γ is not an up per bound of E.
Then β is called the least upper bound of E or the supremum of E, and
we write β= supE or β= lub E.

Definition 1.1.0.6 . Suppose S is an ordered set, E ⊂ S Sand E is bounded


below. Suppose there exists an α ∈ S with the following properties:
(i) α is a lower bound of E.
(ii) If α < δ then δ is not a lower bound of E.
Then a is called the greatest lower bound of E or the infimum of E and we
write
α= inf E or α = glb E.

Example 1 Consider the subset E = { 41 | n = 1, 2; ...} of the ordered set R


Clearly sup E = l and inf E = 0. Note that 1 ⊂ E but 1 ∈ / E.

Example 2 Let E = {r ∈ Q | r < 0}. Then sup E = 0. E is not bounded


below. Note that 0 ∈
/ E.

Definition 1.1.0.7 An ordered set S is said to have the least upper bound
property if the following is true: If E ⊂ S, E 6= ∅ and E is bounded above,
then sup E exists in S. For exam1ple, R has the least upper bound property.

Remark: If every nonempty subset of S that is bounded below has the in-
fimum. in S, then S is said to have the greatest lower bound property.

3
Theorem 1.1.0.1 The set Q does not have the least upper bound property.

P roof . First we show that there is no rational whose square is 2:


Suppose that there is a rational m
n
where (m,n) = 1 such that
m2
(1) n2
=2
Then (1) implies
(2) m2 = 2n2
This shows that m2 is even. Hence m is even (if m is odd, say m = 2k + 1,
then m2 = 4k 2 + 4k + 1 = 2(2k 2 + 2k) + 1, is odd). So let m = 2k. Then (2)
implies
4k 2 = 2n2
or
2k 2 = n2
This shows that m2 is even and hence n is even. Thus m and n are both
even, contradicting the fact that (m, n) = 1. Hence (1) is impossible for any
rational m/n
Next we shall consider two sets A and B:
A = {p ∈ Q|0 < p and p2 < 2} and B = {p ∈ Q|p2 > 2}.
We shall show that A contains no largest number and B contains no smallest.
To do this, we associate with each rational p > 0 the number
2 −2
(3) q = p − pp+2 = 2p+2
p+2
Then 2 −2)
(4) q 2 − 2 = 2(pp+2
If p ∈ A then p2 − 2 < 0. (3) shows that q > p and (4) shows that q 2 < 2.
Thus q ∈ A. If p ∈ B then p2 − 2 > 0. (3) shows that O < q < p and (4)
shows that q 2 > 2. Thus q ∈ B. Since the least upper bound of A, if exists,
is either the largest element of A or the smallest element of B, we conclude
that the subset A of Q has no least upper bound in Q.

Theorem 1.1.0.2 Suppose S is an ordered set. If S has the least upper


bound property then it also has the greatest lower bound property.
Proof. Let B ⊂ S, B 6= 0 and B is bounded below. Denote by L the set of
all lower bounds of B.
Since B is bounded below, L is not empty . For each y ∈ L, we have y ≤ x
for every x ∈ B. So every x ∈ B is an upper bound of L. Thus L is bounded

4
above. The hypothesis about S implies therefore that sup L exists in S. Call
it α.

If γ < α then γ is not an upper bound of L. Hence γ ∈ / B. It follows


that α ≤ x for every x ∈ B . Thus x ∈ L. If α < β then β ∈
/ L , since α is
an upper bound of L.

Thus α ∈ L β ∈ / L if β > α . In other words , α is a lower bound of B


, but β is not if β > α. This means that α= inf B.

1.1.1 Fields
Definition 1.1.1.1 A field is a set F with two operations, called addition
and multiplication, which satisfy the following field axioms (A),(M) and (D
):
(A) Axioms for addition
(Al) If x ∈ F and y ∈ F , then their sum x + y ∈ F .
(A2) Addition is commutative: x + y = y + x for all x, y ∈ F .
(A3) Addition is associative: (x + y) + z = x + (y + z ) for all x, y, z ∈ F .
(A4) F contains an element 0 such that 0 + x = x, for every x ∈ F .
J (A5) To every x ∈ F corresponds an element −x in F such that
x + (−x) = 0.
(M) Axioms for multiplication
(Ml) If x ∈ F and y ∈ F , then their product xy ∈ F .
(M2) Multiplication is commutative: xy = y.1: for all x, y ∈ F .
(M3) Multiplication is associative: ( xy )z = x(yz) for all x, y, z ∈ F .
(M4) F contains an element 1 6= 0 such that 1 ˆ x = x, for every x ∈ F .
(M5) If x ∈ F and x 6= 0 then there exists an element 1/x ∈ F such that x
·(l / x) = l.
(D) The distributive law x( y + z ) = xy +xz holds for all x, y, z ∈ F .
Example 3 The set R with respect to usual addition and multiplication
forms a field .
Example 4 T he set Q with respect to usual addition and multiplication
forms a field.
Example 5 The set Z with respect to usual addition and multiplication does
not form a field.

5
Proposition 1.1.1. The axioms for addition imply the following state-
ments. (a) If x + y = x + z then y = z (b) If x + y = x then y = 0 (c) If x
+ y = 0 then y = -x (d) -(-x) = x.
Proof.

y =0+y (byA4)
= (−x + x) + y (byA5)
= −x + (x + y) (byA3)
= −x + (x + z) (by hypothesis)
= (−x + x) + z (byA3)
= 0 + z (byA5)
= z (byA4).

Thus x + y = x + z im plies y = z, proving (a). Take z=0 in (a) to obtain,


(b).
Ta ke z = -x in (a) to obtain (c). Since - x + x = 0, (c) (with - x in place
of x) gives (d ).
Remark. Statement. (a) is a cancellation law (b) asserts the uniqueness
of 0 of A4 and (c) does the same for - x of A5.

Proposition 1.1.2. The axioms for multiplication imply the following


statements.
(a) If x 6= 0 and xy = xz then y = z
. (b) If x 6= 0 and xy = x then y = 1
(c) If x 6= 0 and x y = l then y = 1/ x
(d) If x 6= 0 then 1/(1/x) = x.
Proof. The proof is similar to that of proposition 1.1.11 and we therefore
leave it to the reader.

Proposition 1.1.3 The field axioms imply the following statements, for
any x, y, z ∈ F .
(a) 0·x= 0
(b) If x 6= 0 and y 6= 0 thenxy 6= 0.
(c) ( - x ) y = -(xy) = x(-y)
(cl) (-x)(-y) = xy.
Proof. 0x+ 0x= (0 + 0)x = 0x. Hence (b) of proposition (1.1.11) implies
that O ·x = 0. This proves (a).

6
Next, assume x 6= 0, x 6= 0 but xy = 0. Then (a) gives
     
1 1 1 1
1= xy − 0=0
y x y x

A contradiction. Thus (b) holds.


Since

(−x)y + xy = (−x + x)y = 0 · y = 0

(c) of Proposition (1.1.1) yields (-x)y = - ( xy ). The proof of second part of


(c) is similar. Using (c) one can easily prove (d).

Definition 1.1.1.2 An ordered field is a field F which is also an ordered set


such that
(i) x + y < x + z, if x , y, z ∈ F and y< z ,
(ii) xy > 0 if x ∈ F, y ∈ F, x> 0 and y>0.
For example R. and Q are ordered fields.

Proposition 1.1.4 In every ordered field we have the following. (a) If x> 0
then -x < 0, and vice versa.
(b) If x> 0 and y < z then xy < xz
(c) If x < 0 and y < z then xy > xz
(d) If x6= 0 then x2 > 0. In particular, 1 ¿ 0.
(e) If O < x < y then O< 1/y <l / x.
Proof. If x > 0 then O = - x + x> -x + 0, so t hat -x ¡ 0 (by the definition
1.1.14 and A4). If x < 0 then 0= - x + x < - x + 0, so that -x> 0 (by the
definition 1.1.14 and A4). This proves (a). Since z ¿ y, we have z-y > y - y
= 0. Hence x( z - y)> 0 and therefore

xz = x(z − y) + xy > 0 + xy = xy.

(c) follows on using (a) and (b) of proposition 1.1.15 and (c) of Proposition
1.1.13. If x> 0, part (ii) of Definition 1.1.14 gives x2 > 0. If x ¡ 0, then by
(a) -x > 0. Thereforex2 = (-x)(-x) > 0. Since 1 = 12 , 1 > 0. This proves

7
   
(d). Finally if y > 0 and v ≤ 0 then yv ≤ 0. But y y1 > 0. Hence y1
1

> 0. Likewise
 x
> 0. If we multiply both sides of the inequality x > y by
1 1

x y
¿ 0, we obtain (e)
Remark. Note that both R and Q are ordered fields but R has the least upper
bound property whereas Q does not have the least upper bound property.
In fact Q is a sub-field of R in the sense that Q itself 1s a field under the
operations of addition and multiplication defined in R.

Definition 1.1.1.3 Let x ∈ R Then the absolute value of x is denoted by |x|


and is defined by

−x x≤0
|x| =
x x≥0
Theorem 1.1.1.1 Let a ≥ 0. Then |x|≤ a if and only if -a≤x≤a.
Proof. From the definition of |x|, we have the in equality -|x|≤x ≤ |x|
. If |x| ≤ a, then we can write - a ≤-|x|≤x ≤ |x|≤a. Next let -a≤x≤a. If x
≥ 0, we have |x| = x ≤a. If x ¡ 0, we have |x|= - x ≤ a. Thus in either case
we have |x| ≤a.

Theorem 1.1.1.2 .(The triangle inequality) For all x , y ∈R, we have |x+y|
6= |x| + |y| .
Proof. We have - |x| ≤ x ≤ |x| and - |y| ≤ y ≤ |y|. Adding and then using
Theorem 1.1.1.1 we obtain
|x + y| =
6 |x| + |y|.

Remark. From Theorem1.1.1.2 it follows that ||a| - |b|| ≤ |a − b|, for


a,b E ∈R and for x1 , x2 , · · · , xn ∈R,
|x1 + x2 + · · · + xn | ≤ |x1 | + |x2 | + · · · + |xn |,
Theorem 1.1.1.3 (Cauchy Schwarz inequality) If a1 , a2 , · · · , an and b1 , b2 , · · · , bn
are arbitrary real numbers, then we have
n
!2 n
! n !
X X X
ak b k ≤ a2k ab2k
k=1 k=1 k=1

8
Moreover, if some ai 6=0 then equality holds if and only if there is a real
number x such that ak x + bk = 0 for k= 1, 2, ... , n.

Proof. Since square of a real number is nonnegative, we have


n
!
X
(ak x + bk )2
k=1

for every real x. This inequality can be written in the form

Ax2 + 2Bx + C ≥ 0
where
n
X n
X n
X
A= a2k , B= ak b k , C= b2k ,
k=1 k=1 k=1

If A ¿ 0, put x = -B / A to obtain B 2 AC ≤ 0 which is t he desired in


equality. If A=0, the proof is trivial. The equality in (1) and hence in the
theorem holds if and only if ak x + bk = 0 for every k.

Theorem 1.1.1.4 (Archimedean property) If x and y are any two real num-
bers with x > 0, then there exists a positive integer n such that n.x > y.
Proof. Let A = {nx — n ∈ Z+ }. If (1) were false then y would be an upper
bound of A. But then A has a least upper bound in R. Put α = supA. Since
x > 0, α - x > α and α- x is not an upper bound of A. Hence α- x > mx
for some positive integer m. But then a>(m+ 1)x ∈A, contradicting the fact
that α is an upper bound of A.

Theorem 1.1.1.5 Between any two real numbers there exists a rational
number. In other words Q is dense in R.
Proof. Let x and y be any two real numbers with x < y. Then y - x > 0. By
the Archimedean property there is a positive integer n such that
n(y − x) > 1.

9
Again by the Archimedean property, there exist positive integers m1 and m2
such that m1 > nx, m2 > −nx . Then −m2 < nx < m1 . Hence there exists
an integer m such that

m − 1 ≤ nx < m.

Combining the above inequalities we obtain

nx < m ≤ 1 + nx < ny.

Since n > 0, it follows that


m
x< <y
n
where m
n
is a rational number.

Theorem 1.1.1.6 For every real x > 0 and every integer n > 0 there is one
and only one real y such that y n = x. This number y is written as y n = x.
Proof. First we shall prove the existence of a real y. So let

E = t ∈ R|t > 0, tn < Xx.

Given x, put t = x/(1 + x). Then O < t < 1 ard so, tn < t < x. Thus t ∈ E
and E is nonempty. If t > 1 + x, then t > 1 and so tn > t > x. Therefore t
∈.
/ E. Thus 1 + x is an upper bound of E. By the least upper bound property
of R, sup E = y
exists in R.
Now, the identity bn − an = (b − a)(bn−1 + bn−2 a + · · · + an−1 +) yields the
inequality

bn − an ≤ (b − a)nbn−1

when 0 < a < b.


Assume y n < x . Choose h so that O < h < l and
x − yn
h<
n(y + 1)n−1

10
Put a= y , b = y + h. Then we have
(y + h)n − y n ≤ hn(y + h)n−1 < hn(y + 1)n−1 < x − y n
Thus (y + h)n < x and so y+ h ∈E. Since y + h > y, this contradicts the
fact that y is an upper bound of E. Assume y n > x. Put
yn − x
k= .
ny n−1
Then O < k < y . If t≥ y - k , we conclude that
y n − tn ≤ y n − (y − k)n < kny n−1 = y n − x
Thus tn> x and t ∈ / E. It follows that y -k is an upper bound of E. But y - k
< y, which contradicts the fact that y is the least upper bound of E. Thus both
y n < x and y n > x are not possible. Hence we must have y n = x. Finally, if
0 < y1 < y2 then y1n < y2n , the uniqueness of y follows.

Corollary 1.1.1.1 If a and b are positive real numbers and n is a positive


integer , then
(ab)1/n = (a)1/n (b)1/n
Proof. Put α= (a)1/n ,, β = (b)1/n . Then

ab = αn β n = (αβ)n
since the multiplication is commutative. The uniqueness assertion of The-
orem ?? shows therefore that

(ab)1/n = αβ = (a)1/n (b)1/n

1.1.2 The extended real number system


Definition 1.1.2.1 The extended real number system consists o the real field
R and two symbols, +∞ and −∞. We preserve the original order in R, and
define
−∞ < x < +∞
for every x ∈R.

11
Remarks: ln the extended real number system every subset is bounded
above and hence every non-empty subset has the least upper bound. That
is, if E is a non empty set of real numbers which is not bounded above in
R, then sup E = +∞ in the extended real number system. Similarly, every
nonempty subset has the greatest lower bound in the extended real number
system. The extended real number system does not form a field, but it is
customary to make the following conventions:
x x
(a) If x is real then X + ∞ = ∞, x - ∞ = - ∞, +∞ = −∞ =0.
(b) If x > 0 then x · (+∞) = +∞, x · (- ∞) = -∞.
(c) If x < 0 then x · (∞) = -∞, x · (-∞) = +∞.

1.1.3 Euclidean spaces


Definition 1.1.3.1 For each positive integer n, let Rn be the set of all or-
dered n-tuples
X = (x1 , x2 , · · · , xn ),
wherex1 , x2 , · · · , xn are real numbers , called the coordinates of x. The
elements of Rn are called points or vectors especially when n > l. We shall
denote vectors by boldfaced letters. If y = (y1 , y2 , · · · , yn ) and a is a real
number we define the addition of vectors and the multiplication of a vector
by a scalar as follows:

x + y = (x1 + y1 , x2 + y2 , · · · , xn + yn
and
αx = (αx1 , αx2 , · · · , αxn )
so that x + y , αx ∈Rn .
One can easily show that with respect to these operations Rn forms a vector
space over the field R. The zero element of Rn (sometimes called the origin
or null vector) is the point 0, all of whose coordinates are 0.

We define inner product (or scalar product ) of x and y by

n
X
X ·Y = xi yi
i=1

12
and the norm of x by

n
!1/2
X
||x|| = (x · x)1/2 = x2i
i=1

Rn , with the inner product and the norm, is called Euclidean n-space.

Theorem 1.1.3.1 Suppose x, y, z ∈ Rn , and α is real. Then we have


(a) ||x||≥0,
(b) ||x||= 0 if and only if x = 0,
(c) ||αx|| = |α|||x|| ,
(d) |x · y| ≤ ||x|| ||y||
(e) ||x + y|| ≤ ||x|| ||y||
(f ) ||x − y|| = ||x − y|| + ||y − z||
Proof. (a) , (b) and (c) are obvious, and (d) is an immediate consequence
of Cauchy-Schwarz inequality. By (d) we have

||x + y||2 = (x + y) · (x + y)
= x · x + 2x · y + y · y
≤ ||x||2 + 2||x|| ||y|| + ||y||2
= (||x|| + ||y||)2

from which (e) follows. Finally, (f ) follows from (e) on replacing x by x - y


and y by y- z.

Remarks. Theorem ?? (a) , (b) and (f) will make Rn a metric space,
which will be defined in unit 2. R1 (the set of all real numbers) is usually
called the line, or the real line. R2 is called the plane. -

1.1.4 Key words.


Integers, rational numbers, real numbers, upper bound, lower bound, least
upper bound , greatest lower bound, inner product , norm.

13
1.1.5 Terminal questions.
1. If r is rational (r 6= 0) and x is irrational, prove that r + x and rx are
rational.
2. Prove that there is no rational whose square is 3.
3. Prove that there is no rational whose square is 8.
4. Prove that there
√ is√no rational whose square is 12.
5. Prove that 2 + 3 is ir rational.
6. Prove that there exists an irrational number between any two real num-
bers.
7. If ab < dc ’ with b > 0, d > 0, prove that (a+ c)/(b + d) lies between a/b
and c/d. √ √
8. Prove that n + 1 + n − 1 is irrational for every integer n≥ 1
. 9. Prove that the supremum and infimum of a set are uniquely determined
whenever they exist .

10. Find the supremum and infimum (if exist) of the following sets of real
numbers:
(a) S = {2−m + 3−n |m, n ∈ Z+ }
(b) S = {x|3x2 − 10x + 3 < 0}
(c) S = {x|(x − a)(x − b)(x − c)(x − d) < 0, wherea < b < c < d}
(d) S = { m1 + n1 |m, n ∈ Z+ }
11. Prove the proposition (1.1.12).
12. Let A be a nonempty set of real numbers which is bounded below. Let
A = {−x|x ∈ A}. Prove that inf A= - sup(-A).
13. Suppose A and B are bounded sets of real numbers. Let A+ B =a+ b |
a ∈ A , b ∈ B. Prove that (a) sup( A + B) = sup A+ sup B (b) inf (A + B)
= inf A+ inf B.
14. Let A and B be sets of positive real numbers that are bounded above
and let AB = ab|a ∈ A, b ∈ B. Prove that sup(AB) = sup A sup B.
15. If x , y ∈ Rn , prove that

||x + y||2 + ||x − y||2 = 2||x||2 + 2||y||2

16. Suppose a , b ∈ Rn . Find c ∈Rn and r > 0 such that ||x − a|| = 2||x − b||
if and only if ||x − c|| = r.

14
References

1. Walter RudiI P.rinci p,les of Mathematical Analysis, McGraw-Hill In-


ternational book company, 3rd Edition.

2. Tom M. Apostal - Mathematical Analysis , Narosa Publishing House.

3. Terence Tao - Analysis I, Hindustan .Book Agency.

15
Unit II
Countable and uncountable sets

2.1.1 Objectives
In this unit we study about the classification of sets as countable and un-
countable set. At the end of this unit students will be able to check whether
a given set is countable or uncountable.

2.1.2 Introduction
In discussing any branch of mathematics it is helpful to use the notation and
terminology of set theory. This subject, which was developed by Boolean
and Cantor in the later part of the 19th century, has had a profound influ-
ence on the development of mathematics in 20th century. It has unified many
seemingly disconnected ideal and has helped to reduce many mathematical
concepts to their logical foundations in an elegant and systematic way.

2.1.3 Basic Notations of set Theory:


Definition 2.1.3.1 A set is a collection of well defined objects. We naturally
use capital letters to denote a set like X, Y, A,... etc. We use small letter x,
y, a,... etc to denote elements of the set.

Definition 2.1.3.2 Cartesian product of two set A and B is denoted by


A × B = {(a, b)/a ∈ A and b ∈ B}.

Definition 2.1.3.3 A relation 0 R0 from a set A and B is a non-empty subset


of Cartesian product A × B.

Definition 2.1.3.4 A function F is a set of ordered pairs (x, y), such that
no two of which have the same first number. That is if (x, y) ∈ F and
(x, z) ∈ F , then y = z.

16
Note: If S is a relation, the set of all elements x that occur as first mem-
ber of pairs (x, y) in S is called the domain of S. The set of second members
y is called the range of S, the set where the second member coming from is
called co-domain.

If S is a subset of domain of F , we say that F is defined on S. In this


case, the set of F (x) such that x ∈ S is called the image of S under F and is
denoted by F (S). If T is any set which contains F (s), then F is also called
a mapping from S to T . This is denoted by writing F : S −→ T .
Definition 2.1.3.5 A function F : S −→ T is said to be onto if F (S) = T
i.e. range of the function is equal to whole of co-domain.
Definition 2.1.3.6 Let f : A −→ B. If E ⊆ B, f −1 (E) denotes the set of
all x ∈ A such that f (x) ∈ E i.e. f −1 (E) = {x ∈ A/f (x) ∈ E}. We call
f −1 (E) the inverse image of E under f . Also it is called as pull back of E.
Definition 2.1.3.7 Let f : A −→ B be a function. If for each y ∈ B,
f −1 (y) = {x ∈ A/f (x) = y} contains at most one element of A, then f is
called one-one function. Alternatively one can define a mapping f : A −→ B
is 1 − 1 mapping if f (x1 ) 6= f (x2 ), whenever x1 6= x2 , for all x1 , x2 ∈ A.
Definition 2.1.3.8 If there exists a map f : A −→ B such that f is both
one-one and onto map, then we say that A and B have 1 − 1 correspondence
or f is bijective or A is equivalent to B.
Definition 2.1.3.9 A relation 0 ∼0 on X is said to be an equivalence relation.
If it satisfies the following three conditions
(i) Reflexive : a ∼ a, ∀a ∈ X
(ii) Symmetric : If a ∼ b then b ∼ a ∀a, b ∈ X
(iii) Transitive : If a ∼ b and b ∼ c then a ∼ c ∀a, b, c ∈ X.

2.1.4 Countable and uncountable sets


Definition 2.1.4.1 Let Jn = {1, 2, 3, ..., n}, where n ∈ N be the set of all
positive integers. For any set A, we say
a)A is said to be finite, if A is equivalent to Jn for some n ∈ N.
b) A is said to be infinite if it is not finite.
c) A is said to be countably infinite if A is equivalent to N.
d) A is said to be countable if it is either finite or countably infinite.
e) A is said to be uncountable if it is not countable.

17
Note:

In other words, we say a set A is countable either if A is finite or if there


is a bijective map from A to N.

Example 6 The set of all integers Z is countable set as the function f :


N −→ N defined as
(
n
2
if n is even
f (n) = −(n−1)
2
if n is odd

is a bijection.

Note: An infinite set is equivalent to its proper subset. Finite sets can’t
enjoy this property.

Now we will define sequence as every countable set can be written as a


sequence. Detailed study on sequence will be done in Block-II.

Definition 2.1.4.2 By a sequence, we mean a function f defined on the set


N of all positive integers. If we denote the image of f at n ∈ N by xn i.e
f (n) = xn , it is customary to denote the sequence f by the symbol {xn }, or
also by x1 , x2 , x3 , ... and the values of f , (i.e the elements xn ) are called the
terms of the sequences.

Remark: Let X be a countable set, then we know there exists a bijective


function f : N −→ X. Since f is one-one and onto, elements of X can be
expressed as a sequence {xn } of distinct terms and conversely if {xn } is a
sequence of distinct terms, then the range of the set {xn /n ∈ N} is countable.

Theorem 2.1.4.1 Every subset of a countable set A is countable.

Proof: Suppose E is finite then we are done, so assume that E is an


infinite subset of A. Since A is a countable set. The elements of A can be
arranged in a sequence {xn } of distinct elements. Construct a sequence {nk }
of natural numbers as follows, let n1 be the smallest positive integer such that

18
xn1 ∈ E. Inductively chosen n
1 , n2 , n3 , ..., nk−1 let
nk be the smallest positive
integer such that xnk ∈ E − xn1 , xn2 , ..., xnk−1 . Putting f (k) = xnk , we
obtain 1-1 correspondence between E and N . Hence E is a countable set.

Question: Let A ⊆ B, given that A is an uncountable set. what can you


say about the set B ?

Theorem 2.1.4.2 N × N is countable.

Proof: Define f : N × N −→ N by f (n, m) = 2n 3m , then it is easy to


see that f is an one-one function from fundamental theorem of arithmetic.
Hence N × N is equivalent to f (N). We know that subset of a countable set
is countable, hence f (N) is equivalent to N. Thus N × N is equivalent to N.
That is N × N is countable.

Note: In the above proof, the function f (n, m) = pn q m , would work


when p and q are distinct prime numbers.

Theorem 2.1.4.3 Countable collection of countable set is countable or


Let {A1 , A2 , ..., An , ...} be a countable collection of countable sets,
then S = ∪∞ n=1 An is countable.

Proof : If suppose the countable collection of countable sets {A1 , A2 , ..., An , ...}
is disjoint collection i.e. Aj ∩ Ak = φ for j 6= k, we are done. If not,
let us construct a countable collection of countable sets {B1 , B2 , ..., Bn , ...}

S ∞
S
such that An = Bn as follows. Let B1 = A1 , B2 = A2 − A1 and
n=1 n=1
Bn = An − (A1 ∪ A2 , ..., ∪An−1 ) for n ∈ N. Now {Bi }∞
i=1 is a disjoint col-
lection of countable sets as Bn ∩ Am = φ for m < n and since Bn ⊆ An for

S ∞
S
each n ∈ N we have Bn ⊆ An .
n=1 n=1

S
Let x ∈ An , then there exist m ∈ N such that m is the least pos-
n=1
itive integer such that x ∈ Am and x ∈
/ Ak for k < m, this implies that

S ∞
S ∞
S
x ∈ Bm ⊆ Bn and hence An = Bn .
n=1 n=1 n=1

S
Now we will prove that Bn is a countable set. Since subset of a countable
n=1

19
set is countable, each Bn 0 s are countable sets. Hence they can be arranged in

S
a sequence as {b1,n , b2,n , b3,n , ...} for n ∈ N. Let x ∈ Bn , then there exist
n=1
(k, p) ∈ N × N such that x = bk,p . this (k, p) is unique as {Bn }∞
n=1 is a dis-

S
joint collection. Thus the map f : Bn −→ N × N defined by f (k) = (k, p)
n=1

S
is well defined and one-one. Thus Bn is a countable set, being equivalent
n=1
to a subset of a countable set.

Theorem 2.1.4.4 Let A be a countable set, and let An = {(a1 , a2 , ..., ak , ..., an )/ak ∈ A, f or 1 ≤ k
be the set of all n − tuples for any n ∈ N. Then An is countable.

Proof: We will prove this using ‘Mathematical induction on n’. If k = 1,


0
A = A , hence it is countable, now assume that the result is true for k = n−1
i.e An−1 is a countable set. Now we will prove that An is countable. By the
definition of An , it can be written that An = {(b, a)/where b ∈ An−1 and a ∈ A}.
For every fixed b, the set {(b, x)/x ∈ A} is a countable set as it has one-one
correspondence with A. Then An =
S
{(b, x)/x ∈ A}, which is a count-
b∈An−1
able union of countable set. From previous result we conclude that An is a
countable set. Hence the proof.
Corollary 2.1.4.1 The set Q of all rational numbers is a countable set.

2
Proof: From previous
n theorem we haveo Z × Z = Z is a countable set.
Since the set Q = pq ; p, q ∈ Z and q 6= 0 . There is a one-one correspon-
dence from Q to a subset of Z × Z i.e the mapping pq 7→ (p, q). Hence Q is a
countable set.

Corollary 2.1.4.2 The set S of intervals with rational endpoints is a count-


able set.

Proof : Since Q is a countable set, elements of Q can be arranged in


sequence say {x1 , x2 , ..., xn , ...}. Let An be the set of all interval whose left
endpoint is xn and whose right endpoint is rational. Then clearly An is a

S
countable set and S = An . Hence S is countable.
n=1

20
Corollary 2.1.4.3 The set of all polynomials with rational coefficients is a
countable set.

Proof: Exercise.

Is there any set which is uncountable ? Following theorem answers this


question it actually help us to conclude that [0, 1] ⊆ R is an uncountable set.

Theorem 2.1.4.5 Let A be the set of all sequences whose elements are the
digits 0 and 1. This set A is uncountable.

Proof: We will prove this theorem by contradiction method i.e on the


contrary assume that A is a countable sets. Then elements of A can be written
as a sequence s1 , s2 , s3 , ... . We now construct a sequence s as follows.
(
1 if n is even
f (n) =
0 if n is odd
Then the sequence 'S 'is different from every sn in its nth place , but s ∈ A
and s 6= sk where k ∈ N. Which is a contradiction to our assumption that A
is a countable set. Hence A is uncountable.

21
2.1.5 Keywords.
Set theory, Cartesian product, Bijective function, equivalence relation, count-
able set and uncountable set.

2.1.6 Terminal problems:


1. Prove that 1 − 1 correspondence is a bijective relation on collection of
all sets.

2. Let f : S −→ T be a function. Prove the following for arbitrary subsets


X of S and Y of T .

(a) X ⊆ f −1 [f (x)].

(b) f [f −1 (Y )] ⊆ Y .

(c) f −1 [Y1 ∩ Y2 ] = f −1 [Y1 ] ∪ f −1 [Y2 ].

(d) f −1 [Y1 ∩ Y2 ] = f −1 [Y1 ] ∩ f −1 [Y2 ].

(e) f −1 (T − Y ) = S − f −1 (Y ).

3. Prove that the empty set is a subset of every set.

4. A complex number Z is s aid to be algebraic if there are integers


a0 , a1 , ..., an , not all zero, such that a0 z n + a1 z n−1 + ... + an−1 z + an .
Prove that the set of all algebraic number is countable.

5. Prove that [0, 1] is an uncountable. Hence prove R is uncountable.

6. Prove that there exist real numbers which are not algebraic.

7. Is the set of all irrational real number countable ? Justify.

22
References:

1. W. Rudin ''Principles of Mathematical analysis ''third edition. Me


Grow hill Education(India) Private limited, New Delhi.

2. S. Kumareson ''Topology of Metric spaces 'S'econd Edition. Narosa


Publishing House. New Delhi.

23
Unit III
Basic Topology

3.1.1 Objectives:
In this unit we study about one of the most fundamental concepts in analy-
sis i.e metric space, open and closed set in a metric space and their properties.

3.1.2 Introduction:
We all know the distance between two cities namely Mysuru and Bengaluru.
Is this the shortest distance? How do we measure the distance between two
point in a plane ? In the set of real number R, what is the distance between
two points? These experiences of finding distance between points gives rise
to a natural question “Can we define distance between any two elements in
a given set ?”The answer is yes, such a set is called METRIC SPACE, where
Metric means distance.

In our day to day life we use some terms such as neighbourhood and
boundary using the concept of distance, so next obvious question is “How
can we define such concepts in general metric space? ”We will study the
several properties of Metric space in this Chapter.

3.1.3 Metric space


Definition 3.1.3.1 Let X be a non-empty set. If there exists a function say
d : X × X −→ R, which satisfies the following conditions
(i) d(x, y) ≥ 0, ∀ x, y ∈ X.
(ii) d(x, y) = 0 iff x = y
(iii)d(x, y) = d(x, y) ∀ x, y ∈ X
(iv)d(x, y) ≤ d(x, z) + d(z, y) ∀ x, y, z ∈ X.
Then d is called a metric on X. (X, d) is called metric space.

Example 7 We know a distance function on R from our school days i.e


modulus function. Let d : R × R −→ R defined by

24
(
x−y if x-y ≥ 0
d(x, y) = |x − y| =
−(x − y) if x-y ≤ 0
now we will prove that d is a metric on R

Solution: (1) From the definition of d, it is clear that d(x, y) ≥ 0 ∀


x, y ∈ R and d(x, y) = 0 iff x = y.

2) Since |x − y| = |y − x|, we get d(x, y) = d(y, x).

3) Since we know that |x + y| ≤ |x| + |y| (triangular inequality), we have


d(x, y) = |x − y| = |(x − z) + (z − y)|
≤ |x − z| + |z − y|
≤ d(x − z) + d(z − y)
i.e d(x, y) ≤ d(x − z) + d(z − y) ∀ x, y, z ∈ R.
hence d(x, y) = |x − y| is a metric on R i.e (R, ||) is a metric space.

Note: On a non-empty set X, we can have more than one metric.

Example 8 Define d : R × R −→ R
(
1 if x 6= y
d(x, y) =
0 if x = y
Then d is a metric on R called discrete metric.

Solution: Clearly d(x, y) ≥ 0 ∀ and d(x, y) = 0 iff x = y Also d(x, y) =


d(y, x). Now we will prove the triangular inequality i.e
d(x, z) ≤ d(x, y) + d(y, z)∀ x,y,z ∈ R
Case i : If d(x, z) = 0 then d(x, y) + d(y, z) ≥ 0 by definition.
Hence d(x, z) ≤ d(x, y) + d(y, z).

Case ii : If d(x, z) = 1 then both d(x, y) and d(y, z) can’t be equal to zero
as if d(x, y) = 0 = d(y, z) implies x = y = z which is not true, hence one of
d(x, y) and d(y, z) is equal to one. Hence d(x, y) + d(y, z) is atleast 1, thus
d(x, z) ≤ d(x, y) + d(y, z) ∀ x, y, z ∈ R.

25
Definition 3.1.3.2 Let (X, d) be a metric space. Let p ∈ X and a neighbour-
hood of p is denoted by Nr (p) and it is defined as Nr (p) = {q ∈ X/d(p, q) < r}
for some r > 0. The number r is called the radius of Nr (p).

Example 9 R is a metric space with metric d(x, y) = |x − y|, for x ∈ R and


r > 0 we have Nr (x) = {y ∈ R/d(x, y) < r} = {y ∈ R/|x − y| < r}. Since x
is fixed we have Nr (x) = {y ∈ R/ − r + x < y < r + x} i.e
Nr (x) = (x − r, x + r)

Similarly identify the following set on the number line

a) N2 (0) = (−2, 2)

b) N1 (−1) = (−2, 0)

c) N100 (1) = (−99, 101)

d) N0.5 (0) = (−0.5, 0.5)

Example 10 Consider X = R2 with the standard Metric(euclidean Metric


), which is defined as follows
p
d((x1 , x2 ), (y1 , y2 )) = (y1 − x1 )2 + (y2 − x2 )2 .

26
Let (a, b) ∈ R2 , then for
n given r > 0,pr−neighbourhood of theopoint n (a, b) is p
defined as Nr ((a, b)) = (x, y) ∈ R / (x − a)2 + (y − b)2 < r = (x, y) ∈ R2 / (x − a)2 + (y − b
2

This is precisely all the points inside the circle with centre (a, b) and radius
r. Example

(a) N2 (1, 0) is all the point inside the circle with centre (1, 0) and radius
1 i.e.

(b) Can you find out N2 (1, 0) ∩ N1 (−2, 0) ?

Example 11 In R3 any neighbourhood Nr (a), where a ∈ R3 and r > 0 is


the set containing all the points inside the sphere at centre a and radius r.

Example 12 In R, if we take the discrete metric then the neighbourhood are


(
{a} if 0 < r ≤ 1
Nr (a) =
R if r > 1

Remark:

1) Now we will recall some well-known as well as important subset of R


and Rn . Let a, b ∈ R then

27
(a) (a, b) = {x ∈ R/a < x < b} (Open interval).

(b) [a, b) = {x ∈ R/a ≤ x < b} (Half -open and half-closed interval).

(c) (a, b] = {x ∈ R/a < x ≤ b} (Half -open and half-closed interval).

(d) [a, b] = {x ∈ R/a ≤ x ≤ b} (Closed interval).

2) Let a = (a1 , a2 , ..., ak )and b = (b1 , b2 , ..., bk ) be two element of Rk , then


k−cell is defined as [a, b] = (x1 , x2 , ..., xk ) ∈ Rk /ai ≤ xi ≤ bi f or i = 1, 2, ..., k .
1 − cell is an interval, a 2 − cell is a rectangle and 3 − cell is a cuboid, etc...

3) We call a set E ⊆ Rk convex if λx + (1 − λ)y ∈ E whenever x, y ∈ E


and 0 < λ < 1.

4) If x ∈ Rn and r > 0, then open ball B with centre at x and radius r is


defined to be the set of all Br (x) = {y ∈ Rn /|y − x| < r}.

5) If x ∈ Rn and r > 0, the closed ball with centre at x and radius r is


defined as Br [x] = {y ∈ Rn /|x − y| ≤ r}. Note that the closure of Br (x) is
Br [x].

Definition 3.1.3.3 (Limit point) Let (X, d) be a metric space, let E ⊂ X


and p ∈ X is said to be a limit point of E if ∀  > 0, N (p) − {p} ∩ E 6= φ.
We denote the set of all limit points of E in X by E 0 .

Example 1: Consider the standard metric on R and let E = (0, 1) ⊆ R,


then 0 is a limit point of E as any − neighbourhood of 0 in R is of the form
(−, ). By Archimedean property ∃ n ∈ N − {1} such that 0 < n1 <  which
implies (−, ) − {0} ∩ E 6= φ as n1 ∈ E.

The point 2 is not a limit point of E, to prove this, it is enough to give


some − neighbourhood of 2 which will not intersect any point of E ex-
cept 2. By observing this, if we take  = 12 then N 1 (2) is (1.5, 2.5) and
2
(1.5, 2.5) ∩ E = φ. Hence 2 is not a limit point of E. verify that E 0 = [0, 1].

28
Example 2: Let X = R be usual metric and A = N ⊆ R, then we will
prove that N0 = φ i.e N has no limit point in R with usual metric.

Consider x ∈ R, if x ∈ N then choose  = 12 such that N 1 x − 12 , x + 21



2
, doesn’t contains any natural number other than x, if x ∈ / N then choose
 = min {x − [x], [x] + 1 − x}, where [x] is the greatest integer less than or
equal to x. Then it is clear that N ∩ N = φ.

Example 3: Prove that, if A is any finite subset of R, then A0 = φ in


standard metric.

Example 4: Find E 0 for the following subsets of R under standard Met-


ric.
1
(a) E = n
/n ∈ N .

(b) E = [0, 1) ∪ [1, 2).

(c) E = Q.

Definition 3.1.3.4 (Isolated point:) Let (X, d) be a metric space, E ⊆ X,


a point p ∈ E is said to be an isolated point of E, if p is not a limit point of
E.

From the previous examples, [0, 1) has no isolated points in R under stan-
dard metric but every point in N is an isolated point.

Definition 3.1.3.5 (Closed set) Let (X, d) be a metric space, E ⊆ X then


E is said to be a closed set if it contains all its limit points i.e E 0 ⊆ E.

Example:

1) N ⊆ R is a closed set as N0 = φ ⊆ N.

2) E = [0, 1) ⊆ R is not a closed set as 1 ∈ E 0 but 1 ∈


/ E.

29
3) A = n1 /n ∈ N is not a closed subset of R under standard metric as


0 ∈ A0 but 0 ∈
/ A.

Definition 3.1.3.6 (Interior point) Let (X, d) be a metric space, let E ⊆ X


a point p ∈ E is said to be an interior point of E, if there exist some  > 0
for which the neighbourhood N (p) ⊆ E. We usually denote the set of all
interior points of E by E̊.

Example 1: E = (0, 1), then every point of E is an interior point as for


any x ∈ E, if we choose  = min {x, 1 − x} then (x − , x + ) ⊆ (0, 1). Hence
E̊ = E.

Example 2: E = [0, 1], then 0 and 1 are not interior points of E, hence
E̊ = E

Example 3: E = N then E̊ = φ.

All the above set are taken as a subsets of R under standard metric.

Definition 3.1.3.7 (Open set) Let (X, d) be a metric space, let E ⊆ X,


then E is said to be an open set if every points of E is an interior point of
it i.e E = E̊.

Example: E = (0, 1) is an open subset of R and F = [0, 1) is not an


open subset of R under standard metric.

Definition 3.1.3.8 Let (X, d) be a metric space E ⊆ X, the set E is said


to be bounded in X if ∃ M ∈ R+ and a point q ∈ X such that d(p, q) < M
for every p ∈ E.

Example : We all know bounded subset of real numbers under standard


metric. i.e let E ⊆ R, then E is said to be bounded subset of R if ∃ M ∈ R+
and choose q = 0, then |p| < M ∀ p ∈ E. It is clear that N is unbounded in
the Euclidean metric. R is itself a unbounded set in this metric.

Now we will consider the discrete metric on R, with this metric N, Q, R


are all bounded sets!

30
Definition 3.1.3.9 Let (X, d) be a metric space, E ⊂ X is said to be a
dense subset of X if every point of X is a limit point of E or a point of E if
X ⊆ E ∪ E 0.
Example: Q and Qc are dense subsets of R under standard metric.

Note: R has no dense proper subset under discrete metric.

Remark: It is important to observe that every subset of Y of a metric


space (X, d) is a metric space in its own right, with the same metric distance
function. Thus every subset of a Euclidean space is a metric space.

The following theorem gives us a large class of open sets.

Theorem 3.1.3.1 Let (X, d) be a metric space. Then every neighbourhood


is an open set.

Proof: Consider a neighbourhood E = Nr (p), for some r > 0 and p ∈ X.


Now we will prove that every point of E is an interior point. Let q ∈ E then
d(p, q) < r. Choose h = r − d(p, q) > 0. Now we will show that Nh (q) ⊆ E.
Let s ∈ Nh (q)
=⇒ d(q, s) < h i.e d(p, s) < r − d(p, q)
=⇒ d(p, q) + d(q, s) < r
=⇒ d(p, s) ≤ d(p, q) + d(q, s) < r
=⇒ s ∈ Nr (p) = E, as s is an arbitrary point we have Nh (q) ⊆ Nr (p) = E,
thus every point of E is an interior point. Hence every neighbourhood in any
metric space is an open set.

31
The following theorems helps us to decide whether a point is a limit point
of a given set or not.

Theorem 3.1.3.2 Let (X, d) be a metric space, p ∈ E ⊆ X. if p is a limit


point of E, then every neighbourhood of p contains infinitely many points of
E.

Proof: We will prove this by contradiction. Let p be a limit point of E, but


there is a neighbourhood Nr (p) which contains only finite number of points
of E. Let q1 , q2 , ..., qn be those points of N ∩ E which are distinct from p.

Choose k = min {d(p, q)/1 ≤ i ≤ n}, since p 6= qi ; for i = 1, 2, ..., n,


k > 0. Now consider the k− neighbourhood of p i.e Nk (p), clearly Nk (p) −
{p} ∩ E = φ which is a contradiction to the assumption that p is a limit point
of E.

Hence the proof.

Corollary 3.1.3.1 A finite set has no limit points in any metric space.

Proof: Exercise.

Given an open set, can you construct a closed set out of it? Similarly if
you get a closed set, can you construct an open set out of it? The following

32
theorem helps us to answer this question.

Theorem 3.1.3.3 Let (X, d) be a metric space and E ⊆ X. E is an open


set if and only if E c is closed set.

Proof: Assume E be an open set in (X, d). Now we will show that
E = {x ∈ X − E} is a closed set. Let x be a limit point of E c . Then
c

every neighbourhood of x i.e ∀  > 0, N (x) ∩ E c 6= φ, implies N (x) * E.


Hence x ∈ / E as x is not an interior point E. Thus x ∈ E c as x is x is an
arbitrary point E c contains all its limit point thus E c is closed.

Conversely, assume E c is a closed subset of X. Now we need to prove


that E is an open set. Let x ∈ E, then x is not a limit point of E c . Hence
there exists a neighbourhood for some  > 0 such that N (x) ∩ E c = φ. Which
implies N ⊆ E. Thus x is an interior point of E and hence E is open.

Corollary 3.1.3.2 Let (X, d) be a metric space and F ⊆ X. Then F is


closed subset of X if and only if F c is open set.

Proof: Exercise.

Lemma 3.1.3.1 Let {Eα }α∈I be a collection of sets Eα for each α ∈ I(I is
any finite or infinite set). Then
!c
[ \
Eα = (Eα c )
α∈I α∈I
c
and B = α∈I (Eα c ). To prove A = B, we
S T
Proof: Let A = α∈I Eα
will prove that A ⊆ B and B ⊆ A.
S
If x ∈ A implies x ∈/ Eα which implies x ∈
/ Eα for each α ∈ I. That is
α∈I
x ∈ Eαc for each α ∈ I i.e x ∈
T
Eα = B. Thus A ⊆ B. Hence the proof of
α∈I
the lemma.

conversely, let x ∈ B, then x ∈ Eαc for each α ∈ I, that is x ∈


/ Eα for
S
each α ∈ I, implies x ∈
/ Eα implies x ∈ A. Thus B ⊆ A.
α∈I

33
Next theorem gives us a way to construct new open and closed sets using
the given ones.

Theorem 3.1.3.4 Let (X, d) be a metric space Fα and Gα be subsets of X


for each α in an arbitrary set I. Then the following are true.
S
(a) If Gα is an open set for each α ∈ I, then Gα is open.
α∈I
T
(b) If Fα is a closed set for each α ∈ I, then Fα is a closed set.
α∈I
S
Proof: (a) let x ∈ Gα , then by the definition of union of set ∃ some
α∈I
α0 ∈ I such that x ∈ Gα0 . Since Gα0 is open ∃  > 0 such that N (x) ⊆ Gα0
S S
implies N (x) ⊆ Gα . Hence Gα is an open set.
α∈I α∈I

(b) From the above lemma we have


!c
\ [
Fα = (Fα c )
α∈I α∈I

Since each Fα is closed set, Fα c is anSopen set from part(a) we have


arbitrary union of open set is open, hence α∈I (Fα c ) is an open set. Thus
T c
α∈I Fα is a closed set.

Note: Above theorem could also be stated as 00 Arbitarary U nion of open sets is open00
and 00 Arbitaray intersection of closed set is closed 00 .

Question: Is the converse of the above theorem true?

Example 1: Is arbitrary intersection of open sets open?

Answer: Need not be true for example let I = N then for each n ∈ N
define
T Gn = −1 , 1 then {Gn }n∈I is an arbitrary collection of open sets but
n n
n∈I Gn = {0}(Prove this). Which is not an open set.

Example 2: Is arbitrary union of closed set closed ?

34
Answer: Need not be true, let I = N, for each n ∈ N define Fn =
2 + n1 , 4 − n1 then

[
Fn = (2, 4) [Justif y?]
n∈N
which is an open set.

Following theorem gives a condition under which the above two question
be true.

Theorem 3.1.3.5 Let (X, d) be a metric space, I be a finite set say I =


{1, 2, 3, ..., n} then the following are true
n
T
(i) G1 , G2 , ..., Gn be open subset of X, then Gi is open.
i=1
n
S
(ii) F1 , F2 , ..., Fn be closed subset of X, then Fi is closed.
i=1
n
T n
T
Proof: (i) We will prove that Gi is open. Let x ∈ Gi , we need to
i=1 i=1
n
T
find some  > 0 such that N (x) ⊆ Gi .
i=1
n
T
Since x ∈ Gi , we have x ∈ Gi for any i = 1, 2, ..., n, since Gi is an
i=1
open set, ∃ ri > 0 such that Nri (x) ⊆ Gi choose  = min {ri /i = 1, 2, ..., n}
then N (x) ⊆ Nri (x) for each i = 1, 2, ..., n, that is N (x) ⊆ Gi for each
n
T
i = 1, 2, ..., n. Thus N ⊆ Gi . Hence the proof.
i=1
(ii) From lemma 3.1.3.1 we have
n
!c n
[ \
Fi = (Fi c )
i=1 i=1
n
since each Fi is closed Fi c is open, from(i) we have (Fi c ) is an open set
T
 n c i=1
S Sn
thus Fi is an open set.That is Fi is a closed set. Hence the proof.
i=1 i=1

Definition 3.1.3.10 (Closure of a set) Let (X, d) be a metric space, if E ⊂


X and E 0 be the set of all limit points of E in X, then the closure of E is
the set Ē = E ∪ E 0 .

35
The following theorem helps us to construct a closed set out of any set
and also to verify whether a given set is closed or not.

Theorem 3.1.3.6 If X is a metric space and E ⊂ X, then the following


holds.
(a) Ē is a closed set.
(b) E = Ē if and only if E is closed.
(c) Ē ⊆ F for every closed set F ⊂ X such that E ⊂ F .
c
Proof: (1) To prove Ē is closed, we will prove that Ē is an open set.
c
let x ∈ Ē = (E ∪ E 0 )c = E c ∩ E 0c
i.e x ∈/ E and x ∈/ E 0 . Which implies ∃ δ > 0 such that Nδ (x) ∩ E = φ and
Nδ (x) ∩ E 0 = φ. i.e Nδ (x) ⊆ E c ∩ (E 0 )c = (Ē)c . Thus Ē is a closed set.

(2) If suppose E is a closed set, then by the definition we have E 0 ⊆ E,


thus Ē ⊆ E and E ⊆ Ē as Ē = E ∪ E 0 . Thus we have E = Ē.

(3) let F be a closed subset of X containing E, then we need to prove that


Ē ⊆ F . Given that E ⊆ F implies that any limit point of set E is also a limit
point of F i.e E 0 ⊆ F 0 . Since F is a closed set we have Ē = E ∪ E 0 ⊆ F .
Hence the proof.

Note: Last statement in the above theorem implies that Ē is the smallest
closed set containing E.

Following Remark help us to clearly understand the difference between


the neighbourhood of a point x ∈ Y ⊂ X with respect to the Metric in Y as
well as in X.

Remark: Let (X, d) be metric space and Y ⊆ X, then we know that


(Y, d) is a metric space. Suppose E ⊆ Y ⊆ X. To say that E is an open
subset of X means for each p ∈ E there is associated a positive number
r such that Nr (p) = {q ∈ X/d(p, q) < r} ⊆ E. To say E is open relative
to Y if for each p ∈ E there is associated a positive number r such that
Nr (p) = {q ∈ Y /d(p, q) < r}.

36
The following theorem helps us to construct an open set in Y by using a
given open set in X.

Theorem 3.1.3.7 Suppose (X, d) be a metric space, Y ⊆ X. A subset E of


Y is open relative to Y if and only if E = Y ∩G for some open subset G of X.

Proof: Assume that E is an open set related to Y i.e for each p ∈ E ∃


rp > 0 such that Nrp (p) = {q ∈ Y /d(p, q) < rp } the neighbourhood of p in Y
is contained in E. Now consider the rp − neighbourhood of p in X i.e say
Vp = {q ∈ X/d(p, q) < rp } = Vp ∩ Y . Then we know that every neighbour-
hood is an open set. Hence Vp is an open set in X. Since arbitrary union of
S S
open sets is open, Vp is an open set in X. Say G = Vp .
p∈E p∈E

Now we will prove that E = G ∩ Y . Since for each p ∈ E, p also belongs


to vp ⊆ G, we have E ⊆ G. Hence E ⊆ G ∩ Y . −→(1)
S
Let q ∈ G ∩ Y implies q ∈ Y and q ∈ G = Vp implies q ∈ Vp for some
p∈E
p ∈ E and q ∈ Y . Since Nrp (p) = Vp ∩ Y, we obtain q ∈ Nrp (p) ⊆ E, thus
we have G ∩ Y ⊆ E −→(2)

From (1) and (2) we have E = G ∩ Y .

Conversely suppose G is an open set in X then we need to prove that


G ∩ Y is open in Y . Let x ∈ G ∩ Y , since x ∈ G and G is open in X, there
exist rx > 0 such that Vrx (x) = {q ∈ X/d(x, q) < rx } ⊆ G.
Let Nrx (x) = {q ∈ Y /d(x, q) < rx } = Vrx (x) ∩ Y , then clearly Nrx (x) is a
neighbourhood of x in Y and x ∈ Nrx (x) ⊆ G ∩ Y. Thus G ∩ Y is an open
set in Y whenever G is an open set in X.

Note: Le (X, d) be a metric space and E ⊆ Y ⊆ X. If suppose E is


open in Y can we conclude E is open in X ? The answer to this question is
''need not be true ''. For example let (R, d) be an Euclidean metric on R,

Y = [0, ∞) ⊆ R consider [0, 1) ⊆ Y , then [0, 1) is open in Y but it is not


open in R.

37
Remark: If suppose Y ⊆ X, where (X, d) be a metric space. If Y is an
open set in X then every open set in Y is also open in X.

3.1.4 Keywords.
Metric space, discrete metric, neighbourhood of a point, interior point, Open
set, Limit point, Closure of a set, Closed set.

3.1.5 Terminal problems:


1. Prove that the following functions are metric on R2
p
(a) d : R2 ×R2 −→ R defined by d((x1 , y1 ), (x2 , y2 )) = (x2 − x1 )2 + (y2 − y1 )2
(b) d1 : R2 × R2 −→ R defined by d1 ((x1 , y1 ), (x2 , y2 )) = |x2 − x1 | +
|y2 − y1 |
(c) d∞ : R2 ×R2 −→ R defined by d∞ ((x1 , y1 ), (x2 , y2 )) = M ax {|x2 − x1 | + |y2 − y1 |}

2. All the above metric defined on R2 can be extended to Rn , where n ∈ N


i.e

(a) d : Rn × Rn −→ R defined by

n
! 21
X
d((x1 , x2 , ...xn ), (y1 , y2 , ..., yn )) = (yi − xi )2 .
i=1

(b) d1 : Rn × Rn −→ R defined by
n
X
d1 ((x1 , x2 , ...xn ), (y1 , y2 , ..., yn )) = |yi − xi |
i=1

(c) d∞ : Rn × Rn −→ R defined by

d1 ((x1 , x2 , ...xn ), (y1 , y2 , ..., yn )) = M ax {|yi − xi | /1 ≤ i ≤ n}

3. Let d, d1 and d∞ are as defined in problem 1, be a metric on R2 . Then


with respect to each Metric find Nr ((0, b)). In particular find N1 ((0, 0))
in (R2 , d), (R2 , d1 ) and (R2 , d∞ )

38
4. Given the following subset of R2 . Under Euclidean metric determine
whether it is closed, bounded, open and find closure of each of them.

(a) A non-empty finite set.


(b) Z
(c) The segment (a, b) in R2
(d) E = n1 /n ∈ N

n p o
2 2 2
(e) A = (x, y) ∈ R / x + y ≤ 1
n p o
2 2 2
(f) B = (x, y) ∈ R / x + y < 1

|x−y|
5. For x, y ∈ R, define d1 : R × R −→ R as d1 (x, y) = 1+|x−y|
. Show that
d1 is a metric on R.

6. Show that the functions d2 , d3 : R × R −→ R defined as d2 (x, y) =


|x2 − y 2 | and d3 = |x − 2y| are not a metric. (Note this is an extension
of Problem 4)

d(x,y)
7. Let (X, d) be a metric space, define δ : X ×X −→ R as δ(x, y) = 1+d(x,y)
show that δ is also a metric on X. Note that X is itself a bounded set
with the Metric δ. Hence it is called a bounded Metric.

8. Let d be a metric on X. Define δ(x, y) = min {1, d(x, y)} for all x, y ∈ X
show that δ is a metric on X.

9. Show that any open interval (a, b) in R is an open set. Is R an open


set in R under standard metric.

10. If in a metric space we have Nr (x) = Ns (y), does it mean that x = y


and r = s ? Justify.

39
11. Find all open sets in a discrete metric space.

12. What are all the open sets in a finite metric space ?

13. Is Q open set in (R, | |).

14. Let E̊ denote the set of all interior points of a set E.


(a) Prove that E̊ is always open.
(b) Prove that E̊ is open if and only if E̊ = E.
(c) If G ⊆ E and G is open, prove that G ⊆ E̊.
(d) Prove that the complement of E̊ is the closure of E c .
(e) Do E and Ē always have the same interior.
(f) Do E and E̊ always have the same closure ?
15. Consider the standard metric on R. For each of the following subspace
Y ⊇ E. Determine whether E is open(closed) in Y or E is open(closed)
in R both or neither.
(a) Y = [0, 1], E = [0, 1).
(b) Y = [−2, 1), E = (0, 1).
(c) Y = n1 : n ∈ N , E = 12 .
 

(d) Y = Z, E = N.
(e) Y = Q, E = (0, 1) − Q.
(f) Y = n1 : n ∈ N ∪ {0}, E = {0}.


References:

1. W. Rudin ''Principles of Mathematical analysis ''third edition. Me


Grow hill Education(India) Private limited, New Delhi.

2. S. Kumareson ''Topology of Metric spaces 'S'econd Edition. Narosa


Publishing House. New Delhi.

40
Unit IV
Compact, Connected and Perfect sets

4.1.1 Objectives:
In this unit we will be studying about the compact, connected and perfect
sets of Rk under standard metric.

4.1.2 Introduction:
The notion of compactness and connectedness are of great importance in
analysis, especially in connection with continuity. We have observed earlier
in the previous unit that if E ⊆ Y ⊆ X, then E may be open relative to Y
without being open relative to X. The property of being open thus depends
on the space in which E is embedded. The same is true of the property of
being closed.

Compactness, however, behaves better, in the sense that if a set is com-


pact relative to a subspace then it is also compact relative to the whole space
and the converse also holds , which we will be proving in the beginning of
this unit.

Connected spaces also plays an equivalently important role in analysis.


This concepts helps to differentiate between two important sets R and R2
analytically(not homeomorphic). We will see that if we remove one point x
from R then R−{x} is not connected but if we remove one point x ∈ R2 from
R2 (infact, any finite number of points from R2 ) it will be still connected set.

Also we study compact, connected and perfect sets in Rk with Euclidean


metric, as well as some general results. We also remark that, when some
property can be generalized to any metric space.

4.1.3 Compact sets


Definition 4.1.3.1 (Open cover) Let (X, d) be a metric space, E ⊆ X and
{Vα }α∈I be a collection of open sets in X. Then we say that the collection

41
S
{Vα }α∈I is said to be an open cover of E if E ⊆ Vα .
α∈I

Example 13 Let us consider the set of all real numbers R with the Euclidean
metric.
a) Let Vn = (−n, n), then the collection {Vn }n∈N is clearly an open cover
for the whole set R and hence it is an open cover for every subset of R.
b) Let Vn = −1, n1 , then the collection {Vn }n∈N is an open cover for (0, 1)
S
as each Vn is open and (0, 1) ⊆ Vn .
n∈N
c) Let Vn = (n − 1, n + 1), then the collection {Vn }n∈Z is an open cover for R

Definition 4.1.3.2 Let (X, d) be a metric space and K ⊆ X. K is said to


be compact if every open cover of K contains a finite subcover.

Example 14 Every finite subset of any metric space X is compact.

We have to wait a little longer to get a non-trivial example of a compact


set as we need to get familiar with some properties of such sets to check for
given any open cover it should have a finite subcover.

Now we will see that compact sets can be regarded as metric spaces in
their own right, without paying any attention to any embedding space.

Theorem 4.1.3.1 Suppose K ⊆ Y ⊆ X. Then K is compact relative to X


if and only if K is compact relative to Y .

Proof: First we will prove that if K is compact relative to X, then K


is compact relative to Y . Let {Vα }α∈I be a collection of sets, open relative
S
to Y . such that K ⊆ Vα i.e {Vα }α∈I is an open cover of K relative
α∈I
to Y . Since Vα is open in Y , there exist a set Gα open in X such that
Vα = Y ∩ Gα for all α ∈ I. Which implies {Gα }α∈I is an open cover of K
relative to X. Since K is compact relative to X, there exist finitely many
indices α1 , α2 , ..., αn ∈ I such that K ⊆ Gα1 ∪ Gα2 ∪, ..., ∪Gαn . Since K ⊆ Y ,
implies K ⊆ Vα1 ∪ Vα2 ∪, ..., ∪Vαn . This proves that K is compact relative to
Y.

42
Conversely, suppose K is compact relative to Y . Let {Gα }α∈I be a col-
lection of open subsets of X which covers K. Consider Vα = Y ∩ Gα , for
each α ∈ I. Then {Vα }α∈I is an open cover of K relative to Y . Since
K is compact relative to Y , there exist finite indices say α1 , α2 , ..., αn such
that K ⊆ Vα1 ∪ Vα2 ∪, ..., ∪Vαn . Since Vαi ⊆ Gαi for 1 ≤ i ≤ n, we have
K ⊆ Gα1 ∪ Gα1 ∪, ..., ∪Gαn . Hence K is compact relative to X.

The following theorem gives us the properties of a compact set in any


metric space.

Theorem 4.1.3.2 Compact subsets of metric space are closed.

Proof: Let (X, d) be metric space and K ⊆ X be a compact set. Now we


will prove that K is a closed set by proving K c is an open set.
Let x ∈ K c be fixed now for each p ∈ K choose Up = N r2 (p) and Vp = N r2 (x)
where r = d(x, p) > 0. The collection {Up /p ∈ K} is an open cover for K.
Since K is compact there exist finite points in K say p1 , p2 , ..., pn such that
K ⊆ Up1 ∪ Up2 ∪, ..., ∪Upn = U . Now consider V = Vp1 ∩ Vp1 ∩, ..., ∩Vpn a
neighbourhood of x, Since Vpi ∩ Upi = φ implies U ∩ V = φ. Hence V ⊆ K c .
Thus x is an interior point of K c . Since x is an arbitrary point we have K c
is an open set. Hence the proof.

Theorem 4.1.3.3 Closed subsets of compact sets are compact.

Proof: Let (X, d) be a metric space and F ⊆ K ⊆ X, F be a closed


subset of a compact set K in X. Then we need to prove that F is a compact
set. Let {Vα }α∈I be an open cover of F . Now consider the open cover Ω =
{Vα }α∈I ∪ {F c } of K. Since K is compact set there exist finite sub collection
Φ of Ω which covers K and hence covers F . If F c is a member of Φ, we may
remove it from Φ and Φ − {F c } is a finite cover of F . We have thus shown
that a finite sub collection of {Vα } covers F .

Corollary 4.1.3.1 If F is closed and K is compact subset of a metric space


X the F ∩ K, is compact.

Proof: Since K is compact subset of a metric space, K is closed. Since


the intersection of closed set is closed K ∩ F is a closed subset of a compact
set K. Hence from the above theorem we have K ∩ F is compact set.

43
Theorem 4.1.3.4 Let {Kα }α∈I be a collection of compact subsets of a met-
ric space (X, d) such that the intersection of every finite sub collection of
T
{Kα }α∈I is non-empty, then Kα is non-empty. (We call such a collection
α∈I
{Kα }α∈I satisfies finite intersection property )
T
Proof: We will prove this by contradiction method. Suppose Kα = ∅,
α∈I
then if we fix one set K1 of {Kα }α∈I , we can say that no point of K1 belongs
to every Kα , α ∈ I i.e for each x ∈ K1 there exist some Kα0 in {Kα }α∈I
such that x ∈/ Kαc 0 . The collection {Kαc }α∈I is an open cover of K1 and it is
a compact set, hence there exist finitely many indices α1 , α2 , ..., αn such that

K1 ⊆ Kαc 1 ∪ Kαc 2 ∪ Kαc 3 , · · · ∪ Kαc n

This means that

K 1 ∩ K α1 ∩ K α2 ∩ K α3 , · · · ∩ K αn = ∅

which is a contradiction to our assumption that the intersection of every


T
finite sub-collection of {Kα }α∈I is non-empty. Hence Kα 6= ∅. Which
α∈I
completes the proof.

Corollary 4.1.3.2 If {Kn } is a sequence of non-empty compact sets such



T
that Kn+1 ⊆ Kn for n = 1, 2, 3, . . . then Kn is not empty.
n=1
Proof: To apply the above theorem we need to show that given any finite
collections of natural numbers say n1 , n2 , . . . nr the intersection

Kn1 ∩ Kn2 ∩ Kn3 , · · · ∩ Knr = ∅

Since we have Kn ⊇ Kn+1 for each n ∈ N, then intersection Kn1 ∩ Kn2 ∩


Kn3 , · · ·∩Knr = Kl where l = min {n1 , n2 , . . . nr }, which is not empty. Hence
we can apply the previous theorem to obtain the required result.

Theorem 4.1.3.5 If E is an infinite subset of a compact set K, then E has


a limit point in K.

Proof: We will prove this by contradiction method. if suppose E has no


limit point in K, then for every point x ∈ K there exists a − neighbour-
hood N (x) such that N (x) − {x} ∩ E = ∅. The collection of neighbourhood

44
{N (x)}x∈K is an open cover of K. Since K is compact there exist finitely
many points x1 , x2 , . . . xn ∈ K such that

K = N1 (x1 ) ∪ N2 (x2 )∪, · · · ∪ Nn (xn )

as E ⊆ K and each Ni (xi ) contains at most one point of E implies that E
is a finite set which is a contradiction. Hence the proof.

Following two lemma helps us to obtain a non-trivial examples of compact


subset of Rn with Euclidean metric.

Lemma 4.1.3.1 If {In } be sequence of intervals i.e In = [an , bn ] in R such



T
that In ⊇ In+1 , for each n ∈ N, then In is not empty.
n=1

Proof: Consider the set E = {an /n ∈ N and In = [an , bn ]}. Then E is


non-empty set which is bounded above by b1 . It has a supremum say x =
sup E. Now we will show that x ∈ In for each n ∈ N. Since for any m and
n in N, we have
an ≤ am+n ≤ bm+n ≤ bm
i.e an ≤ bm ∀ n, m ∈ N . Which implies bm is an upper bound for the set E.
By the definition of supremum, we obtain x ≤ bm for each m ∈ N. Implies
am ≤ x ≤ bm ∀ m ∈ N. =⇒ x ∈ Im ∀ m ∈ N.

T
Hence x ∈ Im 6= ∅.
m=1

Now we will see that above lemma can be generalized to any such k−cells
in Rk .

T∞ if {In } is a sequence of K−cells


Lemma 4.1.3.2 Let K be a positive integer,
such that In ⊇ In+1 for each n ∈ N then n=1 In is non-empty.

Proof: Let In consist of all points x = (x1 , x2 , . . . xk ) such that an,j ≤


xj ≤ bn,j where 1 ≤ j ≤ k and n ∈ N. consider the interval In,j = [an,j , bn,j ].
For each fixed j, the sequence {In,j } of intervals in R satisfies the condition
T∞
In,j ⊇ In+1,j for each n ∈ N, then from Lemma 4.1.3.1 we have In,j 6= ∅.,
n=1

45

let xj ∗ ∈ In,j for each 1 ≤ j ≤ n then an,j ≤ xj ∗ ≤ bn,j for 1 ≤ j ≤ k and
T
n=1
n ∈ N. Now consider
x∗ = (x∗1 , x∗2 , . . . x∗k )

then it is clear from the definition of K−cell that x∗ ∈
T
In . Hence the
n=1
proof of the lemma 4.1.3.2.

Theorem 4.1.3.6 Every k−cell is compact.

Proof:On the contrary, suppose a k−cell I is not compact, then there ex-
ists an open cover {Gα } of I which doesn’t contains a finite sub cover of I.
By the definition of k−cell , I consisting of all points x = (x1 , x2 , . . . xk ) such
( ) 21
k
(bj − aj )2 , then for any
P
that aj ≤ xj ≤ bj for 1 ≤ j ≤ k. Put δ =
j=1
two points x, y in I, we have |x − y| ≤ δ.

Now we will subdivide I by dividing each interval in to two equal halfs.


aj + b j
Put cj = , for each 1 ≤ j ≤ k, then intervals [aj , cj ] and [cj , bj ] then
2
k
determine 2 k−cells Qi whose union is I. At least one of these sets Qi ,
1 ≤ i ≤ 2k say I1 cannot be covered by any finite sub collection of {Gα }, note
δ
for any x, y ∈ I1 , |x − y| ≤ . We next subdivide I1 and obtain I2 such that
2
I1 ⊇ I2 and the collection {Gα } has no finite sub collection which cover I2
and for any x, y ∈ I2 , we have |x − y| ≤ 2δ2 . By continuing this process, we
obtain a sequence {In } with the following properties.
a) I ⊇ I1 ⊇ I3 ⊇ . . .
b) In is not covered by any finite subcollection of {Gα }. c) For each x, y ∈ In ,
we have |x − y| ≤ 2δn .
∞ ∞
In 6= ∅. Let x∗ ∈
T T
Hence by Lemma 4.1.3.2 we have In , then there
n=1 n=1
exist some Gα such that x∗ ∈ Gα . Since Gα is open, there exist r > 0 such
that Nr (x∗ ) ⊆ Gα .
δ
By archimedean property there exist n ∈ N such that n < r, which
2
implies In ⊆ Nr (x∗ ) ⊆ Gα which is a contradiction to (b) as In is covered by
only one element of {Gα }. Hence every k−cell is compact.

46
Using the above theorem we obtain numerous examples of compact sets.
Example 15 [0, 1] is a compact set in R.

Example 16 [0, 1] × [0, 1] is a compact set in R2 .

The following theorem answers the question. ''Given a subset of Rn , can


we check whether it is a compact set or not? ''

Theorem 4.1.3.7 Let E be a subset of Rn , then the following are equivalent


a) E is closed and bounded.
b) E is compact.
c) Every infinite subset of E has a limit point in E.

Proof: O a =⇒ O b we assume that E is a closed and bounded subset of Rn


then we have to prove that E is compact. Since E is bounded there exist a
k−cell I such that E ⊆ I from the previous theorem, we have I is a compact
set since closed subset of a compact set is compact, E is compact.

Ob =⇒ O
c Theorem 4.1.3.5

Oc =⇒ O a We assume that every infinite subset of E has a limit point


in E then we will prove that E is closed and bounded.

On the contrary, suppose E is not bounded then for each n ∈ N, there


exist xn in E such that |xn | > n, the set A = {xn /n ∈ N} is clearly an infinite
subset of E. which doesn’t have a limit point, which is a contradiction to
our assumption that every infinite subset of E has a limit point. Hence E is
a bounded set.

If suppose E is not closed, then there is a point x0 ∈ Rk such that x0 is


a limit point of E but x0 6= E. Now we will construct a sequence {xn } in E
as follows for each n ∈ N choose xn ∈ N 1 (x0 ) ∩ E 6= ∅ =⇒ |xn − x0 | < n1 .
n
Since x0 6= E, the set S = {xn /n ∈ N} is an infinite subset of E. By our
assumption S has a limit point in E say y.

47
If y 6= x0 then |x0 − y| ≤ |x0 − xn | + |xn − y|
|xn − y| ≥ |x0 − y| − |xn − x0 |
|xn − y| ≥ |x0 − y| − n1 ∀ n ∈ N

Choose N1 ∈ N such that m1 ≤ |x02−y| ∀ m ≥ N1 from the above two


inequality we obtain |xn − y| ≥ |x02−y| ∀ n ≥ N1 which is a contradiction to
the fact that y is a limit point of S. Hence, we must have x0 = y. Thus E is
a closed set. Hence the proof.

Theorem 4.1.3.8 (Heine-Borel Theorem) A set E in Rn is compact if and


only if it is closed and bounded.

Proof: Same as the first two parts of Theorem 4.1.3.7

Corollary 4.1.3.3 Every bounded infinite subset of Rk has a limit point in


Rk .

Proof: Exercise.

4.1.4 Connected sets:


Definition 4.1.4.1 Two subsets A and B of a metric space X are said to
be separated if both A ∩ B and A ∩ B are empty.

Definition 4.1.4.2 A set E ⊆ X is said to be connected if E is not a union


of two non-empty separated sets.

Example 17 A = [0, 1) and B = [1, 2) are two subsets of R which are


disjoint sets but A and B not separated as A ∩ B = {1} =
6 ∅, even though
A ∩ B = ∅.

Example 18 let E = n1 /n ∈ N is not a connected set as E canbe written




as separation of two non-empty set E = A ∪ B, where A = 1, 12 and


B = E − A clearly A ∩ B = ∅ and A ∩ B = ∅

Example 19 Any finite set which contains atleast two elements in Rn is not
connected.

48
Note: Separated sets are disjoint, but disjoint sets need not be separated.
Following theorem helps us to classify all connected subsets of R.
Theorem 4.1.4.1 A subset E of the real line R is connected if and only if
it has the following property: If x ∈ E, y ∈ E and x < z < y then z ∈ E.

Proof: let E be a connected subset of E. On the contrary assume x ∈ E


and y ∈ E, but ∃ z in R such that x < z < y but z 6= E. The define
Az = E ∩ (−∞, z) and Bz = E ∩ (z, ∞). Since x ∈ Az and y ∈ Bz , they are
non-empty set. Since (−∞, z) and (z, ∞) are separated set Az and Bz are
separated, which is a contradiction to our assumption that E is a connected
set. Hence E satisfies the later property.

Conversly assume that E is a set in R with the property whenever x, y ∈ E


with x < y then every element z ∈ (x, y) [i.e x < y < z] is in E. Then we
need to prove that E is connected. If suppose E is not connected. Then there
exists non-empty separated sets say A and B such that A ∪ B = E. Pick
x ∈ A and y ∈ B.Without lose of generality assume that x < y. Define
z = sup {A ∩ [x, y]}. Hence z ∈ A, but z ∈ / B. In particular, x ≤ z < y. If
z∈/ A, it follows that x < z < y and z ∈/ E. If z ∈
/ A then z ∈/ B hence there
exists z1 such that z < z1 < y and z1 ∈ / B. Then x < z1 < y and z1 ∈ / E,
which is a contradiction. Hence E has to be a connected set.

4.1.5 Perfect sets:


Definition 4.1.5.1 Let (X, d) be a metric space, E ⊆ X be subset of X. E
is said to be a perfect set if it is closed and if every point of E is a limit point
of E.
Example 20 E = [0, 1] is a perfect set in (R, | |).
Example 21 E = {0} ∪ [2, 3] is not a perfect set even though it is a closed
set as 0 is not a limit point of it.
Example 22 Every closed and connected intervals in R is a perfect set.
Note: Above examples of perfect set are all intervals in R this observa-
tion gives rise to a natural question that '' Is there exists a perfect set in R
which contains no segment(interval)? '' Next we will see a wonderful con-
struction of a set called cantor set which will help us to answer this question!

49
4.1.6 Cantor set:
Let E 0 be the interval [0, 1]. By removing  1the Smiddle
 2  one third segment
1 2
,
3 3
from E0 , we obtain set say E1 =  0, 3 3
, 1 . Again remove the
1 1
middle thirds as the intervals
 S  2 30, 3 
and S
0,3 and  we call the resultant set
as E2 , where E2 = 0, 19 6 7 8
S
,
9 9
,
9 9 9
, 1 .

Continuing this process we obtain a sequence of compact sets En , such


that
a) E1 ⊇ E2 ⊇ E3 ⊇, . . .
b) En is the union of 2n intervals, each of length 3−n .

T
Now consider the set P = En , P is called the cantor set.
n=1

Note:

1) Cantor set is a compact set being closed and bounded subset of R.

2) Cantor set is not-empty as each En 0 s are compact and En ⊇ En+1 in


R.

Theorem 4.1.6.1 Cantor set is a perfect in R which contains no segment(interval).

Proof: let P be the cantor set. By the definition of cantor set, P doesn’t
contains any interval of the form 3k+1 3k+2

3m
, 3m
, where k and m are positive
integer. Now to prove P doesn’t contains any segment, we  will prove that any
interval (α, β) ⊆ [0, 1] ∃ k, m ∈ N such that 3k+13m
, 3k+2
3m
⊆ (α, β), which im-
plies that (α, β) * P . Let ∅ =
6 (α, β) ⊆ [0, 1] be any interval, then β − α > 0,
by archimedean principle there exist m ∈ N such that 3−m (β − α) > 6, i.e the
distance between 3−m β and 3−m α is atleast 6, which implies that there are
atleast 5 consecutive integers between 3−m β and 3−m α. Hence we can find
k ∈ N such that 3−m α < 3k + 1, 3k + 2 < 3−m β .

3k+1 3k+2
=⇒ α < ,
3−m 3−m
< β.

3k+1 3k+2

=⇒ 3m
, 3m ⊆ (α, β)

50
Hence the Cantor set can’t contains any segment. Now we will prove that
the Cantor set is Perfect, we know that P is closed. Hence to prove P is
Perfect set it is enough to prove that every point of P is a limit point of
it. Let x ∈ P, let S = N (x) be any −neighbourhood of x in R. Since
T∞ 2n
S
x∈P = En where En = Ik , x ∈ Ik for some k. note that length of
n=1 k=1
each Ik is 31n for 1 ≤ k ≤ 2n . choose n so large that In ⊆ S. Let xn be an
endpoint of In , such that xn 6= x, implies xn ∈ S ∩ P − {x} =
6 ∅. Hence x is
a limit point of P .

Following theorem gives us a criterion for generating uncountable subset


of Rk .

Theorem 4.1.6.2 Let P be a non-empty perfect set in Rk then P is un-


countable.

Proof: Since P is a perfect set in Rk it has a limit point. Hence P has to


be an infinite set. We will prove that P is an uncountable set using contra-
diction method. If suppose P is a countable set, then elements of P can be
arranged in a sequences as x1 , x2 , . . . we will construct a sequence of neigh-
bourhoods {Vn } as follows. Let V1 = Nr (x1 ) be any neighbourhood of x1 . then
V1 ∩ P − {x1 } 6= ∅ say x2 ∈ V1 ∩ P − {x1 }, since x2 is also a limit point of
P we can choose r' such that V2 = Nr0 (x2 ) ⊆ V1 and x1 ∈ / V2 .

Clearly V2 ∩ P − {x2 } =
6 ∅. By continuing this process we obtained a
sequence of neighbourhood {Vn } , which satisfies the following properties:
i) Vn+1 ⊆ Vn
ii) xn ∈
/ Vn+1
iii) Vn+1 ∩ P 6= ∅

Put Kn = Vn ∩ P . Since Vn is a closed and bounded subset of Rk , it is


compact set. Since Kn is a closed subset of a compact set Kn is also compact

T
set. Since xn ∈
/ Kn+1 no point of P lies in Kn , but Kn ⊆ P implies that
n=1

T
Kn = ∅ where as Kn 6= ∅ for each n ∈ N and Kn ⊇ Kn+1 ∀ n ∈ N which
n=1
is a contradiction. Hence P is an uncountable set.

51
Corollary 4.1.6.1 Every interval [a, b] where (a < b)is an uncountable set
in R. In particular, the set of all real numbers is uncountable.

4.1.7 Keywords.
Open cover, Compact set, Connected set, Perfect set, Cantor set.

4.1.8 Terminal problems:


1. Which of the following subsets of R are compact in the usual metric of
R?

(a) The set of natural numbers


(b) 0, 21 , 14 , 16 , 81 , . . .


(c) 0, 12 , 13 , 14 , . . .


(d) {1, 2, 3, 4, 5, . . .}
(e) (−∞, 2] ∪ (0, 1]
(f) [1, 3] ∪ [2, 4]

2. If X1 and X2 are compact subsets of metric space, prove that the


following

(a) X1 ∪ X2 is compact.
(b) X1 ∩ X2 is compact.

3. Find all the compact subsets of X with the discrete metric.

4. Is the set (0, 1) ∪ 0, 21 , 31 , 14 , . . . compact ?




5. Determine whether the following sets are separated with the usual met-
ric

(a) 0, 21 ∪ 12 , 1 .
 

(b) (0, 1) ∪ 21 , 2 .


(c) (0, 1) ∪ (2, 3).

52
(d) (0, 1) ∪ [1, 2).

6. Find out whether the following subsets of R are connected or discon-


nected in the usual metric:

(a) (−∞, 1].


(b) (−∞, 1) ∪ [1, ∞).
(c) The set of all irrational number.
(d) Set of all integers in
(e) Any finite set in R.

References:

1. W. Rudin ''Principles of Mathematical analysis ''third edition. Me


Grow hill Education(India) Private limited, New Delhi.

2. S. Kumareson ''Topology of Metric spaces 'S'econd Edition. Narosa


Publishing House. New Delhi.

53
BLOCK II
Unit V
Convergent sequences, subsequences and monotone sequences

5.1.1 Objectives
In this unit, we will give a brief development of the theory of sequences.
We study about the convergent sequences, monotone sequences and subse-
quences.

5.1.2 Introduction
Sequences arise naturally when we want to approximate quantities. For in-
stance, when we wish to use decimal expansion for the rational number 31
we get a sequence 0.3, 0.33, 0.333, .... we also understand that each term is
approximately equal to 13 up to certain level of accuracy. What do we mean
by this?

The idea of a sequence {xn } and its convergence to x ∈ R is another way


of saying that we give a sequence of approximation xn to x.

5.1.3 Sequences and Their convergence


Definition 5.1.3.1 Let X be a non-empty set. A sequence in X is a function
from the set of natural numbers to X, i.e f : N → X. We let xn = f (n) and
call xn the nth term of the sequence. One usually denote the sequence f by
{xn } or as an infinite tuple x1 , x2 , x3 , ..., xn , ...

Note:

1. If X = R, then the sequence is said to be a real sequence.

2. If X = C, then the sequence is said to be a complex sequence.

54
3. In the remainder of this unit, all sequence and series under considera-
tion will be real or complex valued.

Now let us look at some examples of sequences.

Example 23 Let f : N → X defined by f (n) = 2n or xn = 2n or the se-


quence is {2, 4, 6, 8, ...}

Example 24 Define xn = a for all n, the sequence {a, a, a, ...} is called a


constant sequence, for example {1, 1, 1, ...}.

Example 25 Let xn = n1 , the sequence is 1, 12 , 13 , 14 , ..., n1 , ... .




(−1)n+1
the sequence is 1, −1 , 1 , 1 , 1 , ... .

Example 26 Let xn = n
, 2 3 2k−1 2k

Example 27 Let xn = (−1)n , the sequence is {1, −1, 1, −1, 1, ...}.

Example 28 Let xn = n, the sequence is {1, 2, 3, 4, ...}.

Example 29 Let xn = 2n , the sequence is {2, 4, 8, 16, ...}.

1
1 1 1

Example 30 Let xn = 2n
, the sequence is , ,
2 4 8
, ... .

Example 31 Let x1 = x2 = 1 and define xn = xn−1 + xn−2 for all n ≥ 3,


the sequence is {1, 1, 2, 3, 5, 8, 13, 21, ...}.

2n+1
Example 32 Let xn = n2 −3
, then the first few terms of this sequence are
 −3
, 5, 67 , 13
9 11 13

2
, 22 , 33 ... .

55
Definition 5.1.3.2 Let {xn } be a real sequence, we say that {xn } converges
to x ∈ R if for every given  > 0, there exist N ∈ N such that for all k ≥ N ,
we have xk ∈ (x − , x + ), (that is, for k ≥ N , we have |x − xk | < ).

The number x is called a limit of the sequence {xn }. We then write


xn −→ x. We also say that {xn } is convergent to x. We write this as
limn→∞ xn = x. We say that a sequence {xn } convergent if there exist x ∈ R
such that xn −→ x.

Remark: Suppose {xn } is a sequence of real numbers converging to 'x'.


Geometrically, it means that if we plot the graph of this sequence and draw
any  - neighbourhood around x say, (x − , x + ), then there exists a positive
integer N such that for all n ≥ N the points xn lie inside this neighbourhood.

The definition of convergence of sequence {xn } can be written in terms


of quantities as follows,

∀  > 0, ∃ n0 ∈ N such that ∀ n ≥ n0 , xn ∈ (x − , x + ), equivalently


∀  > 0, ∃ n0 ∈ N such that ∀ n ≥ n0 , |x − xn | < .

Example 33 Let us consider constant sequence xn = c for all n. It is easy


to see that xn −→ c.

Example 34 If xn = n1 (If we plot the points of this sequence on the real line
it is easy to see that this sequence has limit equal to zero), then lim xn = 0.
n→∞

1
Example 35 If xn = 2n
, then limn→∞ xn = 0.

Example 36 If xn = (−1)n , then xn doesn’t converges.

Theorem 5.1.3.1 (Uniqueness of limit) If xn −→ x and xn −→ y, then


x = y.

56
Proof: Let  > 0 be given. Since xn −→ x and xn −→ y, there exists
integer n1 and n2 such that for k ≥ n1 , we have |xk − x| < 2 and for k ≥ n2 ,
we have |xk − y| < 2 . Consider N = max {n1 , n2 }. Then for all k ≥ N , we
have |x − y| = |x − xk + xk − y| ≤ |xk − x| + |xk − y| < 2 + 2 = . since  > 0
is arbitrary, we have x=y.

Definition 5.1.3.3 A sequence {xn } of real numbers is said to be bounded


if there exist c > 0 such that |xn | ≤ c for all n ∈ N.

Remark: If the sequence {xn } is the function from N to R, then the set
{xn : n ∈ N} is the image of that function. Note that {xn } is bounded iff the
image set {xn : n ∈ N} of the sequence is a bounded subset of R.

Example 37 Let xn = n1 , then xn is bounded.

Example 38 Let xn = n2 , then xn is unbounded.

Example 39 Let xn = (−1)n , then xn is bounded.

Theorem 5.1.3.2 Every convergent sequence of real numbers is bounded.

Proof: Let xn −→ x. By the definition of convergent sequence, for  = 1,


∃ N ∈ N, such that ∀ n ≥ N , |xn − x| < 1.
Then |xn | < |x − xk | + |x| = 1 + |x| for k ≥ N . To get estimate for all
xn , we let C = max {|x1 |, |x2 |, ..., |xN −1 |, 1 + |x|}. Then it is clear from the
definition of C that, |xn | < C ∀ n ∈ N. Hence xn is bounded.

Remark: Using the above theorem one can directly conclude that, the
sequence xn = n is divergent.

Theorem 5.1.3.3 (Algebra of convergent sequences) Let xn −→ x, yn −→ y


and α ∈ R. Then
i) xn + yn −→ x + y.

57
ii) αxn −→ αx.
iii) xn yn −→ xy.
iv) x1n −→ x1 , provided that x 6= 0 and xn 6= 0 ∀ n ∈ N.

Proof: i) Let  > 0 be given. Since xn −→ x there exists a natural number


n1 , such that for all k ≥ n1 , we have |xk − x| < 2 .
Similarly, there exists a natural number n2 , such that for k ≥ n2 , we have
|yk − y| < 2 .
Now choose, N = max {n1 , n2 }. Then for all k ≥ N , we have |xk − x| < 2 .
Then for all k ≥ N , we have |xk − x| < 2 and |yk − y| < 2 . Therefore, for
all k ≥ N

|(xn + yn ) − (x + y)| = |(xn − x) + (yn − y)|


≤ |(xn − x)| + |(yn − y)|
 
< + =
2 2
Hence the proof of (i).

ii) If α = 0,then the result is vacuously true .


Let α 6= 0, and  > 0 be given. Since xn −→ x, there exists N ∈ N such

that for all k ≥ N , we have |xk − x| < |α| . Hence for all k ≥ N , we have

|αxk − αx| = |α||xk − x| < |α| |α| . Hence αxk −→ αx.

iii) Let  > 0 be given. Since {xn } is convergent, there exists c > 0 such
that |xn | ≤ c for all n ∈ N. Since xn −→ x, there exists a n1 ∈ N such that ∀

k ≥ n1 , we have |xk − x| < 2(|y|+1) . Similarly, there exists a number n2 ∈ N,
such that ∀ k ≥ n2 , we have |yk − y| < 2c . Choose N = max {n1 , n2 }. Then
for all k ≥ n, we have

58
|xk yk − xy| = |xk yk − xk y + xk y − xy|
= |xk (yk − y) + y(xk − x)|
≤ |xk ||(yk − y)| + |y||(xk − x)|
≤ |c||(yk − y)| + (|y| + 1)|(xk − x)|
 
≤ c. + (|y| + 1)
2c 2(|y| + 1)
= .
Hence xk yk −→ xy, whenever xk −→ x and yk −→ y.

iv) Let  > 0 be given. Since xn −→ x and x 6= 0, there exists n1 ∈ N


such that ∀ k ≥ n1 , we have |xk | > |x|
2
.
2
Also there exists n2 ∈ N, such that ∀ k ≥ n2 , we have |xk − x| < |x|
2
.
Now choose N = max {n1 , n2 }, then for all k ≥ N ,
we have

− 1 = x − xk = 1 .|xk − x|
1
xk x xxk |x||xk |
1 2
< . .|x − xk |
|x| |x|
2 |x|2
< 2
|x| 2
= .
1
Hence xk
−→ x1 , whenever x 6= 0.

5.1.4 Monotone sequence


Definition 5.1.4.1 1) We say a sequence {xn } of real numbers is monoton-
ically increasing (↑) if for each 0 n0 we have xn ≤ xn+1 .

2) We say a sequence {xn } is strictly increasing if xn < xn+1 for all n ∈ N.

3) We say a sequence {xn } of real numbers is monotonically decreasing


if for each 0 n0 we have xn ≤ xn+1 .

59
4) We say a sequence {xn } is strictly decreasing if xn > xn+1 for all n ∈ N.

5) We say a sequence {xn } is monotonic if it is either monotonically in-


creasing or monotonically decreasing.

Remark: Any strictly monotonically increasing sequence is bounded below


by x1 . Hence such a sequence is bounded iff it is bounded above.

Example 40 Let xn = n1 , then clearly {xn } is strictly decreasing sequence.

Example 41 Let xn = n, then {xn } is a strictly increasing sequence.

Example 42 Let xn = 1, then {xn } is both monotonically increasing as well


as monotonically decreasing.

Example 43 Let xn = (−1)n , then {xn } is not monotonic.

Example 44 Consider the sequence {xn } defined by


1 1 1
xn = + + ··· + , n = 1, 2, . . .
n+1 n+2 n+n
1 1 1 1 1
T hen xn+1 = + + ··· + + + .
n+2 n+3 2n 2n + 1 2n + 2
Therefore,

1 1 1
xn+1 − xn = + −
2n + 1 2n + 2 n + 1
1 1
= − > 0, ∀ n.
2n + 1 2n + 2
This proves that xn < xn+1 , for all n and hence {xn } is a monotonically
increasing sequence.
Next, we have for all n,

60
1 1 1
xn = + + ··· +
n+1 n+2 n+n
1 1 1
< + + ··· +
n n n
1
= n.
n
=1
Since 0 < xn , for all n, we have 0 < xn < 1, for all n. Hence {xn } is
bounded.
Thus, {xn } is a bounded monotonic sequence. Therefore {xn } is convergent.
Example 45 Consider the sequence {xn } defined by
 n
1
xn = 1 + , n = 1, 2, . . .
n
Then we have by binomial theorem,

1 n

xn = 1 + n

= 1 + 1 + n(n−1)
2!
1
n2
+ · · · + n(n−1)...3.2.1
n!
1
nn

= 1 + 1 + 2!1 1 − n1 + 3!1 1 − 1 2
  
 n 1 − n
+ · · · + n!1 1 − n1 1 − n2 . . . 2!1 1 − n−1

n

1 1 1
≤1+1+ 2!
+ 3!
+ ··· + n!

1 1 1
≤1+1+ 21
+ 22
+ ··· + 2n−1
, sincen! ≥ 2n−1 f oralln
1
1−
= 1 + 1−21n
2
< 1 + 2 = 3, for all n.

Also since, xn ≥ 2 for all n, we have 2 ≤ xn < 3, for all n. Hence {xn }
is a bounded sequence.
Again, we have by binomial theorem

61
1
n+1
xn+1 = 1 + n+1

(n+1)(n+1−1) 1 (n+1)(n+1−1)(n+1−2) 1
=1+1+ 2! (n+1)2
+ 3! (n+1)3

+ · · · + (n+1)(n+1−1)...(n+1−(n−1))
n!
1
(n+1)n
+ (n+1)(n+1−1)...(n+1−n)
(n+1)!
1
(n+1)n+1
.

1 1 1 1 2
  
≥1+1+ 2!
1− n+1
+ 3!
1− n+1
1− n+1

1 1 2 n−1
  
+ ··· + n!
1− n+1
1− n+1
... 1 − n+1
.

1 1 1 1 2
  
≥1+1+ 2!
1− n
+ 3!
1− n
1− n

1 1 2 n−1
  
+ ··· + n!
1− n
1− n
... 1 − n
.

= xn , for all n.

Hence, xn ≤ xn+1 , for all n, proving that {xn } is a monotonic sequence.


Being a bounded monotonic sequence, the sequence {xn } converges.

Theorem 5.1.4.1 Let {xn } be an monotonically increasing sequence. Then


it is convergent iff it is bounded above.

Proof: If {xn } is convergent then it is bounded. Hence it is a bounded


above. Conversly suppose {xn } is bounded above. We need to prove that
{xn } is convergent.
Let S = {xn : n ∈ N} be the image of the sequence {xn }. clearly S is non-
empty and bounded above subset of real numbers. Hence it has a supremum.
Let 0 l0 be the supremum of S, we claim that xn −→ l.
Let  > 0, be given, clearly l −  < l, hence 0 l − 0 cannot be an upper
bound of S. Hence there exists N ∈ N such that xN > l − . Since xn
is monotonically increasing, for all n ≥ N , we have xN ≤ xn and hence
l −  < xN ≤ xn ≤ l < l + , that is xn −→ l.

Theorem 5.1.4.2 Let {xn } be an monotonically decreasing sequence. Then


{xn } is convergent iff it is bounded below.

Proof: Let S = {xn : n ∈ N}, clearly S is non-empty and bounded below


hence it has an infimum. Let l be the infimum of S. For any given  > 0,

62
l +  > l hence it is not a lower bound of S. Hence there exists N ∈ N such
that xn < l + . Since {xn } is monotonically increasing, for all n ≥ N , we
have xn ≤ xN and hence l −  < l ≤ xn < xN < l + , thus xn −→ l.
hence we have proved that if a monotonically decreasing sequence is bounded
below then it is convergent.
converse is true as any convergent sequence is bounded.

Lemma 5.1.4.1 (Sandwich lemma) Let {xn }, {yn }, {zn } be sequences such
that xn −→ α, yn −→ α and xn ≤ zn ≤ yn for all n. Then zn −→ α.

Proof: For given  > 0, choose n1 , n2 such that for all k ≥ n1 , we have
|xn − α| < 2 and ∀ k ≥ n2 , we have |y − n − α| < 2 . Let N = max {n1 , n2 }.
Then for k ≥ N , we observe.
α −  < xk ≤ zk ≤ yk < α + . that is |zk − α| <  ∀ k ≥ N . Hence zk −→ α.

Theorem 5.1.4.3 (Cantor’s theorem on nested intervals ) Let {[an , bn ]} be


a sequence of closed and bounded intervals such that
(i) [an+1 , bn+1 ] ⊆ [an , bn ] for all n ∈ N.
(ii) lim Sn = 0, where Sn = bn − an .
n→∞

T
Then [an , bn ] contains precisely one point.
n=1

Proof: Since [an+1 , bn+1 ] ⊆ [an , bn ] for all n ∈ N, we have an ≤ an+1 ,


bn+1 ≤ bn and an ≤ bn for n ∈ N. Therefore a1 ≤ a2 ≤ ... ≤ an ≤ ... ≤ bn ≤
.. ≤ b2 ≤ b1 . This shows that the sequence {an } is monotonically increasing
which is bounded above and the sequence {bn } is monotonically decreasing
which is bounded below. Hence both the sequences are convergent.
Let lim an = l and lim bn = m,
n→∞ n→∞
since lim (bn −an ) = 0, implies that lim bn − lim an = 0. i.e l = m = α say.
n→∞ n→∞ n→∞
Therefore α is the least upper bound of the sequence {an } and the greatest
lower bound of the sequence {bn }. Hence an ≤ α ≤ bn for all n ∈ N. This

T
implies α ∈ [an , bn ], for all n ∈ N. That is α ∈ [an , bn ]. We now prove that
n=1

T ∞
T
α is the only point in [an , bn ]. If suppose β ∈ [an , bn ], then an ≤ β ≤ bn
n=1 n=1
for all n ∈ N. Hence by Sandwich lemma β = α. Hence the proof of Cantor’s
intersection theorem.

63
5.1.5 Subsequences
Definition 5.1.5.1 Given a sequence {pn }, consider a sequence {nk } of pos-
itive integers, such that n1 < n2 < n3 < ....Then the sequence {pnk } is called
a subsequence of {pn }.
If {pnk } converges, its limit is called a subsequential limit of {pn }.
1 1
Example 46 Let xn = n
and sn = 2n
then {sn } is a subsequence of xn .

Example 47 Let {xn } = 1, 12 , 13 , 14 , 51 , 16 , ... and {tn } = 21 , 14 , 16 , 18 , 10


1 1
 
, 12 , ...
then {tn } is not a subsequence of {xn }

Theorem 5.1.5.1 If a sequence {xn } converges to x then every subsequence


of {xn } also converges to x.

Proof: Let {xnk } be a subsequence of xn . Now we need to prove that


xnk −→ x.

let  > 0 be given. Since xn −→ x, ∃N ∈ N such that for all k ≥ N ,


|xk − x| < . From the definition of subsequence if k ≥ N , then nk ≥ k ≥ N .
hence, for k ≥ N , we have |xnk −x| < . This implies that {xnk } converges to
x. As {xnk } is an arbitrary subsequence of {xn } we can conclude that every
subsequence of a convergent sequence converges to the same limit as that of
the sequence.

Theorem 5.1.5.2 The sequence {Pn } converges to P if and only if every


subsequence of {Pn } converges to P .

Proof: If {Pn } converges to P then we need to prove that every subse-


quence {Pnk } of {Pn } converges to 0 P 0 . The proof follows from the Theorem5.1.5.1.
conversly suppose every subsequence of {Pn } to P. then the sequence {Pn }
also converges to P as {Pn } is itself a subsequence of {Pn }.

Theorem 5.1.5.3 (Bolzano- weiestrass Theorem): If {xn } is a bounded real


sequence, then it has a convergent subsequence.

64
Proof: First we will prove that given any real sequence {xn }, there exists
a monotone subsequence.

Consider the set S = {n ∈ N : xm < xn f or m > n}.

Case 1: If S is finite. Let N be any natural number such that k ≤ N


∀ k ∈ S. Let n1 > N , then n1 ∈ / S. Hence there exists n2 > n1 such that
xn2 ≥ xn1 . Since n2 > n1 > N , n2 ∈ / S. We can find an n3 > n2 such that
xn3 ≥ xn2 . Hence by mathematical induction, if nk is so chosen such that
nk ∈/ S, hence ∃ nk+1 ∈ N such that nk+1 > nk and xnk+1 ≥ xnk . Clearly
{xnk } is a monotonically increasing subsequence of {xn }.

Case 2: If S is infinite. Let n1 be the least element of S. Let n2 be the


least element of S − {n1 } and so on. We thus have a listing of S : n1 <
n2 < n3 <, ... since nk−1 ∈ S and nk−1 < nk , we see that xnk < xnk−1 , for all
k. Clearly {xn }. Hence {xn } has a monotone subsequence, say {xnk }. Since
xn is bounded, {xnk } is also bounded. Thus {xnk } is monotone and bounded.
Hence it is convergent.

some typical uses of subsequences

1. It helps to find out the limit of some convergent sequences.

1 1
(a) Let xn = a n , a > 0 , then it is easy to prove that 1 < a n ∀ n ∈ N,
and it is also clear that {xn } is monotonically decreasing sequence.
1
Hence it converges, say a n −→ l. Since every subsequence of of a
1
convergent sequence converges to the same limit. i.e. a 2n −→ l.
1 1
But a n = (a 2n )2 −→ l2 . By the uniqueness of limit we have l = l2 .
1
Which implies l = 0 or l = 1. Since 1 < a n ∀ n ∈ N implies l = 1.
1
Hence a n −→ 1, a > 0.


(b) Let xn = 2 + xn−1 , where x1 = 1. If we could know that this
sequence is√convergent say xn −→ l then xn−1 also converges to l.
Hence l = 2 + l, i.e l2 − l − 2 = 0. Hence l = 2.

65
2. It helps to show few sequences are divergent.

Show that the sequence {(−1)n } is divergent. Suppose {(−1)n } is con-


vergent. Then every subsequence of {(−1)n } converges to the same
limit. But the constant sequences {1} and {−1} are subsequences of
{(−1)n } which are converges to 1 and -1 respectively. Which is a con-
tradiction. Hence we conclude that {(−1)n } is not convergent.

5.1.6 Terminal problems:


1. Given that xn −→ 1, identify the limits of the sequences whose nth
terms are

(a) 1 − xn ,
(b) 2xn + 5,
4+x2n
(c) xn
.

2. Let xn −→ x. assume that xn ≥ 0 for all n. Then show that x ≥ 0.


√ p √
3. If s1 = 2 and sn+1 = 2 + sn then prove that lim sn = 2.
n→∞

4. Prove that the following sequences are divergent.

(−1)n
(i) 5
(ii) cos nπ (iii) n2 (iv) sin nπ
2
.

(n2 +13n−41) cos(2n )


5. Prove that the sequence {xn } where xn = n2 +2n+1
has a con-
vergent subsequence.

6. True or false : A sequence {xn } is bounded iff every subsequence of


{xn } has a convergent subsequence.

7. Find the limit of the sequence {xn } defined as x1 = 2 and xn+1 =


1
2
(xn + x2n ).

66
5.1.7 Keywords:
Sequence, convergent sequence, monotone sequence and subsequence.

5.1.8 References:
1. W. Rudin, ''Principles of Mathematical analysis'' Third edition. Me
Grow hill Education(India) Private limited, New Delhi.

2. S. Kumareson, ''Topology of Metric spaces'' Second Edition. Narosa


Publishing House. New Delhi.

67
g
Unit VI
Cauchy sequence, Upper and lower limits, some special sequences

6.1.1 Objectives
In this unit we will be studying about Cauchy sequence, upper and lower
limits and some special sequences.

6.1.2 Introduction
Knowing limit of some standard sequences helps us to evaluate limit of many
other sequences as well. Hence we will be learning the limit of such sequences.
Many a times, knowing a sequence whether it is convergent or not is impor-
tant even though if we can’t find the limit point of a sequence. Cauchy’s
criterion helps us to do so! There are incidences where it is enough to ap-
proximate the limit point to find its exact value. One such tool is upper and
lower limit. We will be learning in detail, still it is suggested to refer the
below mentioned books for deeper understanding.

6.1.3 Cauchy sequence


Definition 6.1.3.1 A sequence {xn } in R is said to be Cauchy sequence if
∀  > 0, ∃ N ∈ N such that |xn − xm | <  whenever m, n > N.

Theorem 6.1.3.1 Every convergent sequence is Cauchy sequence.

Proof : Let {xn } be a convergent sequence in R, say xn −→ x. Let  > 0


be given, ∃ N ∈ N such that |xn − x| < 2 ∀ n ≥ N . Choose m, n > N ,
consider |xn − xm | = |xn − x + x − xm |
≤ |xn − x| + |xm − x|
< 2 + 2 = .
Hence {xn } is Cauchy.

Theorem 6.1.3.2 A Cauchy sequence of real numbers is convergent.


Proof : Let {xn } be a Cauchy sequence of real numbers. First we will prove
that the sequence {xn } is bounded. Let  = 1. Then there exists a natural

68
number k such that ∀ n, m ≥ k, we have |xn − xm | < 1. Hence for any xn
where n > k, we have |xn | = |xn − xk + xk |
≤ |xn − xk | + |xk |
< 1 + |xk |
Let B = {|x1 |, |x2 |, ..., |xk−1 |, 1 + |xk |} then for every n ∈ N, |xn | < B. Hence
{xn } is a bounded sequence of real number. By Bolzano-Weierstrass theorem,
it has a convergent subsequence. let xnk −→ l be that convergent subsequence.
We now prove that the sequence {xn } converges to l. Let  > 0 be given, then
there exists a k1 ∈ N such that |xm − xn | < 2 for all m, n ≥ k1 . There exists
a k2 ∈ N such that |xnp − l| < 2 for all np ≥ k2 . Let N = max {k1 , k2 }, then
|xn − l| = |xn − xnN + xnN − l| < 2 + 2 = .
Hence xn −→ l.

6.1.4 Upper and lower limits


Introduction: If a sequence {xn } is convergent, then we know that every
subsequence of {xn } converges to the same limit as that of {xn }. If a se-
quence {xn } is not convergent then the following question naturally arises
such as ''does it have a subsequence which is convergent'' ?

Answering these question lead mathematicians to the development of new


concepts such as upper and lower limits. This concepts are very useful in
many context of analysis.

Definition 6.1.4.1 Let {sn } be a sequence of real numbers with the following
property: for every real M there is an integer N such that for every n ≥ N
implies sn ≥ M . We then write sn −→ +∞ and say sn diverges to +∞.

Similarly, if for every n ≥ N implies sn ≤ M . We then write sn −→ −∞


and say sn diverges to −∞.

Definition 6.1.4.2 Let {sn } be a sequence of real numbers. Let E be the


set of all number x(In the extended real number system) such that snk −→ x
for some subsequence {snk } of {sn }. This set E contains all sub sequential
limits including possibly the numbers +∞ and −∞.
Let
s∗ = sup E and s∗ = inf E.

69
The numbers s∗ and s∗ are called the upper and lower limits of sn , we use
the following notation
limn→∞ sup sn = s∗ and limn→∞ inf sn = s∗ .
also, we use the following notation

lim sn = s∗ and lim sn = s∗ .


n→∞ n→∞

Example 48 Let sn = (−1)n , then it has only two subsequence which are
convergent, they are the constant sequences {1} and {−1}. Hence E =
{−1, 1}, s∗ = 1 and s∗ = −1.

lim sup sn = 1 and lim inf sn = −1


n→∞ n→∞

1 + n1 , then it has only two convergent sub-


n
 
Example 49 Let sn = (−1)
1 1
. Hence E = {−1, 1}, s∗ = 1

sequences, they are 1 + 2n and − 1 + 2n−1
and s∗ = −1.

Example 50 Let {xn } be sequence of all rational numbers. Then it has


infinitely(uncountably) many subsequences which are convergent. It is easy
to see that for given any real number x ∈ R there exists a subsequence of
{xn } which converges to x. Hence E = R ∪ {±∞}. hence

lim sup sn = +∞ and lim inf sn = −∞


n→∞ n→∞

Example 51 Let {xn } be defined as {1, 2, 3, 1, 2, 3, 1, 2, 3, ...} then clearly


it has only three convergent subsequences whose sub sequential limits are
E = {1, 2, 3}, then s∗ = 3 and s∗ = 1.

Theorem 6.1.4.1 If {sn } is a sequence of real numbers, then

lim sup sn ≥ lim inf sn .


n→∞ n→∞

proof: Since we know that for any non-empty subset E of R inf E ≤ sup E.
Hence the result follows from the definition of limit superior and limit inferior
or upper and lower limit.

70
Now we will discuss the equivalent definition of upper and lower limit of
a sequence.

Definition 6.1.4.3 1) Let {sn } be a sequence of real numbers. Let Mn =


lub {sn , sn+1 , sn+2 , ...}. Then {Mn } is a monotonically decreasing sequence
as {sn+1 , sn+2 , sn+3 , ...} ⊆ {sn , sn+1 , sn+2 , ...}. Then

lim sup sn = lim Mn .


n→∞ n→∞

2) Let {sn } be a sequence of real numbers. Let mn = glb {sn , sn+1 , sn+2 , ...}.
Then
lim inf sn = lim mn .
n→∞ n→∞

Theorem 6.1.4.2 If {xn } and {yn } are sequences of real numbers then
1. lim sup(xn + yn ) ≤ lim sup(xn ) + lim sup(yn )
n→∞ n→∞ n→∞

2. lim inf(xn + yn ) ≥ lim inf(xn ) + lim inf(yn ).


n→∞ n→∞ n→∞

Proof(a): If one of {xn } and {yn } is not bounded above or both are not
bounded above then lim sup(xn ) + lim sup(yn ) = +∞ and hence
n→∞ n→∞

lim sup(xn + yn ) ≤ lim sup(xn ) + lim sup(yn ) = +∞


n→∞ n→∞ n→∞

(Case2)If both {xn } and {yn } are bounded above, put


An = lub {xn + yn , xn+1 + yn+1 , . . . }
Bn = lub {xn , xn+1 . . . } and Cn = lub {yn , yn+1 , . . . }.
Then, we have
An ≤ Bn + Cn for n = 1, 2, . . .
Now, if {An } , {Bn } and {Cn } are all convergent then we have
lim An ≤ lim Bn + lim Cn
n→∞ n→∞ n→∞

Therefore
lim sup(xn + yn ) ≤ lim sup(xn ) + lim sup(yn )
n→∞ n→∞ n→∞

71
(Case3) If {An } diverges to −∞, then lim sup(xn + yn ) = −∞ and hence
n→∞

lim sup(xn + yn ) ≤ lim sup(xn ) + lim sup(yn ).


n→∞ n→∞ n→∞

(Case4) If {Bn } or {Cn } or both diverges to −∞ then so does {Bn + Cn } .


Hence {An } diverges to −∞. In this case we have

lim sup(xn + yn ) = −∞ = lim sup xn + lim sup yn .


n→∞ n→∞ n→∞

Thus in all the cases, we have


lim sup(xn + yn ) ≤ lim sup xn + lim sup yn .
n→∞ n→∞ n→∞

This completes the proof of (a).

(b) Same as of (a).

Theorem 6.1.4.3 Let {xn } be a sequence of real numbers, E be the set of all
sub sequential limits of {xn } and let x∗ = sup E. Then x∗ has the following
properties:
(a) x∗ ∈ E.
(b) If x∗ < α, then there is an integer N such that xn < α, for all n ≥ N .
Moreover x∗ is the only number with the properties (a) and (b).

Proof: Exercise.

Example 52 Consider the sequence {xn } defined by


1 1 1
xn = 1 + + + · · · + , n = 1, 2, . . .
2 3 n
1
Let  = 4
and N be any positive integer. Then for any n ≤ N , we have
1 1 1
|x2n − xn | = + + ··· +
n+1 n+2 n+n

1 1 1
≥ + + ··· +
2n 2n 2n

72
1 1
=n =
2n 2
Therefore |x2n − x| ≮ 14 , proving that {xn } is not a Cauchy sequence. hence
{xn } is divergent.

6.1.5 Some important limit or some special sequences


Analysis deals with unknown sequences or functions by trying to compare
their behaviour with known things. If we want to be very skilled in this tech-
niques, it is of paramount importance that we have a quite good command
over commonly occurring sequences or functions.

Theorem 6.1.5.1 We have the following important limits:


(1) Let 0 ≤ r < 1 and xn = rn , then xn → 0.
(2) Let −1 < t < 1, then tn → 0.
(3) Let |r| < 1, then nrn → 0.
1
(4) Let a > 0 then a n → 1.
1
(5) n n → 1.
an
(6) Fix a ∈ R, then → 0.
n!
1
(7) If p > 0, then p → 0.
n

(8) If p > 0 and α ∈ R, then → 0.
(1 + p)n
Proof: (1) Let 0 ≤ r < 1 and xn := rn . If r = 0, the sequence is the zero
sequence and hence is convergent. So we shall assume that 0 < r < 1. Since
rn > rn+1 for n ∈ N, the sequence {xn } is decreasing. It is bounded below by
zero. so we conclude that it is convergent. Let xn → l. Now by the algebra of
convergent sequence rxn → rl. But rxn = xn+1 and hence rxn → l. Hence
by the uniqueness of the limit, we conclude that l = 0. hence for 0 ≤ r < 1,
the sequence rn → 0.

(2) Let −1 < t < 1, we know that tn → 0 iff |tn | = |t|n → 0. hence the
result follows from the last result.

(3) It is enough to prove the result for 0 < r < 1. If 0 < r < 1, then
1
we can write r = 1+h for some h > 0. Using binomial theorem, we have

73
(1 + h)n = 1 + nh + n(n−1)
2
h2 + · · · + hn > n(n−1)
2
h2 , since all terms are
n 2
positive. Thus we have 0 ≤ nr ≤ h2 (n−1) for all n ≥ 2. Hence using the
sandwich lemma, we have nrn → 0.
1
(4) case 1: a > 1, first we will prove that a n > 1. We know that if
1
0 ≤ a ≤ b, then 0 ≤ an ≤ bn for any natural number n. If suppose a n < 1
1
then a < 1 which is a contradiction. Thus, we have a n > 1. Hence we
1
can write a n = 1 + hn with hn > 0. This implies by binomial theorem,
a = (1 + hn )n ≥ nhn and hence 0 ≤ hn ≤ na . This means hn → 0. Therefore,
1
a n → 1 as n → ∞.
1
case 2: If 0 < a < 1, we apply the result to b n where b = a1 > 1. Observe
1 1 1
that a n = 1 . By the first case, we have b n → 1,and hence using the algebra
bn
1 1 1
of limits 1 → 1. This proves that lim a n = lim 1 = 1.
bn n→∞ n→∞ b n
1 1
(5) For n > 1, we have n n > 1. So we can write n n = 1 + hn with
hn > 0. it is enough to show that hn → 0. By using binomial theorem we
have n = (1 + hn )n = 1 + nhn + n(n−1)
2
h2n + · · · + hnn > n(n−1)
2
h2n . This implies
2
0 ≤ h2n ≤ n−1 . hence h2n → 0 by the sandwich lemma. Since we know that
1
hn → 0 iff h2n → 0. We obtain hn → 0. Hence n n → 1 as n → ∞.

|a|n
(6) It is enough to prove that → 0. In particular, it is enough to
n!
an
show that → 0 for a > 0. Assume a > 0, then by Archimedean property,
n!
there exists N ∈ N such that a < N . Then for n > N , we have
an
= a1 . a2 . . . Na Na+1 . . . na

n!
≤ crn−N where c = a1 . a2 . . . Na and r = Na
n
Thus we have 0 ≤ an! ≤ cr−N rn .
Since 0 < r < 1 , rn → 0 and hence cr−N rn → 0. Hence by the sandwich
an
lemma we have → 0.
n1
1
(7) let  > 0 be given, then 1 p > 0 . By Archimedean principle ∃ n ∈ N
1
such that n > 1 p . i.e np > 1 =⇒ n1p < . Since  is arbitrary we obtain
1
np
→ 0 as n → ∞.

74
(8) Let k be an integer such that k > α, and k > 0. By Binomial theorem
we have,
(1 + p)n > n(n−1)...(n−(k−1))
k!
pk .

n
n > 2k implies 2
> k. Hence
n(n−1)...(n−(k−1)) k n k pk

k!
p > 2 k!

Hence for n > 2k, we have


n(n−1)...(n−(k−1)) k n k pk

(1 + p)n > k!
p > 2 k!

1 2k k! k
=⇒ 0 < < .n for (n > 2k)
(1 + p)n pk
nα 2k k! α−k
Hence 0 < < n
(1 + p)n pk
Since α − k < 0, nα−k → 0 by (7).

hence → 0 as n → ∞.
(1 + p)n

6.1.6 Terminal problems:


x + xn
1. For what values of x, the sequence where n-th term is xn = is
1 + xn

convergent.

1 + a + a2 + a3 ...... + an−1
2. Find the limits of the sequence where n-th term term is .
n!
n
3. Let an = . Show that lim an = 0.
2n n→∞

1+a
4. Let a ∈ R. Consider x1 = 1, x2 = by induction define
2
1 + xn−1
xn = , discuss the convergence of {xn }.
2

75
√ p √ √
q p
5. Consider the sequence 2, 2 + 2, 2 + 2 + 2, . . . .
Show that it converges to 2.

6. Prove that the sequence {sin n} is divergent.

7. True or False : A sequence {xn } is bounded iff every subsequence of


{xn } has a convergent sub sequences.

an2 + b
8. show that lim = ∞ if a, c > 0.
n→∞ cn + d
9. Let {an } be a sequence of positive real numbers. Assume that
an+1
lim = α. Then show that lim (an )= α.
n→∞ an n→∞

√ √
10. Prove that the sequence n+1− n →0

1 1 1
11. let xn = √ +√ +...+ √ , then prove that lim xn = 1
n2 + 1 n2 + 2 n2 + n n→∞

1
12. Let 0 < a < b. Then prove that the sequence (an + bn ) n → b.

6.1.7 Keywords:
Cauchy sequence, upper and lower limit.

6.1.8 References:
1. W. Rudin, ''Principles of Mathematical analysis'' Third edition. Me
Grow hill Education(India) Private limited, New Delhi.

2. S. Kumareson, ''Topology of Metric spaces'' Second Edition. Narosa


Publishing House. New Delhi.

76
g
Unit VII
Series

7.1.1 Objectives
In this unit, we will be studying about series of numbers and several tests
for checking whether a given series is convergent or not.

7.1.2 Introduction
We all know to add finite number of quantities. This experiance of us with
adding finite numbers helps us to ask the question, ''Can we add infinite set
of numbers ? ''. If we can then, how do we define such a number ?

Previous experience of sequence of numbers helps us to define infinite sum


as the limit of sequence of partial sums. This concept is also a very important
as well as basic for understanding much deeper concepts in analysis.

7.1.3 Series
Definition 7.1.3.1 Given a sequence {an } of real numbers, a formal sum

P P
of the form an or ( an , f or short) is called an infinite series.
n=1

th
For any n ∈ N,P the finite sum sn := a1 + · · · + an is called the n partial sum
of the series an .

P A more formal definition of an infinite series is as follows. By the symbol


an we mean the sequence {sn }, where sn := a1 + a2 + · · · + an .
P
we say that the infinite series an is convergent if the sequence {sn } of
partial sums is convergent. In such a case, the limit s :=Plim sn is called the
sum of the series and we denote this fact by the symbol an = s.
P
Definition 7.1.3.2 We say that the series an is divergent if the sequence
of its partial sums is divergent.

77
Example 53 (Geometric series) This is the most important example. Let

z n . We claim
P
z ∈ R be such that |z| < 1. Consider the infinite series
n=0
1
that the series converges to α = for |z| < 1.
1−z
n 1 − z n+1
Its nth partial sum sn is given by sn := zk =
P
, since |z| < 1,
k=0 1−z
∞ 1
1
lim z n+1 = 0 implies lim sn = 1−z zn =
P
for |z| < 1 . Hence for
n→∞ n→∞ n=−0 1−z
|z| < 1.

Example 54 Telescoping series: let {an } and {bn } be two sequence such
that an = bn+1 − bn , n ≥ 1. We note that s1 = a1 = b2 − b1 , s2 = a1 + a2 =
b2 − b1 + b3 − b2 = b3 − b1 and sn = a1 + a2 + · · · + an = (b2 − b1 ) + (b3 − b2 ) +
· · · + (bn+1 − bn ) =P
bn+1 − b1 .
Then we say that an converges iff lim bn exists, in which case we have
P n→∞
an = −b1 + lim bn .
n→∞

P 1 1 1 1
Example 55 , we can note that = − . Here
n(n + 1) n(n + 1) n n+1
bn = −1 1
= 1 + lim − n1 = 1
P
n
so from the above example n(n+1) n→∞
P 1
i.e = 1.
n(n + 1)
P n
Example 56 Consider .
n→∞ n4
+ n2 + 1
n n
Note that an = 4 2
= 2
n +n +1 (n + 1)2 − n2
n
= 2
(n + n + 1)(n2 − n + 1)
 
1 1 1
= −
2 n2 − n + 1 n2 + n + 1
Hence an = bn+1 − bn
 
1 1
where bn = ,
2 n2 − n + 1
∞ n 1 1
= 21 − 12 lim 2
P
thus 4 2
=
n=1 n + n + 1 n→∞ n + n + 1 2

78
∞ 1
P
Example 57 (Harmonic Series) The series is divergent.
n=0 n
Observe that the harmonic series is a series of positive terms. Hence it is
convergent if and only if its partial sums are bounded above. We show that
the subsequence {S2k } of the partial sums is not bounded above.
1 1 1 1 1
Note that k−1 + · · · + k > 2k−1 k = 12 as each term k−1 > k for
2 +1 2 2 2 +i 2
1 ≤ i ≤ 2k−1 .
we have the following estimate
S1 = 1
S2 = 1 + 12 = 32
S22 = 1 + 12 + 31 + 14  > 1 + 12 + 14 + 41 = 1 + 12 + 24 = 1 + 22
 

S33 = 1 + 12 + 13 + 41 + 15 + · · · + 81 > 1 + 21 + 2 14 + 22 213 = 1 + 23 , so that


S2k > 1 + k2 . It follows that {Sn } is not bounded above. Hence we conclude

1
P
that the harmonic series n
is not convergent.
n=0

We all know that Cauchy criterion for sequence of real numbers, following is
the Cauchy criterion for series of numbers.


P
Theorem 7.1.3.1 (Cauchy criterion) The series an converges if and
n=0
only if for each  > 0 there exists N ∈ N such that if m ≥ n ≥ N
m
P
implies |sn − sm | < . i.e ‘| ak | ≤ .
k=n

P
Proof: Let an be convergent. Then the sequence {sn } of its partial
n=0
sums is convergent. We know that a real sequence is convergent iff it is
Cauchy.Hence {sn } is convergent if and only if it is Cauchy. The result
follows from the definition of Cauchy sequence.

P
Theorem 7.1.3.2 If an converges, then lim an = 0.
n=0 n→∞

P
Proof: Let  > 0 be given. Since the sum an is convergent, the se-
n=0
quence {sn } of partial sums is convergent and in particular, it is Cauchy.
Hence for the given  > 0, there exist N ∈ N such that for n ≥ m ≥ N we
have |sn − sm | < . Now if we take any n ≥ N + 1, then an = sn − sn−1 .

79
Hence we obtain |an | = |sn − sn−1 | <  for n ≥ N + 1. This proves that
lim an = 0.
n→∞


1
P
Note that the converse of the above is not true. for example n
diverges
n=0
1
even though n
→ 0.

Corollary 7.1.3.1 Given  > 0, there exists N ∈ N such that the tail of the

P ∞
P
convergent series an is given by an < .
n=0 n=N +1

P
Proof: Since an is convergent, its partial sum converges say sn → s,
n=0
i.e for  > 0 there exists N ∈ N such that for n ≥ N , |sn − s| ≤ . In

P N
P
particular s −  < sN , that is s − sN < , where s = an and sN an ,
n=0 n=1

P
hence s − sn = an < .
n=N +1

We now turn to a convergence test of a different nature the so called


''comparison test ''.


P ∞
P
Theorem 7.1.3.3 Let an and bn be series of real numbers then:
n=0 n=0

P
(a) If |an | ≤ cn for n ≥ N0 , where N0 is some fixed integer and if cn
n=0

P
converges, then an converges.
n=0

P ∞
P
(b) If an ≥ bn ≥ 0 for n ≥ N0 and if bn diverges, then an diverges.
n=0 n=0

P
Proof: (a) Since cn converges by Cauchy’s criterion, we have that for
n=0
given  > 0, there exists N ∈ N such that N ≥ N0 and if m ≥ n ≥ N implies
m
P
ck ≤ .
k=n

By generalized triangle inequality, we have

80

Pm P m m
P

a k
≤ |a k | ≤ ck ≤  for m ≥ n ≥ N
k=n k=n k=n


P
which implies that an is a convergent series by Cauchy’s criterion.
n=0

P
(b) On the contrary, suppose an converges, then by applying (a) we
n=0

P ∞
P
obtain that bn is convergent, which is a contradiction. Hence an is a
n=0 n=0
divergent series.
P∞ 1
Corollary 7.1.3.2 (Harmonic p-series ) Consider the series p
we prove
n=1 n
that the harmonic p-series is divergent if 0 ≤ p ≤ 1 and is convergent if p > 1.
The case when p = 1 is already done in Example 57.
For 0 < p < 1, let
n 1 1 1
≥ n p = p−1 = n1−p
P
sn = p
k=1 k n n

hence Since p ≤ 1, implies (1 − p) > 0 hence lim n1−p = ∞, thus the series
n→∞

−p
P
n is divergent for 0 < p < 1.
n=0

For p > 1, observe that


1 1 2 2
p
+ p < p = p,
2 3 2 2
 2
1 1 1 1 4 2
p
+ p+ p+ p < p = ,
4 5 6 7 4 2p
 3
1 1 1 8 2
p
+ p + ··· + p < p = ,...
8 9 15 8 2p

 n
P 2
we know that the geometric series is convergent if p > 1. it
n=0 2p
follows from the comparison test that the p-harmonic series is convergent if
p > 1.

81
In many cases the application occur if the terms of the series are monoton-
ically decreasing. The theorem of Cauchy is therefore of particular interest.
The striking features of the following theorem is that a subsequence of {an }

P
determines the convergence or divergence of an .
n=0


P
Theorem 7.1.3.4 Suppose a1 ≥ a2 ≥ a3 ≥ · · · ≥ an . Then the series an
n=0
converges if and only if the series

2n a2n = a1 + 2a2 + 4a4 + 8a8 + . . .
P
n=0

converges.
Proof: Let sn = a1 + a2 + a3 + · · · + an and tk = a1 + 2a2 + 4a4 + 8a8 +
· · · + 2k a2k . Since the sequence {an } is monotonically decreasing, for n < 2k
we have

sn ≤ a1 + (a2 + a3 ) + · · · + (a2k + · · · + a2k−1 −1 )

≤ a1 + 2a2 + 4a4 + 8a8 + · · · + 2k a2k

= tk
so that sn ≤ tn .
∞ ∞
Hence by 00 comparison test 00 we have if 2n a2n converges then
P P
an con-
n=0 n=0
verges.
Conversely, choose n > 2k ,

sn = a1 + a2 + (a3 + a4 ) + · · · + (a2k−1 +1 + · · · + a2k )

1 
≥ a1 + a2 + 2a4 + · · · + 2k−1 a2k
2
1
= tk .
2
We have, if n > 2k then 2sn ≥ tk .
∞ ∞
2k a2k
P P
Thus again using comparison test if the series an convergent then
n=0 n=0
is convergent. Hence the proof the theorem.

The next theorem is a particular example of the above atheorem.

82
Theorem 7.1.3.5 If p > 1, then the series

P 1
p
n=2 n(log n)

converges; if p ≤ 1, the series diverges.


Proof: Since the function log x is monotonically
 ∞increasing implies that
1
{log n} is an increasing sequence. Hence is a decreasing se-
n log n n=2
quence of positive terms. Hence we can apply the above theorem. Which
∞ 1 ∞ 1
2k k
P P
implies p
converges iff converges.
n=2 n(log n) k=2 2 (log 2k )p
∞ 1 ∞ 1 1 ∞ 1
2k k
P P P
= = , and know that harmonic
k=2 2 (log 2k )p k=1 (k log 2)p (log 2)p k=1 k p
P∞ 1
p− series is convergent iff p > 1. Hence p
converges iff p > 1.
n=2 n(log n)


P 1
Example: − On similar line, one can prove that the series
n=3 n log log log n

P 1
is diverges, where as is convergent.
n=3 n log n(log log n)2

Now we will move on to the further tests which will help us to examine the
convergence of numerous series.
P
Theoremp7.1.3.6 (Root test) Let an be a series of rel numbers and α =
lim sup n |an |. Then
n→∞
P
a) If α < 1, an converges.
P
b) If α > 1, an is diverges.

c) If α > 1, test is inconclusive .

Proof: (a) If α < 1, we can choose β so that α < β < 1, and an integer
N such that
pn
|an | < β for n ≥ N .

83
that is, for n ≥ N implies |an | < β n

β n converges. Hence the series
P
since 0 < β < 1, the geometric series
n=0

P
an converges by comparison test.
n=0
p
(b) If α > 1, then there exist a sequence {nk } such that nk |ank | → α.
Hence |an | > 1 for infinitely many values of n, so that the sequence {an }

P
doesn’t converges to 0. Thus the series an doesn’t satisfies the necessary
n=0

P
condition for converges. Hence an diverges.
n=0
P∞ 1 P∞ 1
(c) considfer the series and 2
.
n=1 n n=1 n
q q ∞
Let α1 := lim sup n n1 = 1 and α2 := lim sup n n12 = 1 but the series 1
P
n→∞ n→∞ n
n=1

1
P
diverges where as n2
converges. Hence the result is inconclusive if α = 1.
n=1

P∞
Theorem 7.1.3.7 (Ratio Test) The series an
n=0
(a) Converges if lim sup an+1 < 1,

n→∞ an

(b) Diverges if an+1 ≥ 1 for all n ≥ n0 , where n0 is some fixed integer.

an

an+1
Proof: (i) If lim sup an , then there exists β < 1 and an integer N1 ,
n→∞
such that

an+1
an < β for n ≥ N1 .

In particular,
|aN +1 | < β|aN |,

|aN +2 | < β|aN +1 | < β 2 |aN |,

84
|aN +p | < β p |aN |.

That is,

|an | < |aN |β −N β n for n ≥ N ,


∞ ∞
β n converges, hence β −N β n con-
P P
since β < 1, the geometric series
n=0 n=0

P
verges. Thus by comparison test an converges.
n=0

an+1
(ii) If suppose an ≥ 1 for all n ≥ n0 . Hence lim |an | = 6 0, which
n→∞

P
implies that lim an 6= 0 as lim an = 0 iff lim |an | = 0. Hence an doesn’t
n→∞ n→∞ n→∞ n=0

P
satisfies the necessary condition for convergence, thus an diverges.
n=0

Remark :
an+1
1) The knowledge that lim = 1 implies nothing about the conver-
n→∞ an
∞ ∞ ∞
1 1
P P P
gence of an . Examples n
and n2
demonstrate this as in the proof
n=0 n=1 n=1
of previous theorem.

2) The root test has wide scope there are some incidences where the ratio
test fails but the root test gives the divergence or convergence of the series.
Following example illustrate this.

Example 58 Consider the series


1
2
+ 13 + 1
+
22
1
32
+ 1
23
+ 1
33
+ ...
√ q
1 an+1 3 n
√1 = lim 12
2n

for which lim sup n an = lim 2n
= 2
and lim sup 2
=
n→∞ n→∞ n→∞ an n→∞
+∞. The above series is convergent by root test, but ratio test fails!

Theorem 7.1.3.8 (Integral test) Assume that f : [1, ∞) → (0, ∞) is con-


Rn
tinuous and decreasing. Let an := f (n) and bn = f (t)dt. Then:
1

85

P
(i) an converges if {bn } converges.
n=0

P
(ii) an diverges if {bn } diverges.
n=0
n+1
R
Proof: For n ≥ 2, we have an ≤ f (t)dt ≥ an−1 so that
n

n
P Rn n−1
P
ak ≥ f (t)dt ≥ ak
k=2 1 k=1

If the sequence {bn } converges, then {bn } is a bounded increasing sequence


Pn Pn
and ≥ bn . Hence the partial sum sn = ak is a convergent sequence
k=2 k=1
being a monotonically increasing bounded sequence.
n−1
P ∞
P
If {bn } diverges, then bn → ∞. Since bn ≥ ak , the series an diverges.
k=1 n=0


1
P
Example 59 (Typical applications of the integral test). The p-series np
n=0
converges if p > 1 and diverges if p ≤ 1. Let f : [1, ∞) → [0, ∞) be defined
1 Rn Rn 1 Rn −p x−p n
by f (x) := p then f (x)dx = dx = x dx = |
x 1 1 x
p
1 −p + 1 1
n1−p 1
= −
−p + 1 −p + 1

∞ Z
(
n −1
X
−p+1
if p > 1
f (x)dx =
n=0 1 ∞ if p < 1
Hence by the integral test we obtain the convergence of p-series.

P
Theorem 7.1.3.9 (Abel- Pringsheim) If an is a convergent series of non-
n=0
negative terms with {an } decreasing, then nan → 0.
P∞
Proof: Since an is convergent, there exists N such that |sn − sm | <  for
n=0
n, m ≥ N . For k ≥ N ,
ka2k ≤ ak+1 + · · · + a2k = s2k − sk
since a2n+1 ≤ a2n , we have

86
(2n + 1)a2n+1 ≤ (2n + 1)a2n ≤ (2n)a2n + a2n . → (1)

P
now the series an is convergent, the sequence {an } converges to 0
n=0
and hence the subsequence {a2n } converges to zero. It follows from (1) that
sequence {(2n + 1)a2n+1 } is convergent. We conclude that nan → 0.

7.1.4 Terminal problems:


q−1
P 2n n!
1. Show that n
is convergent.
n=p n

P∞ 7n+1
2. Is n
convergent?
n=0 9

n
3. Use your knowledge of infinite series to conclude that 2n
→ 0.

 n!
4. Show that the sequence nn
is convergent. Find its limit.


P ∞
P
5. Assume that an converges and an = s. Show that
n=0 n=0

P
(a2k + a2k−1 ) converges and its sums is s.
n=0

6. Let {an } be given such that an → 0. Show that there exists a subse-

P
quence {an−k } such that the associated series ank is convergent.
k=0


P 1
7. Show that the series is convergent.
n=0 2n −n

8. Let {an }be given. assume that an > 0 for all n. Let sn denote the nth
∞ ∞ s
P P n
partial sum of the series an . Show that the series is divergent.
n=0 n=0 n

7.1.5 Keywords:
Series and Cauchy’s criteria.

87
7.1.6 Reference:
1. W. Rudin, ''Principles of Mathematical analysis'' Third edition. Me
Grow hill Education(India) Private limited, New Delhi.

2. S. Kumareson, ''Topology of Metric spaces'' Second Edition. Narosa


Publishing House. New Delhi.

88
g
Unit VIII
Power series and Algebra of series.

8.1.1 Objectives
In this unit we will be studying about power series, summation by parts, ab-
solute convergence, addition and multiplication of series, and rearrangements
of series.

8.1.2 Introduction
As we have seen in the previous unit, to ''add'' an infinite tuple, that is, to
add a sequence, {xn } of real numbers , we need analysis to give a sensible

P
meaning to '' xn ''.
n=0
Let an n-tuple (a1 , a2 , . . . , an ) of real numbers be given. Let S = a1 +
a2 + · · · + an be the sum of the finite sequence. Let σ be a permutation of
{1, 2, 3 . . . n}. Note that σ is a bijection of the set {1, 2, 3 . . . n} to itself. Con-
sider new n-tuple (aσ(1) , aσ(2) , . . . , aσ(n) ). Let t := aσ(1) + · · · + aσ(n) . Thanks
to commutativity and associativity of the addition, we know that s = t.
Given an infinite sequence (an ), assume that the associated infinite series
P∞
an is convergent. Fix N ∈ N. Let now σ be a permutation of {1, 2, . . . N }.
n=0
Construct a new sequence (bk ) where bk = aσ(k) for 1 ≤ k ≤ N and bk = ak

P
for k > N. Look at bn , the infinite series associated with the sequence
n=0

P
(bn ). Is bn convergent? and if so what is its sum? Let sn and tn denote
n=0

P ∞
P
the partial sum of the series an and bn respectively. If n = N, the
n=0 n=0
finite sequence (b1 , b2 , . . . , bn ) is just a permutation of (a1 , a2 , . . . , an ). Hence
sn = tn . Also sn = tn for n ≥ N . It follows that tn → s where s = lim sn .
n→∞
Can we extend this if σ is a permutation on N? What do we mean by
this? let (an ) be given. Let σ : N → N be a bijection. Then we construct a
P∞
new sequence (bn ) where bn := aσ(n) . Our question is if an is convergent,
n=0

89

P ∞
P
is bn convergent? If suppose bn converges, what is its sum ? Based on
n=0 n=0
our experience with algebra of finite sums, we may be tented to believe that

P P∞
the answer is bn converges to an . But this is not the case.
n=0 n=0
In this unit we study under what condition the limit of a series and its
rearranged series are same! If they are different, what are all possible attained
values?

8.1.3 Power series:


Definition 8.1.3.1 Given a sequence {cn } of complex numbers, the series

cn z n is called a power series. The numbers cn are called the coefficient’s
P
n=0
of the series; z is a complex number.

In general, the series will converge or diverge, depending on the choice of


z. More specifically, with every power series there is associated a circle, the
circle of convergence, such that the above series converges if z is in the interior
of the circle and diverges if z is in the exterior. The behavior on the circle
of convergence is much more varied and can not be described so simply.
∞ p
cn z n , put α = lim sup n |cn |,
P
Theorem 8.1.3.1 Given the power series
n=0 n→∞

1
cn z n converges if |z| < R, and diverges if |z| > R.
P
R= α
Then
n=0

Proof: Let an = cn z n , by applying root test, we obtain


p p
lim sup n |an | = |z| lim sup n |cn |,
n→∞ n→∞

|z|
= |z|α = R
.

cn z n converges if |z| < R and diverges if |z| > R.
P
Hence the series
n=0


nn z n , then
P
Example 60 1. Consider
n=0
p
α = lim sup n |cn | = lim sup n = +∞.
n→∞ n→∞

90
Hence R = 0. That is the given series converges only if z = 0 and
diverges if |z| > 0.

zn
P
2. Consider the series n!
then
n=0

1 1
R
= lim sup √
1 = 0.
n→∞ n n!

Hence the given series converges for all z.



z n . The radius of convergence of this series is
P
c) Consider the series
n=0

1
= lim sup 1 = 1,
R n→∞
we already know as it is a geometric series i. e the series converges if |z| < 1
and diverges |z| > 1.

Summation by parts:
The tests we have seen so far are for absolute convergence(Which we will
define shortly). There are infinite series which are convergent but not ab-
solutely convergent. There are often quite subtle to handle. We give some
tests which are useful to deal with such series. The basic tool for these tests
is the following Abel’s summation formula.

Theorem 8.1.3.2 (Abel’s Summation by parts) Given two sequences {an }


n
P
and {bn } put An = ak , if n ≥ 0; put A−1 = 0. Then if 0 ≤ p ≤ q, we
k=0
have.
q
P q−1
P
an b n = An (bn − bn+1 ) + Aq bq − Ap−1 bq .
n=p n=p

Proof: We have
q
P q
P
an b n = (An − An−1 )bn
n=p n=p

q
P q
P
= A n bn − An−1 bn
n=p n=p

91
q
P q−1
P
= An bn − An bn+1
n=p p−1

q−1
P
= An (bn − bn+1 ) + Aq bq − Ap−1 bq
n=p

Hence the proof.



P ∞
P
Note: It follows that an bn converges if An (bn − bn+1 ) and the
n=0 n=0
sequence {An bn+1 } converges. The following theorems are applications of
the Abel’s summation formula.
P∞
Theorem 8.1.3.3 Suppose (a) the partial sums An of an form a bounded
n=0
sequence;
(b) b0 ≥ b1 ≥ b2 ≥ . . . ; and
P∞
(c) lim bn = 0. Then an bn converges.
n→∞ n=0

Proof: Since {An } is a bounded sequence, we can choose M such that



|An | ≤ M for all n. Given  > 0, there is an integer N such that bN ≤ 2M .
For N ≤ p ≤ q, we have
q q−1
P P

an bn = An (bn − bn+1 ) + Aq bq − Ap−1 bp
n=p n=p

q−1
P
= M (bn − bn+1 ) + bq bq + bp
n=p

= 2M bp ≤ 2M bN ≤ .
q
P
i.e
an bn ≤  for all q ≥ p ≥ N.
n=p


P
Hence, by the Cauchy’s criterion for convergence of series the series an b n
n=0
convergence.

92
Theorem 8.1.3.4 (Leibnitz theorem) Suppose |c1 | ≥ |c2 | ≥ |c3 | ≥ . . . such
P∞
that c2m−1 ≥ 0 and c2m ≤ 0 for m ∈ N. Also lim cn = 0. Then cn
n→∞ n=0
converges.

Proof: Now by applying the above theorem for an = (−1)n+1 and bn = |cn |.
n
P
Then clearly An := ai is bounded and the sequence {bn } is monotonically
i=1
decreasing and converges to zero. Hence by the above theorem we obtain
P∞ ∞
P ∞
P
cn converges as cn = an b n
n=0 n=0 n=0


cn z n is 1, and
P
Theorem 8.1.3.5 Suppose the radius of convergence of
n=0

cn z n converges at every
P
suppose c0 ≥ c1 ≥ c2 ≥ . . . , lim cn = 0. Then
n→∞ n=0
point on the circle |z| = 1, except possibly at z = 1.
Proof: Let an = z n , bn = cn , then b0 ≥ b1 ≥ b2 ≥ . . . such that lim bn = 0
n→∞
P 1 − z k+1
n
n
≤ 2
P
and An := ak is a bounded sequence as |An | = =
k=1 k=1 1−z 1−z
for z 6= 1. But for |z| = 1 other than z = 1, the hypothesis of the previous

cn z n converges also on the circle |z| = 1 except
P
theorem satisfies hence
n=0
possibly at z = 1.

8.1.4 Absolute convergence



P
Definition 8.1.4.1 A series an is said to be absolutely convergent if the-
n=0

P
series |an | converges.
n=0


P ∞
P
Theorem 8.1.4.1 If an converges absolutely, then an converges.
n=0 n=0

P ∞
P
Proof: If the series an converges absolutely, then |an | converges i.e
n=0 n=0
∀  > 0 ∃ N ∈ N such that m ≥ n ≥ N we have
m
P
|ak | < .
k=n

93
By generalized triangular inequality we have
m
P m
P
| ak | ≤ |ak | <  ∀ m ≥ n ≥ N .
k=n k=n


P
Hence an converges.
n=0


P
Definition 8.1.4.2 (Conditionally convergence) We say a series an con-
n=0

P ∞
P
verges conditionally or converges non-absolutely, if an converges but |an |
n=0 n=0
diverges.
∞ (−1)n ∞ n ∞
| (−1) 1
P P P
Example 61 converges but n
| = n
diverges, hence
n=1 n n=1 n=1
q−1 n
| (−1)
P
n
| converges conditionally.
n=p

Note: We shall see that we may operate with absolutely convergent se-
ries very much as with finite sums. We may multiply them term by term
and we may change the order in which the additions are carried out, with-
out affecting the sum of the series. But for non-absolutely convergent series,
this is no longer true, and more care has to be taken when dealing with them.

Addition and multiplication of series: Following theorem tells us


that two convergent series may be added term by term and the resulting
series converges to the sum of the two series.

P ∞
P ∞
P
Theorem 8.1.4.2 If an = A and bn = B, then (an + bn ) = A + B
n=0 n=0 n=0

P
and can = cA, for any fixed c.
n=0
n
P n
P n
P
Proof: Let An := ak and Bn := bk . Then An + Bn = (ak + bk ).
k=0 k=0 k=0
Since lim An = A and lim Bn = B, we see that
n→∞ n→∞

lim (An + Bn ) = A + B.
n→∞

94

P
Hence (an + bn ) = A + B.
n=0

P
Also lim cAn = cA, we obtain can = cA.
n→∞ n=0

When we consider multiplication of series instead of defining pointwise mul-


tiplication we define it as we define multiplication of two polynomials.


P ∞
P
Definition 8.1.4.3 (Cauchy’s product) Let an and bn be two series,
n=0 n=0
n
P ∞
P
put cn = ak bn−k (n = 0, 1, 2, . . . ) and we call cn the product of the two
k=0 n=0
given series.

P ∞
P
Note: Now, the natural question anyone get is ''If an and bn are
n=0 n=0
convergent sequences, then does this imply the convergence of product of
them?''. The answer to this question is no! the following example illus-
trate that even though both the series converges the product of these series
diverges.
(−1)n P∞ (−1)n
Let an = √ , then the series √ converges as |an | ≥ |an+1 |
n+1 n=0 n+1
for all n and a2m ≥ 0 and a2m−1 ≤ 0 for m = 1, 2, 3, . . . and also lim an = 0.
n→∞
Then the series
∞    
√1 √1 √1 √ 1√ √1
P
cn = 1 − + 2
+ 2
+ + 3 2 2 3
n=0  
− √14 + √31√2 + √21√3 + √14 + + . . . ,
where
n 1
cn = (−1)n
P
p .
k=0 (n − k + 1)(k + 1)
Since
n
2 n
2 n
2
(n − k + 1)(k + 1) = 2
+1 − 2
−k ≤ 2
+1 .
We have
n 2(n + 1)
2
P
|cn | ≥ n+2
=
k=0 n+2

95
as n → ∞ lim |cn | ≥ 2. hence lim cn 6= 0 which implies that the series
n→∞ n→∞
P∞
cn doesn’t satisfies the necessary condition for convergence of the series.
n=0

P
Hence cn diverges.
n=0
.

Now we will be seeing the condition under which the Cauchy product is
convergent. the following theorem is due to Mertens, here we consider the
product of two non absolutely convergent series.

P ∞
P
Theorem 8.1.4.3 If an converges absolutely such that an = A and if
n=0 n=0

P ∞
P n
P
bn = B, then cn = AB, where ak bn−k .
n=0 n=0 k=0
q
P q
P q
P
Proof: Put An = ak , Bn = bk , C n = ck , and βn = Bn − B.
n=p n=p n=p
Then Cn = a0 b0 + (a0 b1 + a1 b0 ) + · · · + (a0 bn + a1 bn−1 + · · · + an b0 )

= a0 Bn + a1 Bn−1 + · · · + an B0

= a0 (B + βn ) + a1 (B + βn−1 ) + · · · + an (B + β0 )

= An B + a0 βn + a1 βn−1 + · · · + an β0 .

Put µn = a0 βn + a1 βn−1 + · · · + an β0 .

Now we need to prove that Cn → AB. Since An B → AB, it suffices to


show that lim µn = 0.
n→∞

P
Put α = |an |. Let  > 0 be given. Since βn → 0. We can choose N
n=0
such that |βn | ≤  for n ≥ N, in which case

|µn | ≤ |β0 an + · · · + βN an−N | + |βn+1 an−N +1 + · · · + βn a0 |

≤ |β0 an + · · · + βN an−N | + α.

96
Keeping N fixed and letting n → ∞, we get

lim sup|µn | ≤ α,


n→∞

Since ak → 0 as k → ∞ and  is arbitrary, we obtain the required result.

8.1.5 Rearrangements
Definition 8.1.5.1 Let {kn } , n = 1, 2, 3, . . . be a sequence in which every
positive integer appears once and only once (that is {kn } is a 1 − 1 function
from N to N ) putting

a0n = akn (n = 1, 2, . . . ),
∞ ∞
a0n is a rearrangement of
P P
We say that an .
n=0 n=0

Following example clearly illustrates that a rearrangement of a convergent


series need not converge to a same point

Example 62 Consider the convergent series


1 1 1 1 1
1− + − + − + ...
2 3 4 5 6
and one of its rearrangements
1 1 1 1 1 1 1 1
1+ − + + − + + − + ...
3 2 5 7 4 9 11 6
in which two positive terms are always followed by one negative term.

(−1)n n1 , then s < 1 − 21 + 1
= 56 .
P
If s = 3
n=1
Since
1 1 1
+ − >0
4k − 3 4k − 1 2k
for k ≥ 1, we see that s03 < s06 < s09 < . . . , where s0n is nth partial sum of the
rearranged series. Hence
5
lim sups0n > s03 =
n→∞ 6
so that the rearranged series doesn’t converges to s.

97
Now we will give the condition under which the rearrangement of a series
also converges to the same limit.
P
Theorem 8.1.5.1 If an is a series ofPcomplex numbers which converges
absolutely then every rearrangement of an converges, and they all con-
verges to the
P 0same sum.
Proof: Let an be a rearrangement, with partial sums s0n . Given  > 0, there
exists an integer N such that m ≥ n ≥ N implies
m
X
|ai | ≤ .
i=n

Now choose p such that the integer 1, 2, . . . , N are all contained in the set
k1 , k2 , . . . , kp . Then if n > p, the number a1 , . . . , aN will cancel in the differ-
ence sn − s0n , so that |sn − s0n | ≤ . Hence {s0n } converges to the same sum as
that of the sequence {sn }. Hence the proof of the theory.

Now we will be seeing a theorem due to Riemann, which helps us to prove


that there exist a convergent series of real number, whose rearrangement
converges to any given real number.
P
Theorem 8.1.5.2 Let an be a series of real numbers which converges, but
not absolutely. Suppose
−∞ ≤ α ≤ β ≤ ∞.
Then there exists a rearrangement a0n with partial sums s0n such that lim inf s0n =
P
n∞
α, and lim sup s0n = β. → (1)
n→∞

Proof: Let
|an | + an |an | − an
pn = , qn = (n = 1, 2, 3, . . . )
2 2

P ∞
P
Then pn − qn = an , pn + qn = |an |, pn ≥ 0, qn ≥ 0. The series pn , qn
n=0 n=0
must both diverges.

For if both were convergent, then



P ∞
P
(pn + qn ) = |an |
n=0 n=0

98
would converge, contrary to hypothesis. since
N
P N
P N
P N
P
an = (pn − qn ) = pn − qn ,
n=1 n=1 n=1 n=1

P ∞
P
divergence of pn and convergence of qn (or vice versa) implies diver-
n=0 n=0

P
gence of an , again contrary to hypothesis.
n=0

P
Now let P1 , P2 , P3 , . . . denote the non negative terms of an , in the order in
n=0
which they occur and let Q1 , Q2 , Q3 , . . . be the absolute values of the negative

P
terms of an , also in their original order.
n=0

P ∞
P ∞
P ∞
P
The series Pn , Qn differ from pn , qn only by zero terms,
n=0 n=0 n=0 n=0
and therefore divergent.
We shall construct sequences {mn }, {kn } such that the series
P1 +P2 +· · ·+Pm1 −Q1 −· · ·−Qk1 +Pm1 +1 +· · ·+Pm2 −Qk1 +1 −· · ·−Qk2 +. . . ,
→ (2)

P
which clearly is a rearrangement of an , satisfies (1).
n=0

Choose real-valued sequences {αn }, {βn } such that αn → α, βn → β,


αn < βn ,β1 > 0.

Let m1 , k1 be the smallest integers such that


P1 + · · · + Pm1 > β1 ,

P1 + · · · + Pm1 − Q1 − · · · − Qk1 < α1 ;

let m2 , k2 be the smallest integers such that


P1 + · · · + Pm1 − Q1 − . . . Qk1 + Pm1 +1 + · · · + Pm2 > β2 ,

P1 + · · · + Pm1 − Q1 − . . . Qk1 + Pm1 +1 + · · · + Pm2 − Qk1 +1 − · · · − Qk2 < α2 ;



P ∞
P
and continue in this way. This is possible since Pn and Qn diverge.
n=0 n=0
If xn , yn denote the partial sums of (2) whose last terms are Pmn ,−Qkn , then

99
|xn − βn | ≥ Pmn , |yn − αn | ≥ Qkn

Since Pn → 0 and Qn → 0 as n → ∞ , we see that xn → β, yn → α.

Finally, it is clear that no number less than α or greater than β can be


subsequent limit of the partial sums of (2).

8.1.6 Terminal problems:


1. Discuss convergene or divergence of the following series.

1
P
(a) 2n
n=1

1
P
(b) 2n−1
n=1

2
P
(c) n2 +3
n=0
P∞ √ √
(d) ( n + 1 − n)
n=0
∞ √ √
P n+1− n
(e) n
n=1
∞ √
( n n − 1)n
P
(f)
n=1

2. Find the radius of convergence of the following series.



n3 z n
P
(a)
n=0

2n n
P
(b) n!
z
n=0

2n n
P
(c) n2
z
n=1

n3 n
P
(d) 3n
z
n=0

z n!
P
(e)
n=0

100

2
zn .
P
(f)
n=0

∞ ∞
a2n is
P P
3. Let an be a convergent series of positive terms. Show that
n=0 n=0

apn is convergent for p > 1.
P
convergent. More generally, show that
n=0

∞ ∞
1 1 π2
P P
4. Compute n2 (2n−1)
, assuming that n2
= 6
.
n=1 n=1

∞ ∞
P (−1)n−1 P 1 π4
5. Compute n4
, assuming that n4
= 90
.
n=1 n=1


an
P
6. Let a > 0. Show that the series 1 is convergent if a < 1.
n=1 (n!) n


1
P 
7. Show that the series log 1 + n
is divergent.
n=1


1
P
8. Show that the series (n+2) log(n+2)
diverges.
n=0


P
9. If an converges, and if {bn } is monotonic and bounded, prove that
n=0

P
an bn converges.
n=0


P ∞
P
10. Let an and bn be convergent series of positive terms. Show that
n=0 n=0
∞ √
P
an bn is convergent.
n=0

8.1.7 Keywords:
Power series, radius of convergence, absolute convergence, conditional con-
vergence and rearrangement of series.

101
8.1.8 References:
1. W. Rudin, ''Principles of Mathematical analysis'' Third edition. Me
Grow hill Education(India) Private limited, New Delhi.

2. S. Kumareson, ''Topology of Metric spaces'' Second Edition. Narosa


Publishing House. New Delhi.

102
BLOCK III
Unit IX
Limit and continuity

9.1.1 Objectives:
We will be studying about limit of a function and continuity of a function in a
general context like metric space, and also particular cases like vector-valued
function.

9.1.2 Introduction:
In our undergraduate mathematics, we are already familiar with concepts as
introduced in elementary calculus, where we have studied about limit of a
function, indicated by notation such as
lim f (x) = A,
x→p

which means that for every  > 0 there is another number δ > 0 such that
|f (x) − A| <  whenever 0 < |x − p| < δ. This convey the idea that f (x) can
be made arbitrarily close to A by taking x sufficiently close to p. Applications
of calculus to geometrical and physical problem in 3D-space and to functions
of several variable make it necessary to extend these concepts to Rn . It is
just as easy to go one step further and introduce limits in the more general
setting of metric spaces. This achieves a simplification in the theory by
stripping it of unnecessary restrictions and at the same time covers nearly
all the important aspects needed in analysis.
First we discuss limit of a functions in metric space, vector valued function
and continuity.

9.1.3 Limit of a function:


In this section we consider two metric spaces (S, dS ) and (T, dT ), where dS
and dT denote the respective metrics.

Definition 9.1.3.1 Let (S, dS ) and (T, dT ) be metic space. Let A be a subset
of S and let f : A → T be a function from A to T . If 'p' is a limit point of
the set A and if b ∈ T , the notation

103
lim f (x) = b,
x→p

is defined to mean the following: for every  > 0, there is a δ > 0 such
that dT (f (x), b) <  whenever x ∈ A, x 6= p and dS (x, p) < δ.

Note:

1. The symbol lim f (x) = b is read as ''The limit of f (x), as x tends to p,


x→p
is b'' or '' f(x) approaches p as x approaches p''. We sometimes indicate
this by writing f (x) → b as x → p.

2. The definition conveys the intuitive idea that f (x) can be made arbi-
trarily close to b by taking x sufficiently close to p.

3. We require that p be a limit point of A to make sure that there will be


points x in A sufficiently close tp p, with x 6= p. Moreover, p need not
be in the domain of f , and b need not be in the range of f .

Remark:
If X and/or, Y are replaced by the real line, or the complex plane or by
some euclidean space Rk , the distances dx , dy are of course replaced by
absolute values, or by norms of differences.

The next theorem relates limits of functions to limits of convergent sequences.

Theorem 9.1.3.1 Let (X, dX ), (Y, dY ) be metric space, E ⊆ X. Let p be a


limit point of E and f be a function from E to Y , then

lim f (x) = b → (1)


x→p

if and only if

lim f (xn ) = b → (2)


n→∞

104
for every sequence {xn } of points in E − {p} which converges to p.

Proof: If suppose lim f (x) = b, then for every  > 0 there is a δ > 0 such
x→p
that dY (f (x), b) <  whenever x ∈ E and 0 < dX (x, p) < δ. Now take any
sequence {xn } in E − {p} which converges to p. For the δ > 0 there exist
N ∈ N such that dX (xn , p) < δ for all n ≥ N. Hence dY (f (xn ), b) <  for
all n ≥ N, and hence {f (xn )} converges to b, therefore (1) implies (2). To
prove the converse, we assume (2) holds and that (1) is false and arrive at a
contradiction.
If (1) is false, then for some  > 0 and for every δ > 0 there is a point x in
A such that 0 < dX (x, p) < δ but dY (f (x), b) ≥ .
Taking δ = n1 , where n = 1, 2, 3, ... this means there is a corresponding se-
quence of points {xn } in A−{p} such that 0 < dX (xn , p) < n1 but dY (f (xn ), b) ≥
. clearly, this sequence {xn } converges to p but the sequence {f (xn )} doesn’t
converges to b, contradiction to (2).

Corollary 9.1.3.1 If the function f has a limit at p, then limit is unique.

Limits of complex valued functions

Let (X, dX ) be a metric space, let A ⊆ X and f is a function f : A → C.


Consider two complex valued functions f and g defined on A, f : A → C,
g : A → C. Then by using the property of addition and multiplication of
complex numbers to define addition and multiplication of complex valued
functions. The sum f + g : A → C is defined as (f + g)(x) = f (x) + g(x).
f
The difference f − g, the product f.g and the quotient (where g(x) 6= 0
g
∀ x ∈ A) are similarly defined. Next theorem gives us the usual rule for
f
calculating the limits of f ± g, f.g and .
g

Theorem 9.1.3.2 Suppose E ⊆ X, a metric space, p is a limit point of E,


f and g are complex functions on E, and

lim f (x) = A, lim g(x) = B.


x→p x→p

Then
(a)lim f (x) ± g(x) = A ± B
x→p

105
(b)lim f (x)g(x) = AB
x→p
f (x) A
(c)lim =
x→p g(x) B
Proof: (a) Since lim f (x) = A and lim g(x) = B for given  > 0, there
x→p x→p
exists δ1 > 0 and δ2 > 0 such that |f (x) − A| < 2 whenever 0 < d(x, p) < δ1
and x ∈ E and |g(x) − B| < 2 whenever 0 < d(x, p) < δ2 and x ∈ E. Choose
δ = min {δ1 , δ2 }, then for every x in E with 0 < d(x, p) < , we have
|(f + g)(x) − (A + B)| = |f (x) − A + g(x) − B| ≤ |f (x) − A| + |g(x) − B| <

2
+ 2 = . Since,  > 0 is arbitrary, we have
lim (f + g)(x) = A + B.
x→p

Similarly lim (f − g)(x) = A − B.


x→p
Proof of (b): Given with 0 <  < 1, choose ' such that 0 < 0 < 1 and

0 = . There is a δ > 0 such that if x ∈ E and 0 < d(x, p) < δ,
(|A| + |B| + 1)
then |f (x) − A| < 0 and |g(x) − B| < 0 .
Then |f (x)| = |A + (f (x) − A)| ≤ |A| + 0 < |A| + 1.
Writing f (x)g(x) − AB = f (x)g(x) − Bf (x) + Bf (x) − AB, we have

|f (x)g(x) − AB| = |f (x)g(x) − Bf (x)| + |Bf (x) − AB| = |f (x)||g(x) −


B| + |B||f (x) − A| < (|A| + 1)0 + |B|0 = 0 (|A| + |B| + 1) = .
Thus |f (x)g(x) − AB| <  whenever x ∈ E and 0 < d(x, p) < δ and this
proves (b).

Proof of (c). For any given  > 0 with |B| −  > 0(This is possible as
|B| =
6 0) there exists δ > 0, such that whenever x in E with 0 < d(x, p) < δ
we have |f (x) − A| < 1 and |g(x) − B| < 2 ,where 1 and 2 will be dependent
on  and will be defined later.
Since ||g(x)| − |B|| < |g(x) − B| < .

we obtain, |B| −  < |g(x)| < |B| + .

Let M = |B| −  > 0, then M < |g(x)| for all x in E with 0 < d(x, p) < δ.

f (x) A |f (x)B − Ag(x)|
Consider,
− =
g(x) B |g(x)||B|

106
|f (x)B − AB + AB − Ag(x)|
=
|g(x)||B|
|B||f (x) − A| + |A||g(x) − B|

|g(x)||B|
|f (x) − A| |A|
≤ + |g(x) − B|
|g(x)| |B|
1 |A|
< 1 + 2 .
M |B|
M |B|‘
By taking 1 = > 0 and 2 =
2 2(|A| + 1)

f (x) A
We obtain, − < .
g(x) B
f (x) A
Hence lim = .
x→p g(x) B

Limits of vector valued functions: Let (S, d) be a metric space and let
A be a subset of S. Consider two vector valued functions f and g on A i.e
f : A → Rk and g : A → Rk .
Quotients of vector valued functions are not defined if (k > 2), but we can
define the sum f ± g, the product λf (if λisreal) and the inner product f.g
by the respective formulas.

(f ± g)(x) = f (x) ± g(x), (λf )(x) = λf (x), (f.g)(x) = f (x).g(x)

for each x in A. We then have the following rule for calculating with limits
of vector-valued functions.

Theorem 9.1.3.3 Let 'p' be an limit point of A and assume that lim f (x) =
x→p
a, lim g(x) = b. Then
x→p
1) lim (f (x) + g(x)) = a + b
x→p
2)lim λf (x) = λa for every real number λ.
x→p
3) lim f (x).g(x) = a.b
x→p
4) lim ||f (x)|| = ||a||
x→p

107
Proof of 1): Since lim f (x) = a and lim g(x) = b, For a given  > 0,
x→p x→p
 
there exist δ > 0 such that ||f (x) − a|| < 2
and ||g(x) − b|| < 2
for each x in
A with 0 < d(x, p) < δ. Hence

||(f (x) + g(x)) − (a + b)|| = ||f (x) − a + g(x) − b||

≤ ||f (x) − a|| + ||g(x) − b||

<

for each x in A with 0 < d(x, p) < δ.

Proof of (2) Since lim f (x) = a, and λ ∈ R.


x→p
If λ = 0 then λf (x) = 0 ∀ x ∈ A, hence the result holds true vacuously.
6 0, then for given any  > 0, there exists δ > 0 such that ||f (x) − a|| <
If λ =

|λ|
for each x in A with 0 < d(x, p) < δ.

Hence ||λf (x) − λa|| = |λ|||f (x) − a|| <  for each x in A with 0 <
d(x, p) < δ.

Hence lim λf (x) = λa.


x→p

Proof of (3) We have

f (x)g(x) − a.b = [f (x) − a][g(x) − b] + a[g(x) − b] + b[f (x) − a].

By the triangular inequality and the Cauchy-Schwarz inequality, we have

||f (x).g(x) − a.b|| ≤ ||f (x) − a||||g(x) − b|| + ||a||||g(x) − b|| + ||b||||f (x) − a||.

as x → p, f (x) → a and g(x) → b, hence ||f (x)−a|| → 0 and ||g(x)−b|| → 0.


Thus we obtain

||f (x).g(x) − a.b|| → 0 as x → p.

i.e lim f (x).g(x) = a.b


x→p

108
Note here (.) represents the inner product of two vectors.

Proof of (4) Since lim f (x) = a, for given  > 0, there exists δ > 0 such
x→p
that ||f (x) − a|| <  whenever x in A with 0 < d(x, p) < δ.
We have
|||f (x)|| − ||a||| ≤ ||f (x) − a|| < 
for each x in A with 0 < d(x, p) < δ. Thus lim ||f (x)|| = ||a||.
x→p

9.1.4 Continuous functions:


We now know the sequential definition of a existence of limit point. Let {xn }
be a convergent sequence say lim xn = x, in the domain of a function f .
n→∞  
Then can we write lim f (xn ) = f lim xn ?
n→∞ n→∞
Notice that, not every functions satisfies this property.
Define f : R → R as
(
0 if (−∞, 0]
f(x)=
1 if (0, ∞)

and let xn = n1 , then we know that lim xn = 0. Where as lim f (xn ) = 1


  n→∞ n→∞
1
but f (0) = f lim n = 0 Hence lim f (xn ) 6= f ( lim xn ), but continuous
n→∞ n→∞ n→∞
functions satisfies this required relation i.e interchanging limit in function.

Now, the definition of continuity presented in elementary calculus can be


extended to functions from one metric space to another.

Definition 9.1.4.1 Let (X, dX ) and (Y, dY ) be metric space and let f : X →
Y be a function. The function f is said to be continuous at a point 0 P 0 in
X if for every  > 0 there is a δ > 0 such that dY (f (x), f (p)) <  whenever
dX (x, p) < δ. If f is continuous at every point of a subset E of X, we say f
is continuous on E.

This definition reflects the intuitive idea that points close to p are mapped
by f into points close to f (p).

109
Alternative definition :
Let (X, dX ) and (Y, dY ) be metric spaces and let f : X → Y be a function.
The function f is said to be continuous at a point p in X if ∀  > 0, there is
a δ > 0 such that

f (Nδ (p)) ⊆ N (f (p))

Here Nδ (p) is a δ−neighborhood of p in X, its image under f must be


contained in the −neighborhood of f (p), N (f (p)) in Y .

Note:

1. If 'p' is an accumulation point of the set E, then the definition of


continuity implies that

lim f (x) = f (p).


x→p

2. If 0 p0 is an isolated point of the set E, then every f defined at 0 p0 will


be continuous at p because for sufficiently small δ > 0 there is only one
point satisfying dX (x, p) < δ, namely x = p, and dY (f (p), f (p)) = 0.

Following theorem is often described by saying that for continuous func-


tions the limit symbol can be interchanged with the function symbol.

Theorem 9.1.4.1 Let f : X → Y be a function from metric space (X, dX )


to another (Y, dY ) and assume p ∈ X. Then f is continuous at p if and only
if for every sequence {xn } in X converging to p, the sequence {f (xn )} in Y
converges to f (p); in symbols,
 
lim f (xn ) = f lim xn .
n→∞ n→∞

Proof: Assume f is continuous at p. Let {xn } be a sequence in X such that


lim xn = p, then we need to show that f (xn ) → f (p). Let  > 0 be given
n→∞
then there exist δ > 0 such that dY (f (x), f (p)) <  whenever d(x, p) < δ.
Since lim xn = p there exist N ∈ N such that d(xn , p) < δ for each n ≥ N .
n→∞
Hence

110
d(f (xn , f (p))) <  for each n ≥ N .

Thus f (xn ) → f (p).

Conversely, suppose for every sequence {xn } in X convergent to p, the


sequence {f (xn )} in Y converges to f (p). Then we need to prove that f
is continuous at p. If suppose f is not continuous at p, then there exist
 > 0 such that for each n1 > 0, there exist xn in X with d(xn , p) < n1 but
d(f (xn ), f (P )) ≥ , which implies xn → p but {f (xn )} doesn’t converges to
f (p). A contradiction to our assumption. Hence f has to be continuous at
p.

Continuity of composite functions:

We now turn to composition of functions. A brief statement of the fol-


lowing theorem is that a continuous function of a continuous function is
continuous.

Theorem 9.1.4.2 Suppose X, Y, Z are metric spaces, E ⊆ X, f maps E


into Y , g maps the range of f , i.e g maps f (E) into Z and h is the mapping
of E into Z defined by

h(x) = g(f (x)) for each x ∈ E.

If f is continuous at a point p ∈ E and if g is continuous at the point f (p),


then h is continuous at p.

Proof: Let  > 0 be given. Since g is continuous at f (p), there exists


η > 0 such that

dZ (g(y), g(f (p))) <  if dY (y, f (p)) < η and y ∈ f (E).

Since f is continuous at p, there exists δ > 0 such that

dY (f (x), f (p)) < η if dX (x, p) < δ and x ∈ E.

It follows that

dZ (h(x), h(p)) = dZ (g(f (x)), g(f (p))) < 

111
if dX (x, p) < δ and x ∈ E. Thus 0 h0 is continuous at p.

Continuous complex-valued and vector-valued functions:

We now turn to complex-valued, vector-valued functions and to functions


defined on subset of Rk

Theorem 9.1.4.3 Let f and g be complex-valued functions continuous at a


point 0 p0 in a metric space (X, d). Then f + g, f − g and f.g are each con-
tinuous at p. The quotient f /g is also continuous at p if g(p) 6= 0.

Proof: If p is an isolated point, then the reset is trivial. If p is a limit


point of X, we obtain the result from the Theorem 9.1.3.2.

Theorem 9.1.4.4 Let f and g be functions continuous at a point 0 p0 in a


metric space (X, d) and assume that f and g have values in Rn . Then each
of the following is continuous at p; the sum f + g, the product λf for every
real λ, the inner product f.g, and the norm ||f ||.

Proof: Proof follows directly from the corresponding theorem on limit of


vector valued function.

Theorem 9.1.4.5 Let f1 , f2 , . . . , fk be real valued functions on a metric


space X, and let f be the mapping of X into Rk defined by

f (x) = (f1 (x), f2 (x), . . . , fk (x)) for each x ∈ X;

then f is continuous if and only if each of the functions f1 , f2 , . . . , fk is con-


tinuous.

Proof: Let p be a point of X, since each fi 1 ≤ i ≤ k is continuous at


p. For any given  > 0, there exists δi > 0 for each 1 ≤ i ≤ k such that
|fi (x) − fi (p)| < √k whenever d(x, p) < δi and x ∈ X.

For δ := min {δ1 , δ2 , . . . , δk } and for each x in X with d(x, p) < δ, we


have |fi (x) − fi (p)| < √k for each 1 ≤ i ≤ k.

Since

112
 k
 12
2
P
|f (x) − f (p)| = |fi (x) − fi (p)|
i=1

 k
 12
2
P
≤ k
i=1

= for each x in X with d(x, p) < δ.

Hence f is continuous at p.

Example 63 Let x = (x1 , x2 , . . . , xk ) ∈ Rk , the function φi defined by


φi (x) = xi ∀ x ∈ Rk are continuous on Rk , since the inequality
 k
 12
2
P
|φi (x) − φi (y)| = |xi − yi | ≥ |xi − yi | = |x − y|
i=1

shows that we may take δ = . φ are sometimes called the co-ordinate


'' i

functions''.

Example 64 Repeated application of the previous theorem, then shows that


every monomial xn1 1 xn2 2 . . . xnk k where n1 , n2 , . . . , nk are non-negative integers,
is continuous on Rk . Since constant multiple of continuous function is con-
tinuous

P (x) = Cn1 n2 ...nk xn1 1 xn2 2 . . . xnk k (x ∈ Rk ),


P

is continuous on Rk . Here Cn1 n2 ...nk are complex numbers and the sum has
finitely many terms.

Example 65 Since real valued functions of elementary calculus, such that as


the exponential, trigonometric, and logarithmic functions, are all continuous
wherever they are defined. For example

lim ex = e0 = 1.
x→0

113
9.1.5 Continuity and inverse images of open or closed set.
The concept of inverse image can be used to give two important global de-
scriptions of continuous function.
Definition 9.1.5.1 Let f : X → Y be a function from a set X to a set Y .
If E is a subset of Y , the inverse image of E under f , denoted by f −1 (E),
is defined to be the largest subset of X which f maps into E; that is
f −1 (E) = {x : x ∈ X and f (x) ∈ E}
The following theorem gives a very useful characterization of continuity.
Theorem 9.1.5.1 A mapping f of a metric space X into a metric space Y
is continuous on X if and only if f −1 (V ) is open in X for every open set V
in Y .

Proof:Let f be continuous on X, let V be an open set in Y , and let 0 p0 be


any point of f −1 (V ). We will prove that p is an interior point of f −1 (V ). Let
y = f (p). Since V is open and y ∈ V , there exist  > 0 such that N (y) ⊆ V .
Since f is continuous at p, there is a δ > 0 such that f (Nδ (p)) ⊆ N (y) ⊆ V .
Hence Nδ (p) ⊆ f −1 [f (Nδ (p))] ⊆ f −1 [N (y)] ⊆ f −1 (V ), so p is an interior
point. Hence f −1 (V ) is an open set.

Conversely, assume that f −1 (V ) is open in X for every open set V in


Y . Choose p in X and let y = f (p). We will prove f is continuous at
p. For every  > 0, the −neighborhood N (y) is an open set in Y , so
f −1 (N (y)) is open set in X. Now p ∈ f −1 (N (y)) so there is a δ > 0 such
that Nδ (p) ⊆ f −1 (N(f (p)) ), therefore f (Nδ(p) ) ⊆ N (f (p)) so f is continuous
at p.

Corollary 9.1.5.1 A mapping f of a metric space X into a metric space Y


is continuous if and only if f −1 (V ) is closed in X for every closed set V in Y .

Proof: Since if V is closed in Y .then V c = Y − V is open in Y and


f (Y − V ) = X − f −1 (V )
−1

Now by applying the above theorem we obtain the required result.


Note:

114
1. The image of an open set under a continuous mapping is not necessarily
open. For example, let (X, d) be any metric space, define f : X → R
by f (x) = 1 for each x ∈ X, then clearly f is a continuous function
but f (X) = {1}, X is open where as {1} is not open in R.

2. The image of a closed set under a continuous mapping need not be


closed. For example f : R → R defined by f (x) = arctan x, then
f (R) = − 2 , 2 . R is closed where as − π2 , π2 is not closed.
π π


9.1.6 Terminal problems:


1. Suppose f is a real function defined on R which satisfies lim [f (x + h) −
h→0
f (x)] = 0 for every x ∈ R. Does this imply that f is continuous ?

2. Let f be a continuous real function on a metric space X. Let Z(f ) :=


{p ∈ X : f (p) = 0}. Prove that Z(f ) is a closed set.

3. Let f and g be continuous mappings of a metric space X into metric


space Y and let E be a dense subset of X. i.e (E = X) Prove that f (E)
is dense in f (X). If g(p) = f (p) for all p ∈ E, prove that g(p) = f (p)
for all p ∈ X.

4. Determine for each of the following functions defined on R2 , weather


the limits lim [lim f (x, y)], lim [lim f (x, y)] and lim f (x, y) exists
x→0 y→0 y→0 x→0 (x,y)→(0,0)
and evaluate those limits that do exist:
x2 − y 2
(a) f (x, y) = if (x, y) 6= (0, 0), f (0, 0) = 0.
x2 + y 2
(xy)2
(b) f (x, y) = if (x, y) 6= (0, 0), f (0, 0) = 0.
(xy)2 + (x − y)2
1
(c) f (x, y) = sin(xy) if x 6= 0, f (0, y) = y.
x
(
(x + y) sin( x1 ) sin( y1 ) if x 6= 0 and y 6= 0
(d) f (x, y) =
0 if x = 0 or y = 0

115
 sin x − sin y

if tan x =6= tan y
(e) f (x, y) = tan x − tan y
 3
cos x tan x = tan y

5. Let f be continuous on [a, b] and let f (x) = 0 when x is rational. Prove


that f (x) = 0 for every x in [a, b].

6. Let f, g and h be defined on [0, 1] as follows: f (x) = g(x) = h(x) = 0,


whenever x is irrational; f (x) = 1 and g(x) = x, whenever x is ratio-
nal h(x) = n1 , if x is the rational number mn
(in lowest term) h(0) = 1.
Prove that f is not continuous anywhere in [0, 1], that h is continuous
only at the irrational points in [0, 1].

7. For each x in [0, 1], let f (x) = x if x is rational and let f (x) = 1 − x if
x is irrational. Prove that

(a) f (f (x)) = x for all x in [0, 1].

(b) f is continuous only at the point x = 12 .

(c) f assumes every value between 0 and 1.

9.1.7 Keywords:
Limits, continuity and inverse image.

9.1.8 References:
1. W. Rudin, ''Principles of Mathematical analysis'' Third edition. Me
Grow hill Education(India) Private limited, New Delhi.

2. Tom. M. Apostol, '' Mathematical analysis '' Second Edition. Narosa


Publishing House. New Delhi.

116
g
Unit X
Continuity and compactness

10.1.1 Objectives
In this unit, we will be studying about the global properties of continuous
functions on a compact set.

10.1.2 Introduction
One, after studying about the continuous functions on a metric space gets
numerous questions such as. Is every continuous function bounded ? Is
every continuous function uniformly continuous ? When does an image of
a open/closed set under a continuous function open/closed? In this unit we
will be answering all these questions.

10.1.3 Functions continuous on compact sets.


The following theorem shows that the continuous image of a compact set is
compact. This is another global property of continuous functions.

Theorem 10.1.3.1 Suppose f : X → Y is a continuous mapping of a com-


pact metric space X. Then f (X) is compact subset of Y .

Proof: To prove f (X) is a compact subset of Y . Consider an open cover


{Vα } of f (X). Since f is continuous, we know that {f −1 (Vα )} is an open
cover for the set X. Since X is compact set there exist finitely many indices,
say α1 , α2 , . . . , αn , such that

X ⊆ f −1 (Vα1 ) ∪ · · · ∪ f −1 (Vαn ).

Since f (f −1 (E)) ⊆ E for every E ⊆ Y , we obtain

f (X) ⊆ Vα1 ∪ . . . Vαn .

Hence f (X) is a compact subset of Y .

The above theorem can also be stated as follows

117
Theorem 10.1.3.2 Let f : X → Y be a function from metric space X to
Y . If f is continuous on a compact subset E of X, then f (E) is a compact
subset of Y .

Definition 10.1.3.1 A function f : S → Rk is called bounded on S if there


is a positive number M such that ||f (x)|| ≤ M for all x in S.

Theorem 10.1.3.3 Let f : X → Rk be a continuous mapping of a compact


metric space X in to Rk , then f (X) is closed and bounded. Thus f is bounded.

Proof: Since X is a compact metric space and f is a continuous map-


ping, from the previous theorem we have f (X) is a compact subset of Rk .
Since every compact subset of Rk is closed and bounded, f (X) is closed and
bounded. Hence f is bounded.

This theorem has important implications for real-valued functions. If f


is real-valued and bounded on X, then f (X) is a bounded subset of R, so it
has a supremum, sup f (X), and an infimum, inf f (X). Moreover,

inf f (X) ≤ f (x) ≤ sup f (X) for every x in X.

The next theorem shows that a continuous function f actually takes the
values sup f (X) and inf f (X) if X is compact.

Theorem 10.1.3.4 If f : X → R is a continuous real valued function on a


compact metric space X, and

M = sup f (x),m = inf f (x).


x∈X x∈X

Then there exists points p, q ∈ X such that f (p) = M and f (q) = m.

Proof: Since continuous image of a compact set is compact, f (X) is a


compact subset of R. Hence f (X) is closed and bounded subset of R. Since
m = inf f (X), and M = sup f (X) are limit point of f (X) and f (X) is
x∈X x∈X
closed, implies m ∈ f (X) and M ∈ f (X), which again implies that there
exists p and q in X such that f (q) = m and f (p) = M .

118
Note: Since f (q) ≤ f (x) ≤ f (p) for all x in X, the number f (p) and f (q)
are called absolute or global maximum and minimum values respectively of
f on X.

Following theorem answers our question that ''when does continuous im-
age of an open/closed set is open/closed ?'' under certain cases.

Theorem 10.1.3.5 Suppose f is a continuous one-to-one mapping from a


compact metric space X on to a metric space Y . Then the inverse mapping
f −1 defined on Y by

f −1 (f (x)) = x ∀ x ∈ X

is a continuous mapping of Y onto X.

Proof: As f is a bijective function, f −1 is well defined. Now to prove


that f −1 is a continuous function it is enough to prove that inverse image of
open set in X under f −1 is open i.e we need to prove that f (V ) is an open
set in Y for every open set V in X.

Let V be an open set in X, the complement V c is closed in X. Since X


is compact and a closed subset of a compact set is compact, V c is compact
subset of X. Hence f (V c ) is a compact subset of Y and so it is closed in Y .
Since f is bijective f (V ) = [f (V c )]c . Hence f (V ) is open. Hence the proof.

Example 66 This example shows that compactness of X is an essential part


of the above theorem. Let X = [0, 1) with the usual metric on R and consider
the complex-valued function f defined by

f (x) = e2πix for 0 ≤ x < 1.

This is clearly a continuous mapping and one-to-one and onto unit circle
|z| = 1 in the complex plane. However, f −1 is not continuous at the point
f (0). For example, if xn = 1 − n1 , the sequence {f (xn )} converges to f (0)
but {xn } doesn’t converges in X = [0, 1).

119
Topological mappings (Homeomorphisms).

Definition 10.1.3.2 Let f : X → Y be a function from one metric space


(X, dX ) to (Y, dY ). Assume also that f is one-to-one and onto, so that the
inverse function f −1 exists. If f is continuous on X and f −1 is continuous
on Y , then f is called a topological mapping or a homeomorphism and the
metric space X and Y are called homeomorphic.

Note:

1. The above theorem gives us a wonderful criterion that, if X is a com-


pact set then a bijective continuous function of X is homeomorphic.

2. A property of a set which remains invariant under every topological


property mappings is called a topological property. Thus the property
of being open, closed and compact are topological properties.

3. An important example of a homeomorphism is an isometry. A function


f : X → Y which is one-to-one on X and which preserves the metric,
that is

dY (f (x), f (y)) = dX (x, y)

for all points x and y in X, is called an isometry.

Bolzano’s theorem

In this section, we will be studying about the famous theorem of Bolzano


which concerns a global property of real-valued functions continuous on com-
pact intervals [a, b] in R. If the graph of f lies above the x − axis at a and
below the x − axis at b Bolzano’s theorem asserts that the graph must cross
the x-axis somewhere in between.

120
Theorem 10.1.3.6 Let f be defined on an interval S in R. Assume that f is
continuous at a point in S and that f (c) 6= 0. Then there is a δ−neighborhood
of c say Nδ (c) such that f (x) has the same sign as f (c) in Nδ ∩ S.

Proof: Assume, without loss of generality that f (c) > 0. For every  > 0
there is a δ > 0 such that

|f (x) − f (c)| <  whenever x ∈ Nδ ∩ S.

For  = f (c) 2
> 0, then there exist δ corresponding to this  such that
|f (x) − f (c)| < f (c)
2
for each x in Nδ (x) ∩ S. i.e −f2(c) < f (x) − f (c) < f (c)
2
for each x in Nδ (x) ∩ S.
i.e f (c)
2
< f (x) < 3f2(c) whenever x in Nδ ∩ S.
So f (x) > 0 for each x in Nδ ∩ S.

So f (x) > 0 for each x in Nδ (x)∩S. Proof is similar for the case f (x) < 0.
Hence the result holds.

Theorem 10.1.3.7 (Bolzano) Let f be real-valued and continuous on a com-


pact interval [a, b] in R, and suppose that f (a) and f (b) have opposite signs;
that is, assume f (a)f (b) < 0. Then there is at least one point c in the open
interval (a, b) such that f (c) = 0.

Proof: For definiteness, let us assume f (a) > 0 and f (b) < 0. Let
A = {x : x ∈ [a, b] and f (x) ≥ 0}. Then A is non-empty since a ∈ A, and A
is bounded above by b. Let c = sup A. Then a < c < b. We will prove that
f (c) = 0.

If f (c) 6= 0, then there exist Nδ (c) in which f has the same sign as f (c).
If f (c) > 0, there are points x > c at which f (x) > 0, contradicting the
definition of c. If f (c) < 0, then c − 2δ is an upper bound for A, again
contradicting the definition of c. Therefore we must have f (c) = 0.

From Bolzano’s theorem we can easily deduce the intermediate value the-
orem for continuous functions.

Theorem 10.1.3.8 Assume f is real-valued and continuous on a compact


interval S in R. Suppose there are two points α < β in S such that f (α) 6=

121
f (β). Then f takes every value between f (α) and f (β) in the interval (α, β).

Proof: Let k be a number between f (α) and f (β). Now define g(x) =
f (x) − k, then g(x) < 0 and g(β) > 0. Hence by Bolzano’s theorem there
exists a point p in (α, β) such that g(p) = 0 i. e f (p) = k.
Hence the proof of intermediate value theorem.

Note: Continuous image of a compact interval S under a real-valued


function is f (S) = [inf f (S), sup f (S)]

10.1.4 Uniform Continuity


Suppose f is defined on a metric space (X, dX ) with values in another metric
space (Y, dY ) and assume that f is continuous on X. Then for given any
point p in X and any  > 0, there is a δ > 0 such that, if x ∈ A, then

dY (f (x), f (p)) <  whenever dX (x, p) < δ.

In general we cannot expect that for a fixed  the same value of δ will serve
for EVERY POINT in X. If a function satisfies this property, we call such a
function as uniformly continuous on X.

Definition 10.1.4.1 Let f : X → Y be a function from one metric space


(X, dX ) to (Y, dY ). Then f is said to be uniformly continuous on X if the
following condition holds:

For every  > 0, ∃ δ > 0 such that dY (f (x), f (y)) <  for all x, y ∈ X
with dX (x, y) < δ.

The following examples helps us to clearly understand the difference be-


tween continuous and uniformly continuous function.

Example 67 f : (0, 1] → R, defined by f (x) = x1 . This function is continu-


ous but not uniformly continuous. Let  = 10, suppose we should find a δ > 0
such that 0 < δ < 1, satisfies the definition of uniform continuity. Taking
δ
x = δ, p = 11 , we obtain |x − p| = 10δ
11
< δ and |f (x) − f (p)| = | 1δ − 11
δ
|=
10
δ
> 10. Hence the given function is not uniformly continuous.

122
Note: Every uniformly continuous function is continuous. The converse
need not be true always (as we have seen in the above example!). But the
following theorem gives us a condition, under which the converse holds
true: i.e if a function f is continuous and something happens then f is
uniformly continuous.

Theorem 10.1.4.1 Let f be a continuous mapping of a compact metric


space X into a metric space Y . Then f is uniformly continuous on X.

Proof: Let  > 0 be given. Since f is continuous, for each point p in X,


there exist δp > 0 such that

dY (f (p), f (q)) < whenever dX (p, q) < δp and q ∈ X.
2
n o
let Jp := q ∈ X/dX (p, q) < δ2p = N δp (p). Since p ∈ Jp , and every neigh-
2
borhood is an open set, the collection {Jp }p∈X is an open cover of X. Since X
is compact, there exists finitely many points say p1 , p2 , p3 , . . . , pn in X such
that

X ⊆ Jp1 ∪ Jp2 ∪ · · · ∪ Jpn . → (1)

We choose δ = 12 min {δp1 , δp2 , . . . , δpn }.


Then clearly δ > 0. We will now show that this δ satisfies the required
criterion. Now let p and q be points of X, such that dX (p, q) < δ. From (1)
there exist an integer m, 1 ≤ m ≤ n, such that p ∈ Jpm ; hence

dX (p, pm ) < 21 δpm

also, we have

dX (q, pm ) ≤ dX (q, p) + dX (p, pm ) < δ + 21 δpm ≤ δpm

Hence

dY (f (p), f (q)) ≤ dY (f (p), f (pm )) + dY (f (q), f (pm )) < .

Hence f is uniformly continuous.

Note: Compactness is essential in the hypothesis of the above theorems.

123
Theorem 10.1.4.2 Let E be a non-compact set in R. Then
(a) there exists a continuous function on E which is not bounded;
(b) there exists a continuous and bounded function on E which has no max-
imum.
(c) If E is bounded non-compact set, then there exists a continuous function
on E which is not uniformly continuous.

Proof of (a): If E is an unbounded set, then the function f : E → R


defined by f (x) = x is an unbounded function. If E is bounded, then E is
not closed as E is a non-compact set. So there exists a limit point x0 of E
1
which is not a point of E, then f : E → R defined by f (x) = x−x 0
for x ∈ E
is a continuous function on E, but not bounded on E.

Proof of (b): If E is bounded. Let x0 be a limit point of E which is not


1
in E, then define g : E → R by g(x) = 1+(x−x 0)
2 (x ∈ E). Then it is clear

that g is a continuous and bounded as 0 < g(x) < 1. Also sup g(x) = 1, where
x∈E
as g(x) < 1 for each x in E, hence g doesn’t attains its maximum value on E.

x 2
If E is an unbounded set, then define g : E → R as g(x) = 1+x 2 (x ∈ E),

then g is continuous and bounded as 0 < g(x) < 1 for each x ∈ E. Also
sup g(x) = 1 but there is no point in E. Hence g has no maximum on E.
x∈E

Proof of (c): Since E is bounded, there exists a limit point x0 of E


which is not a point of E. The function
1
h : E → R defined by h(x) = x−x0
for x ∈ E.

Then h is a continuous function on E, but it is clearly unbounded. Now


we will prove that h is not uniformly continuous.

Let  > 0 be given and δ > 0 be arbitrary and choose a point x ∈ E


such that |x − x0 | < δ. Taking t close enough to x0 , we can then make the
difference |f (t) − f (x)| ≥ , although |t − x| < δ. Hence f is not uniformly
continuous.

Note: Considering E to be a bounded set is necessary in part C. For ex-


ample the set of all integers Z is an non-empty set but every function defined
on Z is both continuous and uniformly continuous. To see this, for any given

124
 > 0, choose 0 < δ < 1, then this δ would do.

We will conclude this unit by showing that compactness is also essential


in proving a continuous bijective map is a homeomorphism.

Example 68 Let X = [0, 2π) and Y = {(x, y) ∈ R2 /x2 + y 2 = 1} . Then


define f : X → Y by f (t) = (cos t, sin t). Since sin t and cos t are continuous
function and one-to-one on [0, 2π), it is clear that f is a bijective continuous
map but it is not a homeomorphism as f −1 is not continuous at the point
(1, 0) = f (0). This is so as X is not compact. It is interesting to observe
that f −1 is not continuous even though Y is a compact set.

10.1.5 Terminal problems:


1. In each case, give an example of a function f continuous on S such that
f (S) = T (i. e f is on-to) or else explain why there can be no such f :

(a) S = (0, 1), T = (0, 1]


(b) S = [0, 1] × [0, 1], T = R2
(c) S = [0, 1] ∪ [2, 3], T = {0, 1}
(d) S = [0, 1] × [0, 1], T = (0, 1) × (0, 1)

2. Give an example of a continuous function f and a Cauchy sequence


{xn } in some metric space X for which {f (xn )} is not a Cauchy se-
quence in Y .

3. Prove that the interval (−1, 1) in R is homeomorphic to R. This shows


that neither boundedness nor completeness is a topological property.

4. Prove that a function which is uniformly continuous on X is also con-


tinuous on X.

5. Let f : R → R, defined by f (x) = x2 for x in R, prove that f is not


uniformly continuous on R.

6. Assume f : X → Y is a uniformly continuous on X, where X and


Y are metric space. If {xn } is any Cauchy sequence in X, prove that
{f (xn )} is a Cauchy sequence in Y .

125
7. In a metric space (X, d), let A be a non-empty subset of X. Define a
function fA : X → R by

fA (x) = inf {d(x, y)/y ∈ A} for each x ∈ X.

The number fA (x) is called the distance from x to A.

(a) Prove that fA is uniformly continuous on X.


(b) Prove that A = {x : x ∈ S and fA (x) = c}

10.1.6 Keywords:
Compactness, homeomorphism and uniform continuity.

10.1.7 References:
1. W. Rudin, ''Principles of Mathematical analysis'' Third edition. Me
Grow hill Education(India) Private limited, New Delhi.

2. Tom. M. Apostol,, '' Mathematical analysis '' Second Edition. Narosa


Publishing House. New Delhi.

126
g
Unit XI
Continuity and Connectedness

11.1.1 Objectives
In this unit we will be recalling the definition of connecdeness and exploring
its relation to continuity. Also we will be studying about path wise connected
sets.

11.1.2 Introduction
Most of the times it is difficult to check whether a given set is connected or
not by trying to find separation,, instead we can use some properties and
special continuous function to decide whether a given set is connected set or
not.

11.1.3 Connectedness
Definition 11.1.3.1 A metric space X is said to be disconnected if X =
A ∪ B, where A and B are disjoint non-empty open sets in X. We call X
connected if it is not disconnected.

Example 69 The metric space X = R − {0} with usual euclidean metric is


disconnected, since it is the union of two disjoint nonempty open sets, the
positive real number and the negative real number.

Example 70 Every metric space X contains a non-empty connected subsets.


In fact, for each p in X the set {p} is connected.

Example 71 The set Q of rational numbers, regarded as a metric subspace


of Euclidean space R0 is disconnected.
√ In fact, Q = A ∪ B, where
√ A consists
of all rational numbers < 2 and B of all rational numbers > 2.

To relate connectedness with continuity we introduce the concept of a


two-valued function.

127
Definition 11.1.3.2 A real-valued function f which is continuous on a met-
ric space X is said to be two-valued on X if f (X) ⊆ {0, 1} .
In other words, a two valued function is a continuous function whose only
possible values are 0 and 1. This can be regarded as a continuous function
from X to the metric space T = {0, 1}, where T has the discrete metric. We
recall that every subset of a discrete metric space T is both open and closed
in T .

Theorem 11.1.3.1 A metric space X is connected if and only if every two-


valued function on X is connected.

Proof: First, we assume X is connected and let f be a two -valued function


on X. We must show that f is constant. Let A = f −1 ({0}) and B =f −1 ({1})
be the inverse images of the subsets {0} and {1}. Since {0} and {1} are
open subsets of the discrete metric space {0, 1}, both A and B are open in
X. Hence X = A ∪ B, where A and B are disjoint open sets. But since X is
connected, either A is empty and B = X or else B = ∅ and A = X. In either
case, f is constant on X.

conversely assume that every two valued function on X is constant, then


we need to prove that X is connected. If suppose X is not connected, then
there exist disjoint non-empty open sets in X say A and B such that X =
A ∪ B. Then f : X → R as
(
0 if x ∈ A
f(x) =
1 if x ∈ B
Since A and B are non-empty, f takes both value 0 and 1, so f is not constant.
Also f is continuous on X, inverse image of every open subset of {0, 1} is open
in X. Which is a contradiction to our assumption. Hence X is connected.

The above theorem helps us to answer many questions like


1. Let f : (0, 1) → {0, 1} be a continuous function, then what is f ((0, 1))?
solution :- From the above theorem, we have that this function f has to
be constant as the interval (0, 1) is connected.
2. Is the set A = (0, 1) ∪ (1, 2) connected ?
solution:- If we define f : A → R as

128
(
0 if x ∈ (0, 1)
f(x) =
1 if x ∈ (1, 2)

then clearly f is continuous, as inverse image of every open set is open.


Hence f is two-valued on A, f is not constant. Thus A is not connected.

Now we will show that the continuous image of a connected set is connected
,which intern helps us to obtain many results.

Theorem 11.1.3.2 Let f : X → Y be a function from a metric space X to


another metric space Y. Let X be a connected set. If f is continuous on X,
then f(X) is a connected subset of Y.

Proof: Let g be a two- valued function on f (X), then prove f (X) is a


connected set, we have to show that g is constant. Consider the composite
function h : X → R defined by h(x) = g (f (x)), for each x ∈ X.

Since f and g are both continuous functions, h is also continuous on X


and can only take the values 0 and 1, so h is two-valued function on X.
Since X is connected, h is constant on X and this implies that g has to be a
constant on f (X). Therefore f (X) is connected set.

Note: This theorem helps to answer a wonderful question that ”if X is an


interval in R1 and f : X → R is a continuous map then how does the set
f(X) looks like ?” Since we know that a subset of R is connected iff it is an
interval, from the above theorem we can conclude that f(X) has to be a
connected subset of R, hence f(X) has to an interval.
Hence there can’t exists a continuous function from the set (0, 1) onto
(0, 21 ) ∪ ( 21 , 1).

Example 72 The above theorem also helps us to find many examples of


connected set in Rn . Since any interval X is connected in R, if f: X → Rn
is a continuous map then f(X) is a connected subset of R.
As a corollary of the above theorem, we have the following extension of
Bolzano’s theorem.

129
Theorem 11.1.3.3 (Intermediate-value theorem for real continuous func-
tion)

Let f be a continuous real function on the interval [a, b]. If f (a) < f (b)
and if c is a number such that f (a) < c < f (b), then there exists a point x ∈
(a, b) such that f(x) = c.

Proof: Since [a, b] is a connected subset of R and continuous image of


connected set is connected. f ([a, b]) is connected in R . Hence f ([a, b]) must
be a interval in R. Thus there exists a point x ∈ (a, b) such that f(x) = c as
c ∈ f ([a, b]).

Note: A similar result holds, of course, if f (a) > f (b). Roughly speaking,
the theorem says that a continuous real function assumes all intermediate
values on an interval.

11.1.4 Path wise connected or Arcwise connected


In this section, we will be studying about special property of few connected
set called Path wise connected or Arc wise connected, which is passed by
some connected sets in Euclidean space Rn
Definition 11.1.4.1 A set S in Rn is called path wise (arcwise) connected
if for any two points a and b in S there is a continuous function f : [0, 1] →
S such that f(0) = a and f(1) = b.

Example 73 Let B = {(x, y)/max {|x|, |y|}} ≤ 1, then this is clearly a path
wise connected set as for any two points (x1 , y1 ) and (x2 , y2 ) in B, define f:
[0, 1] → B as
f(t) = t (x2 , y2 ) + (1-t) (x1 , y1 )= (x2 t + (1 − t)x1 , ty2 + (1 − t)y1 )
clearly f is a continuous map and f(0) = (x1 , y1 ) and f(1) = (x2 , y2 ).

Example 74 Every convex set in R is path wise (arc wise) connected.

Note:
1. A function f : [0, 1] → X defined by f(t) = t.b + (1-t) a is a curve
joining a and b, it is called a line segment.

130
2. A continuous function f : [0, 1] → S with f(0) = a and f(1) = b is called
a path from a to b. Thus S is said to be path wise connected(arc wise
connected) if every pair of distinct points in S can be joined by a path
lying in S.

Remark: One gets a natural question that ”Is every connected set path
wise connected?” The answer to this question is need not be true. For
example a set consisting of these points on the curve described by y =
sin( x1 ), 0 < x ≤ 1, along with the points on the horizontal segment
−1 ≤ x  ≤ 0,
i.e A = (x, y) : 0 < x ≤ 1 and y = sin( x1 ) ∪

{(x, y) : y = 0 and − 1 ≤ x ≤ 0}, A is connected but not path wise


connected.

The next theorem shows that the converse of the above remark is true i.e
every path wise connected set is connected for subset of Rn

Theorem 11.1.4.1 Every path wise connected set S in Rn is connected

Proof: Let g be two-valued on S.We will prove that g is constant on S.


Choose a point a in S. Since S is path connected, for any point x in S there
is a path joining a and x. i.e there exists a continuous function f : [0, 1] → S
such that f(0) = a and f(1) = x. Since continuous image of connected set is
connected implies f ([a, b]) ⊆ S is a connected subset of S. Hence g must be
a constant on f ([a, b]) so g(x) = g(a) for each x in f ([a, b]). But since x is
an arbitrary point of S, this shows that g is constant on S, so S is connected.

We have already noted that there is a connected set in Rn which are not
path wise connected. However the following theorem shows that the two
concept are equivalent for open sets in Rn .

Theorem 11.1.4.2 Every open connected set in Rn is path wise connected.

Proof: Let S be an open connected set in Rn and assume x ∈ S. We will


show that x can be joined to every point y in S by a path lying in S. Let A be
the set consisting of all y in S which has a path to x lying in S.

i.e A = {y ∈ S : x and y are path connected}

131
Let B = S - A. Then S = A ∪ B, where A and B are disjoint sets, we will
show that A and B are both open in Rn .

Assume that a ∈ A. Since a ∈ A and S is an open subset of Rn , there


exists  > 0 such that Na () ⊆ S. Now if y ∈ Na (), then line segment joining
y and a lies in Na (). Which implies that y ∈ A. Hence a ∈ Na () ⊆ A. i. e
A is an open set.

To prove that B is also open, assume that b ∈ B. Then there is a δ- neigh-


borhood of b in S, as S is open. i.e Nδ (b) ⊆ S.

Now we need prove show that Nδ (b) ∩ A = ∅ i.e Nδ (b) ⊆ S − A = B. If


suppose y ∈ Nδ (b) ∩ A, which implies that there is a path from a point in A
say 'a' to y. Since, every neighborhood is a path wise connected set implies
there is a path from y to b. Which implies there is a path from a to b, which
is not possible as b∈
/ A. Hence Nδ (b) ∩ A = ∅.
i.e Nδ (b) ⊆ B. Hence B is also an open set.

Therefore we have S = A ∪ B, where A and B are disjoint open sets in


n
R . Moreover, A is not an empty set, by the definition of A, x ∈ A. Since
S is connected, it follows that B must be empty. So S = A. Now clearly A
is path wise connected, because any two of its points can be joined by first
joining each of them to x. Therefore S is path wise connected and the proof
is complete.
Note: A path f : [0, 1] → S is said to be a polygonal if the image of [0, 1]
under f is the union of a finite number of line segments.

11.1.5 Terminal problems:


1. In each case, give an example of a function f , continuous on S and
such that f (S) = T , or else explain why there can be no such f .
(a) S = (0, 1) and T = (0, 1)∪(0, 2)
(b) S = R and T = Q
(c) S = [0, 1] ∪ [2, 3] and T = {0, 1}.
2. Prove that a metric space X is connected if and only if ,the only subsets
of X which are both open in X are empty set and X itself.

132
3. Prove that the only connected subsets of R are

(a) The empty set


(b) The sets consisting of a single point
(c) Interval ( pen, closed, half open, infinite intervals ).

4. Prove that a metric space X is connected iff every non-empty proper


subset of X has a non-empty boundary.

11.1.6 Keywords:
Connecded set, Pathwise connected (arc wise connected).

11.1.7 References:
1. W. Rudin, ''Principles of Mathematical analysis'' Third edition. Me
Grow hill Education(India) Private limited, New Delhi.

2. Tom. M. Apostol, '' Mathematical analysis '' Second Edition. Narosa


Publishing House. New Delhi.

133
g
Unit XII
Discontinuity and Monotonic functions

12.1.1 Objectives
In this unit, we will be studying about discontinuous of a function, in par-
ticular we study about discontinuous of monotonic functions and limit of a
function at infinity and infinite limit.

12.1.2 Introduction
This unit is completely devoted to special properties of real- valued functions
defined on sub intervals of R. Let f be such a function i.e f : S → R where
S⊆ R and S is an interval. We know if such a function is continuous at a
point p, then it satisfy some special property. When do we say a function
is not continuous at a point p? In this unit, we can actually say how many
such points are there for a monotonic function.

12.1.3 Discontinuities
If 'x' is a point in the domain of definition of the function f at which f is not
continuous, we say that f is discontinuous at x, or that f has a discontinuity
at x. If f is defined on an interval or on a segment, then we can classify the
discontinuity into two types. To do so, we need to define the right-hand and
the left- hand limits of f at x, which we denote by f (x+ ) and f (x− ) respec-
tively.

Definition 12.1.3.1 [Right hand limit f (c+ )]

Let f : (a, b) → R be a function and c ∈ [a, b). If ∃ A ∈ R such that for


each  > 0, there is a δ > 0 such that

|f (x) − A| <  when ever c < x < c + δ < b

134
Then we say right hand limit exists and lim+ f(x) = A and we denote it as
x→c
f (c+ ) = A.

Note:

1. Right hand limits of f at c is the limit f (x) tends to A as x tends to


through values greater than c.

2. f (c+ ) = A if f (tn ) tends to A as n → ∞ for all sequence {tn } in (c,


b) such that tn → c. This is the sequential definition of right hand limit.

On similar lines we define Left hand limit of a function

Definition 12.1.3.2 [Lef t hand limit f (c− )]

Let f : (a, b) → R be a function and c ∈ (a, b]. If ∃ A ∈ R such that for


each  > 0, there is a δ > 0 such that

|f (x) − A| <  when ever a < c + δ < x < c


Then we say left hand limit exists and lim− f(x) = A and we denote it as
x→c
f (c− ) = A.

Example 75 Let f : (0, 2) → R be defined as


(
x if 0 < x ≤ 1
f(x) =
2 + x if 1 < x < 2

Then f (1+ ) = 3 where as f (1− ) = 1


However if x 6= 1 then f (x+ ) = f (x− ) for each x ∈ (0, 2).

Hence the above example clearly shows that right hand limits and left hand
limit of a function need not be same at every point.

Definition 12.1.3.3 [Discontinuous] A function f is said to be discontinu-


ous at a point x, if f is not continuous at x

135
Classification of discontinuity :-

Based on right hand and left hand limit of a function f at a point x, the
discontinuous is classified in two kinds

1. First kind or simple discontinuity : If f is discontinuous at a point x,


and if both f (x+ ) and f (x− ) exists then we say f has a discontinuity of
the first kind or a simple discontinuity at x.

The discontinuity of the first kind or a simple discontinuity can be


further classified as two kinds

(a) Removable discontinuity : If f has first kind discontinuity at x and


if f (x+ ) = f (x− ), then f is said to have a removable discontinuity
at x.
(b) Irremovable discontinuity: If f has first kind discontinuity at x and
if f (x+ ) 6= f (x− ), then f is said to have an irremovable disconti-
nuity at x

2. Second kind : If f is discontinuity at a point x and if either f (x+ ) or


f (x− ) doesn’t exists then f is said to have a discontinuity of second
kind. This discontinuity is also irremovable discontinuity

Note: Removable discontinuity, as the name itself says the discontinuity


could be removed by redefining the function f at x as f(x) = f (x+ ) = f (x− ).

Example 76 Define f : R → R, as
(
x
|x|
if x 6= 0
f(x) =
A if x = 0

Then clearly f has a discontinuity at x = 0, let us find out the kind of dis-
continuity by finding f (0+ ) and f (0− ).

x
f (0+ ) = lim+ f(x) = lim+ |x|
(since x > 0, |x| = x)
x→0 x→0

x
= lim+ x
= 1.
x→0

136
x
Where as f (0+ ) = lim− f(x) = lim− |x|
(since x < 0, |x| = −x)
x→0 x→0

x
= lim− −x
= -1.
x→0

i,e Both f (0+ ) and f (0− ) exists, hence f is discontinuous at 0. Also since
f (0+ ) 6= f (0− ) it is an irremovable discontinuous of first kind or simple
discontinuity at 0.

Example 77 Define f : R → R, by
(
1 if x 6= 0
f(x) =
0 if x = 0

It is clear that f is discontinuous at x = 0, then it is easy to check that f (0+ )


= 1 and f (0− ) = 1 . Hence f is removable discontinuity of first kind at 0.

Example 78 Define f : R → R, by
(
1
x
if x 6= 0
f(x) =
A if x = 0

Then the only point where f is discontinuous is x = 0, to find its kind, lets
find f (0+ ) and f (0− ).
Let f (0+ )= lim+ x1 = doesn’t exists (why?)
x→0
f (0− )= lim− 1
x
= doesn’t exists Similarly. Hence f has a discontinuity of
x→0
second kind at 0.

Example 79 Define f : R → R by
(
sin x1 if x 6= 0
f (x) =
A if x = 0

Then clearly f is not continuous at 0. Also lim+ f (x) and lim− f (x) doesn’t
x→0 x→0
exists, hence f has a discontinuity of second kind at x = 0, which is irremov-
able.

Example 80 Define f : R → R by

137
(
x sin x1 if x 6= 0
f (x) =
1 if x = 0

Then clearly f has a discontinuity at x = 0, let f (0+ ) = lim+ f (x) =


x→0
lim+ x sin x1 .
x→0
Since |x sin x1 | ≤ |x|

we obtain lim+ x sin x1 = 0 = f (0+ ).


x→0

For the same reason, we obtain lim− x sin x1 = 0 = f (0− ).


x→0
Hence f (0+ ) = f (0− ) 6= 1 = f (0) i.e f has a simple discontinuity at
x = 0 and this discontinuity can be removed i.e the function g : R → R
defined by
(
x sin x1 if x 6= 0
g(x) =
0 if x = 0

is a continuous function at every point on real line.

Example 81 Define f : R → R by
(
1 if x is rational
f (x) =
0 if x is irrational.

Then f is discontinuous at every point of R and the discontinuity is of the


second kind as neither f (x+ ) nor f (x− ) exists.

Example 82 Define f : R → R
(
x if x is rational
f (x) =
0 if x is irrational.

Then the function is continuous only at x = 0 and since f is discontinuous


on R − {0}. Since irrational are dense the limits f (x+ ) and f (x− ) doesn’t
exists. Hence f has a discontinuity of second kind at every x ∈ R − {0}.

138
12.1.4 Monotonic functions:
We shall now study those functions which either never decrease or never
increase on a given segment.

Definition 12.1.4.1 Let f : (a, b) → R be a function. Then f is said to


be monotonically increasing on (a, b) if a < x < y < b implies f (x) ≤ f (y).
f is said to be monotonically decreasing on (a, b) if a < x < y < b implies
f (x) ≥ f (y).

Note: Function f : (a, b) → R is said to be monotonic if either f is


monotonically increasing or monotonically decreasing.

Remark: If f is monotonically increasing function, then −f is a


monotonically decreasing function. Because of this simple fact, in many
situations involving monotonic functions it suffices to consider only the case
of increasing functions. We have seen in the previous section that there
exists several functions for which left hand limit and right hand limit
doesn’t exists. Now, we shall prove that functions which are monotonic on
compact intervals always have finite right hand and left hand limits. Hence
their discontinuities must be always simple removable discontinuities.

Theorem 12.1.4.1 If f is monotonically increasing on (a, b). Then f (c+ )


and f (c− ) both exists for each c in (a, b) and we have
sup f (t) = f (c− ) ≤ f (c) ≤ f (c+ ) = inf f (t)
a<t<c c<t<b

Further more, if a < x < y < b, then


f (c+ ) ≤ f (y − )
Proof: Let A = {f (x) : a < x < c}. Since f is monotonically increasing,
this set is bounded above by f (c). Hence A has a supremum say α = sup A.
Then α ≤ f (c) and we shall prove that f (c− ) exists and it is equal to α. Let
 > 0 be given, since α = sup A, α −  is not an upper bound for the set A
i.e there exists some a < c − δ < c such that α −  < f (c − δ) ≤ α since f is
monotonically increasing, we have
f (c − δ) ≤ f (t) ≤ f (c) whenever c − δ < t < c
Hence we obtain, |f (t) − α| <  whenever t in (c − δ, c). Hence

139
f (c− ) = α

Similarly we define B = {f (x) : c < x < b}. Since f is monotonically


increasing, this set is bounded below by f (c). Hence the set B has infimum
say β = inf B. Then f (c) ≤ β and we shall prove that f (c+ ) = β.

Let  > 0 be given, since β < β + , β +  is not a lower bound for B.


Hence there exists δ > 0 such that β < f (c + δ) < β + .
Since f is an increasing function implies that |f (t)−β| <  whenever |c−t| <
δ. Thus lim+ f (t) exists and equal to β i.e f (c+ ) = β. Next if a < x < y < b,
t→c
we have proved that

f (x+ ) = inf f (t) = inf f (t)(as f is monotonically increasing function )


x<t<b x<t<y

and f (y − ) = sup f (t) = sup f (t) (as f is monotonically increasing


a<t<y x<t<y
function)

Thus, we obtain inf {f (t)/x < t < y} ≤ sup {f (t)/x < t < y}
i.e f (x+ ) ≤ f (y − )
Hence the proof of the theorem.

Corollary 12.1.4.1 Monotonic functions have no discontinuities of second


kind.

Note: Analogous result of the above theorem holds for monotonically


decreasing function.

Theorem 12.1.4.2 Let f be monotonic on (a, b). Then the set of points of
(a, b) at which f is discontinuous is at most countable.

Proof: First we will assume that f is monotonically increasing. Let


E := {x ∈ (a, b)/f is discontinuous at x} Now, we need to prove that E is
a countable set. From the previous theorem we know that, for each x ∈ E,
f (x+ ) and f (x− ) exists. Since f is monotonically increasing, f (x− ) < f (x+ ).
Since set of all rational numbers Q is dense in R, there exists some rational
number say r(x) such that f (x− ) < r(x) < f (x+ ). Let x1 , x2 ∈ E with

x1 < x2 , implies f (x+ 1 ) ≤ f (x2 ) which intern implies that r(x1 ) 6= r(x2 ).

140
Since the set of all rational number Q is countable, if we define g : E → Q
g(x) = r(x) then clearly this is a one-one function, hence E has to be a
countable set. The proof is similar if f is monotonically decreasing.

Remark: Discontinuities of a monotonic function need not be isolated.

12.1.5 Infinite limits and limits at infinity.


The extended real number system consists of all the real numbers including
±∞. Can we define a limit of a function at ±∞? YES. To define a limit of
a function we have used the concept of a neighborhood of a point in R. Can
we define neighborhood of ±∞? YES. We define it as follows.

Definition 12.1.5.1 For any real number c. The set Nc (+∞) := {x ∈ R/x > c} =
(c, +∞) is called a c−neighborhood of +∞. Similarly the set (−∞, c) is a
c−neighborhood of −∞.

Now, by using the definition of neighborhood of ±∞, we can define


lim f (t) and lim f (t), using the following general definition of a limit of a
t→∞ t→−∞
function involving neighborhood.

Definition 12.1.5.2 Let E ⊆ R and f : E → R be a function. We say that


lim f (t) = A where A, x ∈ R ∪ {±∞} if for every neighborhood U of A there
t→x
is a neighborhood V of x such that V ∩ E, t 6= x.

Note: Particular cases of the above definition

1. When x = +∞, and A ∈ R. The lim f (t) = A is equivalent to say ∃


t→∞
A ∈ R such that ∀  > 0 there is a δ ∈ R such that |f (t) − A| < 
whenever t > δ and t ∈ E.

2. When x = −∞, and A ∈ R. Then the lim f (t) = A is equivalent to


t→−∞
say ∃ A ∈ R such that ∀  > 0 ∃ δ ∈ R such that |f (t) − A| < 
whenever t ∈ E with t < δ.

3. When A = +∞, and x ∈ R. The lim f (t) = ∞ is equivalent to say


t→x
that if ∀ c ∈ R, there exist δ > 0 such that f (t) > c for all t in E with
0 < |t − x| < δ.

141
4. When A = −∞, and x ∈ R. The lim f (t) = −∞ is equivalent to say
t→x
that ∀ c ∈ R, ∃ δ > 0 such that f (t) < c for all t in E with
0 < |t − x| < δ.

Remark:

1) The first two cases, in which x = ±∞ is called the limit at infinity of


the function f .

2) The last two cases, in which A = ±∞ is called the infinite limit of the
function f .

Example 83 Let f : (0, 1) → R be defined by f (x) = x1 , then lim f (x) = ∞


x→0
1
as for any real number, c ∈ R, there exist δ > 0 such that x
> c if 0 < x < δ.
Hence the above function has infinite limit at x = 0.

Note: In the above example we can’t talk about lim f (t), a the domain of
t→±∞
the definition of f doesn’t intersects infinitely many neighborhood of ±∞.

Example 84 Let f : (0, ∞) → R be defined by f (x) = x, then


lim f (x) = ∞.
x→∞

Example 85 Let f : (1, ∞) → R be defined as f (x) = x1 , then lim f (t) = 0.


t→∞

It is interesting to see that algebra of limits also holds for this extended
real number system. The following theorem illustrate this.

Theorem 12.1.5.1 Let f and g be real valued functions defined on E lim R.


n→∞
Suppose lim f (t) = A and lim f (t) = B. Then
t→x t→x

a) f (t) → A0 implies A0 = A i.e limit is unique,

b) (f + g)(t) → A + B,

c) (f g)(t) → AB and

142
d) (f /g)(t) → A/B, provided the right members of (b), (c) and (d) are
defined.

Proof: Proof is similar to that of the earlier proof, here we need to apply
the neighborhood definition of limit of a function.

∞ A
Note: ∞ − ∞, 0.∞, , and are not define. Hence A and B be should
∞ 0
be the value other than mentioned one.

12.1.6 Terminal problems:


1. Find and classify the function f defined on R by the following equations.

 sin x
if x 6= 0
(a) f(x) = x
0 if x = 0.
( 1
e x if x 6= 0
(b) f(x) =
0 if x = 0.
( 1
e x + sin x1

if x 6= 0
(c) f(x) =
0 if x = 0.
 1

1 if x 6= 0
(d) f(x) = 1 − e x
0 if x = 0.

2. Let f be continuous on a compact interval [a, b] and assume that f


doesn’t have a local maximum or a local minimum at any interior point.
Prove that f must be monotonic on [a, b].

3. Discuss the limit at infinity and infinite limit for all the function in
problem 1.

12.1.7 Keywords:
Discontinuity, Removable discontinuity, Discontinuity of first kind and second
kind, monotonically increasing function, monotonically decreasing function.

143
12.1.8 References:
1. W. Rudin, ''Principles of Mathematical analysis'' Third edition. Me
Grow hill Education(India) Private limited, New Delhi.

2. Tom. M. Apostol, '' Mathematical analysis '' Second Edition. Narosa


Publishing House. New Delhi.

144
BLOCK IV
Unit XIII
Differentiation of Real-valued functions.

13.1.1 Objectives
In this unit, we will be studying about derivative of a real-valued functions
defined on an interval in R. Its basic properties and algebra of derivatives.

13.1.2 Introduction
Derivatives are the central concepts of differential calculus. Two different
types of problems leads us to the same answers i.e derivatives, they are

1)The physical problem of finding the instantaneous velocity of a moving


particle.

2) The geometrical problem of finding the tangent line to a curve at a


given point.

Here, we shall not be concerned with applications to mechanics and geom-


etry, but instead will confine our study to general properties of derivatives.

13.1.3 Derivatives
Let c ∈ (a, b), let f : (a, b) → R, then for any x ∈ (a, b) with x 6= c, we can
define
f (x) − f (c)
x−c
we keep c fixed and study the behavior of this quotient as x → c.

Definition 13.1.3.1 Let f : (a, b) → R and assume that c ∈ (a, b). Then f
is said to be differentiable at c if the limit
f (x) − f (c)
lim
x→c x−c
exists. We denote this limit by f 0 (c), it is called the derivative of f at c.

145
Remark: This limit process defines a new function f 0 , whose domain
consists of those points in (a, b) at which f is differentiable. The function f 0
is called the first derivative of f . Similarly, the nth derivative of f , denoted
by f (n) , is defined to be the first derivative of f (n−1) , for n = 2, 3, . . . other
notations with which the reader may be familiar are
df dy
f 0 (c) = Df (c) = (c) = |x=c , where y = f (x).
dx dx
The function f itself is sometimes written as f (0) . The process which
produces f 0 from f is called differentiation.

Example 86 Let f (a, b) → R be a constant function say f (x) = k ∀ x ∈


(a, b). Then the quotient
f (x) − f (c) k−k
= = 0 ∀ x ∈ (a, b) with x 6= c.
x−c x−c
f (x) − f (c)
Hence lim = 0.
x→c x−c
i.e f 0 (c) = 0 for each c ∈ (a, b).

Example 87 Let f : J → R, (where J be an interval in R) be given by


f (x) = ax + b. Let c ∈ J, consider the quotient
f (x) − f (c) (ax + b) − (ac + b) a(x − c)
= = = a.
x−c x−c x−c
f (x) − f (c)
Thus lim = a = f 0 (c).
x→c x−c
Example 88 Let f : J → R, be given by f (x) = x2 , let c ∈ J, consider the
quotient.
f (x) − f (c) x 2 − c2
= = x + c.
x−c x−c
f (x) − f (c)
Hence lim = 2c = f 0 (c).
x→c x−c
Derivatives and continuity Is there any relationship between derivatives
and continuous function? Following theorem makes it possible to reduce
some of the theorem on derivatives to theorems on continuity.

146
Theorem 13.1.3.1 If f is defined on (a, b) and differentiable at a point c
in (a, b), then there is a function f ∗ (depending on f and an c) which is
continuous at c and which satisfies the equation

f (x) − f (c) = f ∗ (x)(x − c) → (1)

for all x in (a, b), with f ∗ (c) = f 0 (c). Conversely, if there exists a function
f ∗ , continuous at c, which satisfies (1), then the f is differentiable at c and
f 0 (c) = f ∗ (c).

Proof: If suppose f is differentiable at c, then define



 f (x) − f (c)
∗ 6 c
if x =
f (x) = x−c
f 0 (c) if x = c

Then clearly f ∗ satisfies the condition (1) and also


f (x) − f (c)
lim f ∗ (x) = lim = f 0 (c) = f ∗ (c).
x→c x→c x−c
Hence f ∗ is continuous at c.
Conversely, suppose there exist a function f ∗ on (a, b) which satisfies (1) and
f (x) − f (c)
continuous at c, then lim = lim f ∗ (x) = f ∗ (c)
x→c x−c x→c
the limit exist, hence f is differentiable at c and f 0 (c) = f ∗ (c).

Above theorem, we have approximated a given differentiable function by a


linear function or a first degree polynomial. i. e

f (x) = f (c) + f 0 (c)(x − c) + error term

where error term goes to zero as x → c.

1
Example 89 Let f (x) = for x 6= 0. Then f is differentiable at c 6= 0 and
x
−1
0
f (c) = 2 for c 6= 0. We have f (x) − f (c) = x1 − 1c = c−x xc
= −1
xc
(x − c).
c
This suggests that we have f ∗ (x) = −1cx
for x 6= 0. Clearly f ∗ is continuous
for x 6= 0. This proves that f (x) = x1 is differentiable at any c 6= 0 and that
lim f ∗ (x) = −1
c2
= f 0 (c).
x→c

147
Example 90 Let n ∈ N, we shall now show that f : (0, ∞) → (0, ∞) defined
1
by f (x) = x n is differentiable. Let a > 0 be fixed.
1 1
Consider f (x) − f (a) = x n − a n

Using the formula,

(t − s)(tn−1 + tn−2 + · · · + sn−1 ) = (tn − sn ).


1 1
and substituting t := x n and s := a n , we obtain.
1 1 n−1 n−1
(x n − a n )(x n + ··· + a n ) = (x − a)

1 1 1
(x n − a n ) = (x − a) n−1 n−1
(x n + ··· + a n )

1 1 1
Clearly f ∗ (x) = n−1 n−1 and lim f ∗ (x) = n−1 = n1 a n −1 .
(x n + ··· + a n ) x→a na n

1
f 0 (a) = a −1 .
1 n
n

Now we will prove that if a function is differentiable at a point then it is


continuous at that point, we will also see that the converse of this statement
need not be true.

Theorem 13.1.3.2 Let f : (a, b) → R be a function and c ∈ (a, b). If f is


differentiable at c, then f is continuous at c.

Proof: Since f is differentiable at c,


f (x) − f (c)
lim exists.
x→c x−c
f (x) − f (c)
Consider f (x) − f (c) = (x − c).
x−c

f (x) − f (c)
lim(f (x) − f (c)) = lim lim(x − c)
x→c x→c x−c x→c

= f 0 (c) 0

148
=0

which implies lim f (x) = f (c)


x→c

Hence f is continuous at c.

Note: Converse of the above theorem need not true for example f (x) = |x|
is continuous everywhere, whereas it is not differentiable at x = 0, as
f (x) − f (0)
lim+ = lim+ 1 = 1
x→0 x−0 x→0

f (x) − f (0)
but lim− = lim− (−1) = −1.
x→0 x−0 x→0

f (x) − f (0)
implies lim doesn’t exist!
x→0 x−0
Algebra of Derivatives:
The next theorem describes the usual formulas for differentiating the sum,
difference, product and quotient of two functions.

Theorem 13.1.3.3 Let f, g : J → R be differentiable at c ∈ J. Then the


following holds:
(a) f + g is differentiable at c with (f + g)0 (c) = f 0 (c) + g 0 (c).
(b) αf is differentiable at c with (αf )0 (c) = αf 0 (c).
(c) f g is differentiable at c with (f g)0 (c) = f (c)g 0 (c) + g(c)f 0 (c).
(d) If f is differentiable at c with f (c) 6= 0, then φ := f1 is differentiable at c
−f 0 (c)
with φ0 (c) = .
(f (c))2
Proof: Since f and g are differentiable at c, there exist continuous func-
tion f ∗ and g ∗ respectively such that

f (x) = f (c) + f ∗ (x)(x − c) → (1)

and

g(x) = g(c) + g ∗ (x)(x − c) → (2).

149
Proof of (a) Define h(x) := f (x) + g(x). Then using (1) and (2)
h(x) = f (x) + g(x) = [f (c) + g(c)] + [f ∗ (x) + g ∗ (x)](x − c). Define h∗ (x) =
f ∗ (x) + g ∗ (x). Then h∗ is continuous at c. Hence f + g is differentiable at c
and h0 (c) = lim h∗ (x) = f 0 (c) + g 0 (c).
x→c

Hence (f + g)0 (c) = f 0 (c) + g 0 (c).

Proof of (b) Using (1), we can write


αf (x) = αf (c) + αf ∗ (x)(x − c).
Since f ∗ is continuous at c, αf ∗ is continuous at c. Therefore, αf is differ-
entiable at c and its derivative is αf 0 (c).

Proof of (c) Let φ(x) = f (x)g(x).


From (1) and (2), we obtain
φ(x) = [f (c) + f ∗ (x)(x − c)][g(c) + g ∗ (x)(x − c)]
= φ(c) + f (c)g ∗ (x)(x − c) + g(c)f ∗ (x)(x − c) +
∗ ∗ 2
g (x)f (x)(x − c)

= φ(c) + [f (c)g ∗ (x) + g(c)f ∗ (x) + f ∗ (x)g ∗ (x)(x − c)](x − c)

If we define
φ∗ (x) := [f (c)g ∗ (x) + g(c)f ∗ (x) + f ∗ (x)g ∗ (x)(x − c)]
then we know that

φ0 (c)(c) = lim φ∗ (x) = f (c)g 0 (c) + g(c)f 0 (c) + 0.


x→c

Hence the proof of (c).

Proof of (d) Since f (c) 6= 0, there exists a neighborhood, in which


f (x) > 0 for x ∈ (c − δ, c + δ) ⊆ J, we have f (x) 6= 0. Hence the function f1
is well defined on the interval (c − δ, c + δ).
1
Let φ(x) = f (x) .

1 1
Consider φ(x) − φ(c) = f (x)
− f (c)
from (1), we have

150
1 1
φ(x) − φ(c) = −
f (c) + f ∗ (x)(x − c) f (c)

−f ∗ (x)
= (x − c).
f (c)[f (c) + f ∗ (x)(x − c)]

Since f (x) = f (c) + f ∗ (x)(x − c) 6= 0 in the open interval (c − δ, c + δ). We


can define
−f ∗ (x)
φ∗ (x) =
f (c)[f (c) + f ∗ (x)(x − c)]
clearly φ∗ (x) is continuous at 'c' and
−f 0 (c)
φ∗ (c) = lim φ∗ (x) = .
x→c [f (c)]2

Note: Using the definition, we have seen that if f is constant on (a, b) then
f 0 = 0 on (a, b). Also if f (x) = x, then f 0 (x) = 1 for all x. Now by applying
the above theorem repeatedly we obtain if f (x) = xn , then f 0 (x) = nxn−1 .
Again, by applying this, we see that every polynomial has a derivative
every where in R and every rational function has a derivative where ever it
is defined.

Chain rule

This is a much deeper result then the above proved results, it gives us
the derivative of composite function.

Theorem 13.1.3.4 (Chain rule) Let f : J → R be differentiable and f (J) ⊂


J1 , an interval and if g : J1 → R is differentiable at f (c), then g ◦ f is
differentiable at c with
(g ◦ f )0 (c) = g 0 (f (c)).f 0 (c).
Proof: Since f is differentiable at x = c, there exist f ∗ : J → R, which is
continuous at c such that
f (x) = f (c) + f ∗ (x)(x − c) and f 0 (c) = f ∗ (c). →(1)
Similarly, there exists g ∗ : J1 → R, continuous at d = f (c) ∈ J1 , such that
g(y) = g(d) + g ∗ (y)(y − d) and g 0 (d) = g ∗ (d) for y ∈ J1 → (2)

151
Let h = g ◦ f , we have to find a function h∗ : J → R which is continuous at
'c', such that

h(x) = h(c) + h∗ (x)(x − c) = g(f (c)) + h∗ (x)(x − c).

Consider the composition,

(g ◦ f )(x) = g(f (x))

= g[f (c) + f ∗ (x)(x − c)] from (1)

= g[d + f ∗ (x)(x − c)]

= g(y) where y = d + f ∗ (x)(x − c).

Using (2), the last expression on the right can be written as

g(y) = g(d) + g ∗ (y)(y − d)

= g(d) + g ∗ [f (c) + f ∗ (x)(x − c)][f (c) + f ∗ (x)(x − c) − f (c)]

= g(f (c)) + g ∗ [f (c) + f ∗ (x)(x − c)]f ∗ (x)(x − c).

If we choose,
h (x) = g ∗ [f (c) + f ∗ (x)(x − c)]f ∗ (x)

then clearly h∗ is continuous at c and it satisfies the required relation, hence


h = g ◦ f is differentiable at c. Also, we have

h0 (c) = (g ◦ f )0 (c) = h∗ (c) = g ∗ (f (c)).f ∗ (c)

= g 0 (f (c)).f 0 (c).
This completes the proof.

152
Some important derivatives

(a) Let f be defined by


(
x sin x1 if x 6= 0
f (x) =
0 if x = 0

(Assuming that derivative of sin x is cos x). Lets find the the derivative of f .
If x 6= 0, then by applying algebra of derivative(product rule) we obtain

f 0 (x) = sin x1 − x1 cos x1 for x 6= 0.


1
At x = 0, we can’t apply product rule as x
is not defined at this point. Hence
we directly use the definition.

f (t) − f (0) t sin 1t − 0


Consider for t 6= 0, =
t−0 t−0
= sin 1t

f (t) − f (0)
Then, lim = lim sin 1t doesn’t exists.
t→0 t−0 t→0

Hence f is not differentiable at x = 0.

(b) Let f be defined by


(
x2 sin x1 if x 6= 0
f (x) =
0 if x = 0 .

Again, for x 6= 0, we can apply product rule to obtain


1 1
f 0 (x) = 2x sin
− cos .
x x
At x = 0, we use the definition directly. To do so,
Consider,
f (t) − f (0) t2 sin 1t − 0
=
t−0 t−0

153
= t sin 1t

f (t) − f (0)
Since
= t sin 1 ≤ |t| (t 6= 0);
t
t−0

As t → 0, we obtain
f (t) − f (0)
lim = lim t sin 1t = 0.
t→0 t−0 t→0

Hence f is differentiable at all points but f ' is not a continuous function,


since cos x1 doesn’t tend to a limit as x → 0.

(c) Let α ∈ R and x > 0. Define, φ : (0, ∞) → R by φ(x) := xα = eα log x .

Clearly φ is the composition of two functions namely f : (0, ∞) → R and


g : R → R defined by f (x) = α log x and g(x) = ex .

Hence φ = g ◦ f i.e φ(x) = g(f (x)).

Hence by chain rule

φ0 (x) = g 0 (f (x))f 0 (x)

= eα log x . αx = (α.xα ) x1 = αxα−1 .

Note: The above formulae can’t be used if x ≤ 0, see terminal problem (2).

154
13.1.4 Terminal problems:
1. Show that f : R → R given by f (x) = |x| is not differentiable at x = 0.

1
2. Show that f (x) = x 3 is not differentiable at x = 0.

3. Let n ∈ N, define f : R → R by
(
xn if x ≥ 0
f (x) =
0 if x < 0 .

(a) For which values of n, is f continuous at 0?

(b) For which values of n, is f ' continuous at 0?

(c) For which values of n, is f differentiable at 0?

(d) For which values of n, is f ' differentiable at 0?

4. Let f : R → R be differentiable at x = 0. Define g(x) = f (x2 ). Show


that g is differentiable at 0 by using the chain rule.

5. Let m, n be positive integers. Define


(
xn if x ≥ 0
f (x) =
xm if x < 0 .

Discuss the differentiablity of f at x = 0.

6. Let f be defined for all real x, and suppose that

|f (x) − f (y)| ≤ (x − y)2

for all real number x and y. Then prove that f is a constant.

7. Let f1 , f2 , g1 and g2 be four functions having derivatives in (a, b). Define


F by means of the determinant

155

f1 (x) f2 (x)
F (x) = , if x ∈ (a, b).
g1 (x) g2 (x)

(a) Show that F 0 (x) exists for each x in (a, b) and that
0
0
f 1 (x) f 0 2 (x) f1 (x) f2 (x)
F (x) = +
g1 (x) g2 (x) g 0 1 (x) g 0 2 (x)
.
(b) State and prove a more general result for nth order determinants.

8. Derive Liebnitz’s formula for the nth derivative of the product h of two
function f and g i.e if h = f.g. Then
n n!
h(n) (x) = ()f (k) (x)g (n−k) (x), where () =
P
k=0 k!(n − k)!

13.1.5 Keywords
Differentiability of a function, derivative, chain rule.

13.1.6 References
1. W. Rudin, ''Principles of Mathematical analysis'' Third edition. Me
Grow hill Education(India) Private limited, New Delhi.

2. Tom. M. Apostol, '' Mathematical analysis '' Second Edition. Narosa


Publishing House. New Delhi.

3. Ajit kumar and S. Kumaresan, '' A Basic course in Real analysis'',


CRC press.

156
g
Unit XIV
Mean value Theorems

14.1.1 Objectives
In this unit, we will be studying about Rolle’s theorem, mean value theorem
and generalized mean value theorem and its applications.

14.1.2 Introduction
Given a function 'f ', how do we find out the point at which f is either
maximum or minimum? We have already learned from calculus courses,
differential calculus is powerful tool in problems of maxima-minima. In this
unit, we shall establish the so called first derivative test and its applications.

Definition 14.1.2.1 (Local Maximum) Let f be real valued function defined


on a metric space X. We say that f has a local maximum at a point p ∈ X
if there exists δ > 0 such that f (q) ≤ f (p) for all point q ∈ Nδ (p) and q ∈ X.

Definition 14.1.2.2 (Local minimum)Let f be a real valued function defined


on a metric space X. We say that f has a local minimum at a point p ∈ X
if there exists δ > 0 such that f (p) ≤ f (q) for all q ∈ Nδ (p) and q ∈ X.

Since we are dealing with real-valued function of real variable, following


is a precise definition for this case.

Definition 14.1.2.3 (Local maximum) Let J ⊆ R be an interval and f :


J → R be a function, we say that a point c ∈ J is a point of local maximum
if there exists δ > 0 such that (c − δ, c + δ) ⊆ J and f (x) ≤ f (c) for all
x ∈ (c − δ, c + δ).

A local minimum is defined similarly.

Note: A point c is said to be a local extremum if it is either a local


maximum or a local minimum.

157
The next theorem shows a connection between zero derivatives and local
extrema at interior points.

Theorem 14.1.2.1 Let J ⊆ R be an interval, let f : J → R be differentiable


on J and c ∈ J. If c is a local extremum of f then f 0 (c) = 0.

Proof: Let f have local maximum at c. Then there exists a δ > 0 such
that f (x) ≤ f (c) for all x ∈ (c − δ, c + δ). That is f (c + h) ≤ f (c) and
f (c − h) ≤ f (c) for all h ∈ (−δ, δ). Since f is differentiable at c, we have
f (c + h) − f (c)
f 0 (c) = lim
h→0 h
Also, we have
f (c + h) − f (c) f (c + h) − f (c)
f 0 (c) = lim+ ≤ 0 and f 0 (c) = lim− ≥ 0.
h→0 h h→0 h
Hence from above two inequality, we deduce f 0 (c) = 0.
The proof is similar if f has a local minimum at c.

Remark:
1) The converse of the above theorem need to be true. In general, knowing
that f 0 (c) = 0 is not enough to determine whether f has an extremum at c.
In fact, it may have neither, as an example, consider f : R → R, defined by
f (x) = x3 and at c = 0. In this case f 0 (0) = 0 but f is increasing in every
neighborhood at 0.
2) Furthermore, it should be noted that f can have a local extremum at c
without f 0 (c) being zero. For example, f (x) = |x| has a local minimum at
x = 0, but, of course, there is no derivative at 0.

158
14.1.3 Rolle’s Theorem.
It is geometrically evident that a sufficiently '' smooth '' curve which crosses
the x−axis at both endpoints of an interval [a, b] must have a '' turning
point'' somewhere between a and b. The precise statement of this fact is
known as Rolle’s theorem.

Theorem 14.1.3.1 (Rolle’s theorem) Let f : [a, b] → R be such that


i) f is continuous on [a, b],
ii) f is differentiable on (a, b) and
iii) f (a) = f (b).
Then there exists c ∈ (a, b) such that f 0 (c) = 0.

Proof: We assume f 0 is never 0 in (a, b) and obtain a contradiction.


Since f is continuous on a compact set, it attains its maximum M and its
minimum 'm' some where in [a, b]. Neither 'M' or 'm' is attained at an
interior point, otherwise f would be vanishes there so both are attained at
the endpoints. Since f (a) = f (b), then m = M and hence f is constant
on [a, b]. This contradicts the assumption that f 0 is never zero on (a, b).
Therefore f 0 (c) = 0 for some c in (a, b).

Now we will look at the most important result in differentiation.

Theorem 14.1.3.2 (Mean value theorem) Let f : [a, b] → R be such that


(i) f is continuous on [a, b],
(ii) f is differentiable on (a, b). Then there exists c ∈ (a, b) such that
f (b) − f (a) = f 0 (c)(b − a).

(f (b) − f (a))(x − a)
Proof: Consider g(x) = f (x)−l(x), where l(x) := f (a)+ .
b−a
Clearly g(a) = g(b) = 0 and g is continuous on [a, b] and differentiable on
(a, b) i. e g satisfies all conditions of Rolle’s theorem. Hence there exists
c ∈ (a, b) such that g 0 (c) = 0. This implies f 0 (c) − l0 (c) = 0
f (b) − f (c)
=⇒ f 0 (c) − = 0.
b−a
i.e f (b) − f (a) = f 0 (c)(b − a). This completes the proof.

f (b) − f (a)
Remark: We can write the concluding identity as f 0 (c) = .
b−a
Note that the right hand side is the slope of the chord joining two points

159
(a, f (a)) and (b, f (b)). As mentioned earlier, we may consider f 0 (c) as the
slope of the tangent line at (c, f (c)) to the graph of f . Thus we arrive at the
geometric interpretation of mean value theorem: under the given conditions,
there exists c such that the slope of the tangent to the graph of f at c equals
that of the chord joining the two points (a, f (a)) and (b, f (b)).

Applications of Mean value theorem

The mean value theorem is also known as Lagrange’s mean value theorem.
It is the single most important result in the theory of differentiation. Below
are some typical applications. Perhaps the only way to derive the following
result is to use the mean value theorem.

Result 14.1.3.1 Let J be an interval and f : J → R be differentiable with


f 0 (x) = 0 for all x ∈ J. Then f is a constant on J.

Proof: For any x, y ∈ J, by mean value theorem, we have f (y) − f (x) =


f (z)(y − x) for some z between x and y. Since f 0 (z) = 0, we get f (y) = f (x)
0

for all x, y ∈ J. This implies that f is a constant function.

Note:

1. The above result need not be true if J is not an interval. For example
f : V → R where U = (−1, 0) ∪ (0, 1) defined as
(
−1 if x ∈ (−1, 0)
f (x) =
1 if x ∈ (0, 1) .

Then f 0 (x) = 0 for all x, but f is not a constant function.

Result 14.1.3.2 Let J be an interval and f : J → R be differentiable with


f 0 (x) > 0 for x ∈ J. Then f is increasing on J.

Proof:For any x, y ∈ J with x < y by mean-value theorem, we have


f (y) − f (x) = f 0 (z)(y − x) for some z between x and y. Since f 0 (z) > 0 and
y − x > 0, the right side is positive. Thus f (x) < f (y). Thus f is strictly
increasing. Similarly the corresponding results holds for each x ∈ J.

160
Result 14.1.3.3 Let f : J → R be differentiable. Assume that there exists
M > 0 such that |f 0 (x)| < M . Then f is uniformly continuous on J.

Proof:We need to estimate |f (x1 ) − f (x2 )| for x1 , x2 ∈ J. Since J is an


interval by the mean-value theorem, there exists c in between x1 and x2 such
that f (x1 ) − f (x2 ) = f 0 (c)(x1 − x2 ), hence |f (x1 ) − f (x2 )| = |f 0 (c)||x1 − x2 | ≤
M |x1 −x2 |. We conclude that f is Lipschitz and hence uniformly continuous.

The mean value theorem is quite useful in proving certain inequalities. For
example.

Result 14.1.3.4 We have ex > 1 + x for all x ∈ R.

Proof:Suppose x > 0. Consider the function f (x) = ex on [0, x]. Since ex


is differentiable on R, we can apply mean value theorem to f on the interval
[0, x]. Hence there exists c ∈ (0, x) such that

ex − e0 = f 0 (c)(c − 0) = ec x.

Note that f 0 (x) = ex > 1 for x > 0. Hence we obtain

ex − 1 = ec x > x.

If x < 0, then considering the interval [x, 0], on similar lines, we obtain
the required inequality.

y−x y−x
Result 14.1.3.5 We have y
< log xy < , 0 < x < y.
x
Proof:Let is a differentiable function on x > 0. Hence using mean value
theorem, there exists c ∈ (x, y) such that

log y − log x = 1e (y − x)

=⇒ log xy = (y − x) 1c .
Since 0 < x < c < y, we have y1 < 1c < x1 . Hence we get
y−x y−x y−x
< log xy = < .
y c x

161
Result 14.1.3.6 If e ≤ a < b, then ab > ba .
Proof:Using the above inequality, we have
b−a b−a
< log ab < .
b a
b
 

a log
ba

Since a log b

< b − a, we have
=e a < eb−a . That is, ba < eb−a aa .
a
aa
If e ≤ a, then et ≤ at for t ≥ 0 and hence we conclude that ba < eb−a aa ≤
ab−a aa = ab .

Theorem 14.1.3.3 (Inverse function theorem) Let f : I := (a, b) → R be


continuously differentiable with f 0 (x) 6= 0 for all x. Then
(i) f is strictly monotone.
(ii) f (I) = J is an interval and
(iii) g = f −1 is differentiable on the interval J := f ((a, b)) and we have

g 0 (f (x)) = 1
f 0 (x)
= 1
f 0 (g(y))
for all x = g(y) ∈ [a, b].

Proof: Since f 0 is continuous on J, by the intermediate value theorem ex-


actly one of the following holds: either f 0 > 0 or f 0 < 0 on I. Hence f is
strictly monotone on I. Hence, the inverse function g is continuous on J.
Fix c ∈ (a, b), let d := f (c). Then
g(y) − g(d) x−c 1
= = f (x)−f (c)
.
y−d f (x) − f (c)
x−c

Since g is continuous, if yn is a sequence in J converging to d, then xn :=


g(yn ) → c. Hence
g(yn ) − g(d) 1 1
lim = lim = .
n→∞ yn − d n→∞ f (xn )−f (c) f 0 (c)
x−c

Since this is true for any sequence {yn } in J converging to d, we conclude


that
g(y) − g(d) 1
lim = 0
y→d y−d f (c)
Hence the proof.

162
Theorem 14.1.3.4 (Cauchy’s form of mean value theorem) Let f, g : [a, b] →
R be continuous on [a, b] differentiable on (a, b). Assume that g 0 (x) 6= 0 for
any x ∈ (a, b). Then there exists c ∈ (a, b) such that
f (b) − f (a) f 0 (c)
= 0 .
g(b) − g(a) g (c)
Proof: Put h(t) = [f (b) − f (a)]g(t) − [g(b) − g(a)]f (t) where a ≤ t ≤
b. Since f and g are continuous on [a, b] h is continuous on [a, b], h is
differentiable in (a, b) and h(a) = f (b)g(a) − f (a)g(b) = h(b). To prove the
theorem, we have to show that h0 (x) = 0 for some x ∈ (a, b).
If h is constant, then h0 (x) = 0 for every x ∈ (a, b). If h(t) > h(a) for
some t ∈ (a, b), let x be a point on [a, b] at which h attains its maximum,
thus x ∈ (a, b) and h0 (x) = 0. If h(t) < h(a) for some t ∈ (a, b), the
same argument applies if we choose for x a point on [a, b] where h attains its
minimum, then h0 (x) = 0. Hence the proof.

Theorem 14.1.3.5 Assume f has a derivative at each point of an open in-


terval (a, b) and that f is continuous on [a, b].
(a) If f 0 takes only positive values in (a, b), then f is strictly increasing on
[a, b].
(b) If f 0 takes only negative values in (a, b), then f is strictly decreasing pn
[a, b].
(c) If f 0 is zero everywhere in (a, b) then f is constant on [a, b].

Proof: Choose x < y and apply the mean value theorem to the sub
interval [x, y] of [a, b] to obtain

f (y) − f (x) = f 0 (c)(y − x) where c ∈ (x, y) −→ (1)

(i) If f 0 > 0, then from (1) f is strictly increasing.

(ii) If f 0 < 0, then from (1) f is strictly decreasing.

(iii) If f 0 = 0, then from (1) f is constant on [a, b].

Hence the proof.

163
14.1.4 Terminal problems:
n
ak X k , n ≥ 2 be a real polynomial. Assume that all
P
1. Let P (X) :=
k=0
the roots of P lie in R. Show that all the roots of its derivative P 0 (X)
also are real.

2. Does there exist a differential function f : R → R such that f 0 (x) = 0


if x < 0 and f 0 (x) = 1 if x > 0 ?

3. Let f : R → R be differentiable such that |f 0 (x)| ≤ M for some M > 0


for all x ∈ R.

(a) Show that f is uniformly continuous on R.

(b) If  > 0 is sufficiently small, then show that the function g (x) :=
x + f (x) is one-one.

4. Let f : (a, b) → R be differentiable at x ∈ (a, b). Prove that

f (x + h) − f (x − h)
lim = f 0 (x).
h→0 2h

Give an example of a function where the limit exists but the function
is not differentiable.

5. Let f : [0, 2] → R be given by f (x) := 2x − x2 . Show that f satisfies
the conditions of Rolle’s theorem. Find a c such that f 0 (c) = 0.

6. Use MVT to establish the following inequalities:


1 1 1
(a) Let b > a > 0. Show that b n − a n < (b − a) n .
(b) Show that | sin x − sin y| ≤ |x − y|.

(c) Show that


nxn−1 (y − x) ≤ y n − xn ≤ ny n−1 (y − x) for 0 ≤ x ≤ y.
(d) Bernoulli’s Inequality. Let α > 0 and h ≥ −1. Then
(1 + h)α ≤ 1 + αh. for 0 < α ≤ 1,

164
(1 + h)α ≥ 1 + αh, for α ≥ 1.

7. Assume that f : (a, b) → R is differentiable on (a, b) except possibly at


c ∈ (a, b). Assume that lim f 0 (x) exists. Prove that f 0 (c) exists and f 0
x→c
is continuous at c.

8. Show that the function f (x) = x3 − 3x2 + 17 is not one-one on the


interval [−1, 1].

9. Prove that the equation x3 − 3x2 + b = 0 has at most one root in the
interval [0, 1].

10. Show that cos x = x3 + x2 + 4x has exactly one root in [0, π/2].

11. Let f (x) = x + 2x2 sin x1 for x 6= 0 and f (0) = 0. Show that f 0 (0) = 1


but f is not monotonic in any interval around 0.

12. Let J be an open interval and f, g : J → R be differentiable. Assume


that f (a) = 0 = f (b) for a, b ∈ J with a < b. Show that f 0 (x) +
f (c)g 0 (c) = 0 for some c ∈ (a, b).

13. Let f, g : R → R be differentiable. Assume that f (0) = g(0) and


f 0 (x) ≤ g 0 (x) for all x ∈ R. Show that f (x) ≤ g(x) for x ≥ 0.

14. Let f : R → R be differentiable. Assume that 1 ≤ f 0 (x) ≤ 2 for x ∈ R


and f (0) = 0. Prove that x ≤ f (x) ≤ 2x for x ≥ 0.

15. Let f, g : R → R be differentiable. Let a ∈ R. Define h(x) = f (x)


for x < a and h(x) = g(x) for x ≥ a. Find necessary and sufficient
conditions which will ensure that h is differentiable at a. (This is a
gluing lemma for differentiable functions.)

16. Let f : [2, 5] → R be continuous and be differentiable on (2, 5). True


or false: f (5) − f (2) = 3.

17. Let f : (0, ∞) → R be differentiable. If f 0 (x) → l as x → ∞, then


f (x)
show that → l as x → ∞.
x

18. Let f : (a, b) → R be differentiable. Assume that lim+ f (x) = lim− f (x).
x→a x→b
Show that there exists c ∈ (a, b) such that f 0 (c) = 0.

165
19. Let f : (0, 1] → R be differentiable with |f 0 (x)| < 1. Define an :=
f (1/n). Show that {an } converges.

20. Let f : [a, b] → R be continuous and differentiable on (a, b). Assume


further that f (a) = f (b) = 0. Prove that for any given λ ∈ R, there
exists c ∈ (a, b) such that f 0 (c) = λf (c).

21. Let f : [a, b] → R be continuous and differentiable on (a, b). Assume


further that f (a) = f (b) = 0. Prove that for any given λ ∈ R, there
exists c ∈ (a, b) such that f 0 (c) + g 0 (c)f (c) = 0.

22. Show that f (x) := x|x| is differentiable for all x ∈ R. What is f 0 (x) ?
Is f 0 continuous ? Does f 00 exist?

23. Let f : R → R be differentiable with f (0) = −3. Assume that f 0 (x) ≤ 5


for x ∈ R. How large can f (2) possibly be?

24. Let f : R → R be differentiable with f (1) = 10 and f 0 (x) ≥ 2 for


1 ≤ x ≤ 4. How small can f (4) possibly be?
(
1/x if x > 0
25. Let f (x) = x1 for x 6= 0 and g(x) =
1 + (1/x) if x < 0 .
0
Let h = f − g. Then h = 0 but h is not a constant. Explain.

26. Let f : [0, 1] → R be continuous. Assume that f 0 (x) 6= 0 for x ∈ (0, 1).
Show that f (0) 6= f (1).

27. Show that on the graph of any quadratic polynomial f the chord join-
ing the points (a, f (a)) and (b, f (b)) is parallel to the tangent line at
the midpoint of a and b.

28. Let n = 2k − 1 ∈ N. Let f (x) = xn for x ∈ R. Show that f maps R


bijectively onto itself.

29. Let f (x) = x2k + ax + b, k ∈ N, a, b ∈ R. Show that f has at most two


zeros in R.

30. Let f : (0, ∞) → R be differentiable. Assume that f (0) = 0 and f 0 is


increasing. Prove that f (x)/x is increasing.

166
14.1.5 Keywords
Local maximum, Local minimum and monotone function.

14.1.6 References
1. W. Rudin, ''Principles of Mathematical analysis'' Third edition. Me
Grow hill Education(India) Private limited, New Delhi.

2. Tom. M. Apostol, '' Mathematical analysis '' Second Edition. Narosa


Publishing House. New Delhi.

3. Ajit kumar and S. Kumaresan, '' A Basic course in Real analysis'',


CRC press.

167
g
Unit XV
The continuity of derivative and L'hospital’s Rule

15.1.1 Objectives
In this unit, we will be studying about intermediate value theorem for deriva-
tives, which is also know as the continuity of derivatives and we will see as
an application of knowing the derivatives to find the limit of ratio of function
i.e L'Hospital's rule.

15.1.2 Introduction
There are few ratio of functions, determining whose limit using the definition
would be impossible ! Knowing the function has a derivative helps us to
evaluate the limit, that rule is known as L' Hospital's rule. Also, we will be
studying about intermediate value property of derivatives.

15.1.3 L'Hospital's rule


As an application to Cauchy mean value theorem, we look at L'hospital's
rules.
Theorem 15.1.3.1 (L'Hospital's Rule) Let J be an open interval. Let either
a ∈ J or a is an end point of J(a = ±∞). Assume that
(i) f, g : J − {a} → R is differentiable,
(ii) g(x) 6= 0 6= g 0 (x) for x ∈ J − {a},
(iii) A := lim f (x) = lim g(x) where A is either 0 or ∞ and
x→a x→a
f 0 (x)
(iv) B := lim 0 exists either in R or B = ±∞.
x→a g (x)
Then
f (x) f 0 (x)
lim = lim 0 =B
x→a g(x) x→a g (x)

Proof: Case 1 : When A = 0, a ∈ R and B ∈ R. Since A = 0, by setting


f (a) = 0 = g(a), the functions f and g are continuous on J. Let {xn } be a
sequence in J such that either xn > a or xn < a for all n ∈ N and xn → a.
By Cauchy’s mean-value theorem, there exists cn between a and xn such that

168
f (xn ) − f (a) f 0 (cn )
= 0 .
g(xn ) − g(a) g (cn )
Since f (a) = 0 = g(a), it follows that
f (xn ) f (xn ) − f (a) f 0 (cn )
= = 0 .
g(xn ) g(xn ) − g(a) g (cn )
f 0 (cn )
Clearly, cn → a(by sandwitch lemma). By hypothesis, the sequence 0 →
g (cn )
f (xn )
B and hence → B. Thus the result.
g(xn )
Case2: When A = ∞, a ∈ R and B ∈ R. Consider h(x) = f (x) − Bg(x),
x ∈ J − {a}. Then h0 (x) = f 0 (x) − Bg 0 (x) so that
h0 (x)
lim = 0.
x→a g 0 (x)

h(x)
We want to show that lim = 0. Let  > 0 be given. Then there
x→a g(x)
0
h (x) 
exists δ1 > 0 such that g(x) > 0 and 0 < for x ∈ (a, a + δ1 ] → (1)
g (x) 2
If x ∈ (a, a + δ1 ), then by Cauchy’s mean value theorem,
h(x) − h(a + δ1 ) h0 (c)
= 0 for some c ∈ (x, a + δ1 )→ (2)
g(x) − g(a + δ1 ) g (c)
From equation (1) and (2), we get

h(x) − h(a + δ1 ) 
g(x) − g(a + δ1 ) . <

2
for x ∈ (a, a + δ1 ) → (3)

Since lim g(x) = ∞, there exists δ2 < δ1 such that


x→a

g(x) > g(a + δ1 ) for x ∈ (a, a + δ2 ) → (4)

From equation (1) and (4), we deduce

0 < g(x) − g(a + δ1 ) < g(x), for x ∈ (a, a + δ2 ) → (5)

From equation (3) and (5), we get

169
|h(x) − h(a + δ1 )| |h(x) − h(a + δ1 )|
< < /2, for x ∈ (a, a + δ2 ) → (6)
g(x) g(x) − g(a + δ1 )
Now choose δ3 < δ2 so that
|h(a + δ1 )|
< /2 for x ∈ (a, a + δ3 ). → (7)
g(x)
We have
h(x) h(x) − h(a + δ1 ) h(a + δ1 )
= + .
g(x) g(x) g(x)
Using this, if x ∈ (a, a + δ3 ), we have

h(x) |h(x) − h(a + δ1 |) |h(a + δ1 )|
g(x) ≤ + . → (8)

g(x) g(x)
Hence by equation (6) and (7)

h(x)
g(x) < , for x ∈ (a, a + δ3 ), → (9)

h(x)
equation (9) says that lim = 0. Since
x→a g(x)

f (x) h(x)
= + B, the result follows. The other cases are similar.
g(x) g(x)

We now discuss few typical applications. The conclusions of these examples


say something about the behavior of functions such as which of the two
goes to zero or to infinity faster/slower. Analysis most often deals with the
comparison of the behavior of functions.

log x
Example 91 f (x) = log x and g(x) = x for x > 0, to calculate lim
x n→∞
Clearly lim f (x) = ∞ and lim g(x) = ∞. Hence we can apply L'hospital's
x→∞ x→∞
rule i.e
f (x) f 0 (x) 1
lim = lim 0 = lim = 0.
x→∞ g(x) x→∞ g (x) x→∞ x

log x
Hence lim = 0.
x→∞ x

170
xn
Example 92 Evaluate lim .
x→∞ ex

Solution: Let f (x) = xn and g(x) = ex . Clearly lim f (x) = ∞ and


x→∞
lim g(x) = ∞ Hence, by L'Hospital's rule
x→∞
xn f (x) f 0 (x) nxn−1
lim = lim = lim = lim .
x→∞ ex x→∞ g(x) x→∞ g 0 (x) x→∞ ex
Now again consider f1 (x) = nxn−1 and g1 (x) = ex .
Also, since lim f1 (x) = ∞ = lim g1 (x), we can again apply L'Hospital's
x→∞ x→∞
xn n(n − 1)xn−1
rule again to obtain lim x = lim .
x→∞ e x→∞ ex
Repeatedly applying L'Hospital's rule n times, we obtain
xn nxn−1 n!
lim x = lim x
= ... = lim x = 0.
x→∞ e x→∞ e x→∞ e
xn
Thus, we obtain lim x = 0.
x→∞ e

−1
−n x
Example 93 To evaluate lim+ x e .
x→0
1 −1
Let f (x) = n and g(x) = e x
x
Then by using L'Hospital's rule repeatedly and the fact lim h(x) = lim+ h( x1 )
x→∞ x→0
−1
−n
We can conclude that lim+ x e x = 0[from (8) and example 92 ]
x→0

Note: L'Hospital's rule also helps us to find out the derivatives of some
function.
(
e−1/x if x > 0
Example 94 Let f (x) := evaluate f 0 (0).
0 if x ≤ 0 .
f (a) − f (0)
Solution: Consider f 0 (0) = lim+
−1
x→0 x −1
e x ex
= lim+ = 0. From example 93. Note to evaluate lim+ , we have used
x→0 x x→0 x
L'Hospital's rule.
( 2
e1/x if x 6= 0
Example 95 Let g(x) := Then evaluate g 0 (0).
0 if x = 0 .

171
2
0 g(x) − g(0) e−1/x
Solution: Consider g (0) = lim+ = lim+ = 0.
x→0 x x→0 x

15.1.4 Intermediate - Value theorem for derivatives:


In the previous block, we have proved that a function f which is continuous
on a compact interval [a, b] assumes every value between its maximum and
its minimum on the interval. In particular, f assumes every value between
f (a) and f (b). A similar result will now be proved for functions which are
derivatives.

Before going to that, we first discuss one-sided derivatives and infinite


derivatives. Up to this point, the statement that f has a derivative at c has
meant that c was interior to an interval in which f was defined and that the
limit defining f 0 (c) was finite. It is convenient to extend the scope of our
ideas some what in order to discuss derivatives at endpoints of intervals. It
is also desirable to introduce infinite derivatives, so that the usual geometric
interpretation of a derivative as the scope of a tangent line will still be valid
in case the tangent line happens to be vertical. In such a case we cannot
prove that f is continuous at c. Therefore, we explicitly require it to be so.

Definition 15.1.4.1 Let f be defined on a closed interval S and assume


that f is continuous at the point c in S. Then f is said to have a right hand
derivative at c if the right hand limit.
f (x) − f (c)
lim+
x→c x−c
exists as a finite value, or if the limit is ±∞. This limit will be denoted by
f+ 0 (c). Left hand derivatives, denoted by f− 0 (c), are similarly defined. In
addition, if c is an interior point of S, then we say that f has the derivative
f 0 (c) = +∞ if both the right and left hand derivatives at c are +∞.

It is clear that f has a derivative at an interior point c if and only if f+ 0 c =


f− 0 c, in which case f+ 0 c = f− 0 c = f 0 (c).

Theorem 15.1.4.1 (Intermediate-value theorem for derivatives) Assume that


f is defined on a compact interval [a, b] and that f has a derivative at each
interior point. Assume also that f has finite one-sided derivatives f+ 0 (a) and
f− 0 (b) at the endpoints, with f+ 0 (a) 6= f− 0 (b). Then, if c is a real number

172
between f+ 0 (a) and f− 0 (b), there exists at least one interior point x such that
f 0 (x) = c.

Proof: Define a new function g as follows:



 f (x) − f (a)
if x 6= a
g(x) = x−a
f 0 (a) if x = a .
+
Then g is continuous on the closed interval [a, b]. By the intermediate-
value theorem for continuous functions, g takes on every value between f+ 0 (a)
f (b) − f (a)
and in the interior (a, b). By the mean value theorem, we have
b−a
g(x) = f 0 (k) for some k in (a, x) whenever x ∈ (a, b). Therefore f 0 takes on
f (b) − f (a)
every value between f+ 0 (a) and in the interior (a, b). A similar
b−a 
 f (x) − f (a)
if x 6= b
argument applied to the function h, defined by h(x) = x−a
f 0 (a) if x = b ,

f (b) − f (a)
shows that f 0 takes on every values between and f− 0 (b) in the in-
b−a
terior (a, b). Combining there results, we see that f 0 takes on every value
between f+ 0 (a) and f− 0 (b) in the interior (a, b), and this proves the theorem.

Note: The above theorem is still valid if one or both of the one-sided
derivatives f+ 0 (a), f− 0 (b) is infinite. The proof in this case can be given by
considering the auxiliary function g defined by the equation g(x) = f (x)−cx,
if x ∈ [a, b].
The above theorem can also be stated as follows which is due to Darboux.

Theorem 15.1.4.2 (Darboux Theorem) Let f : [a, b] → R be differential.


Assume that f 0 (a) < λ < f 0 (b). Then there exists c ∈ (a, b)such that
f 0 (c) = λ.(thus, through f 0 need not be continuous, it enjoys the interme-
diate value property).

Proof: Let f 0 (a) < λ < f 0 (b) and consider g(x) = f (x) − λx.
Then g is differentiable on [a, b]. Note that g 0 (a) < 0 and g 0 (b) > 0.

It attains a global minimum at some point c ∈ [a, b]. We claim that c


cannot be any of the endpoints. For, if c = a, then g(a + h) − g(a) ≥ 0 for

173
h > 0 and hence
g(a + h) − g(a)
g 0 (a) = lim+ ≥ 0.
h→0 h
This contradicts our observation that g 0 (a) < 0. Similarly, if c = b, then
g(b − h) − g(b)
g(a−h)−g(b) ≥ 0 so that the difference quotient ≤ 0. Hence
−h
we conclude that
g(b + h) − g(b)
g 0 (b) = lim− ≤ 0.
h→0 h
This contradicts the fact that g 0 (b) > 0. Hence we conclude that a < c < b.
Hence c is a local minimum for the differentiable function g. It follows that
g 0 (c) = 0, that is f 0 (c) = λ.

The following example illustrate that the Darboux theorem can be applied
to a large class of functions, we look at some functions which are differentiable
and whose derivatives are not continuous.
(
x2 sin x1 if x 6= 0
Example 96 Define g(x) = Then f is differentiable
0 if x = 0 .
(
2x sin x1 − cos x1 if x 6= 0
at all points including 0. Then f 0 (x) = It is
0 if x = 0 .
easy to see that f 0 is not continuous as if we consider xn = nπ 1
, then xn → 0.
But cos xn = cos nπ = (−1) is not convergent. Hence we conclude that f 0 is
n

not continuous at x = 0.

According to Darboux theorem, f 0 enjoys the intermediate value property,


even though it is not continuous.

The intermediate-value theorem shows that derivative cannot change sign


in an interval without taking the value 0.

Theorem 15.1.4.3 Assume f has a derivative on (a, b) and is continuous


at the endpoints a and b. If f 0 (x) 6= 0 for all x in (a, b) then f is strictly
monotonic on [a, b].

174
The intermediate -value theorem also shows that monotonic derivatives are
necessarily continuous.

Theorem 15.1.4.4 Assume f 0 exists and is monotonic on an open interval


(a, b). Then f 0 is continuous on (a, b).
Proof: We assume f 0 has a discontinuity at some point c in (a, b) and arrive
at a contradiction. Choose a closed sub interval [α, P ] of (a, b) which contains
c in its interior. Since f 0 is monotonic on [α, β] the discontinuity at c must
be a jump discontinuity. Hence f 0 omits some value between f 0 (α) and f 0 (β),
contradicting the intermediate value theorem.

15.1.5 Terminal Problems:


1. Evaluate the following limits,
tan 3x
(a) limπ
x→ 2 tan 5x

x−a
(b) lim
x→a ln x − ln a

ln(t + 2)
(c) lim
t→∞ log2 t
√ √
1 − tan x − 1 + tan x
(d) lim
x→π sin 2x
(π/2) − tan−1 x
(e) lim
x→∞ x−1
x4 − 4x
(f) lim
x→0 sin(πx)

1 + tan(x/4)
(g) lim
x→3π cos(x/2)
1
(h) lim x (ln x)
x→∞
x ln x
(i) lim+ ,a>0
x→0 ln(1 + ax)
√ √
x ln x ( ln x)x
(j) lim √ ln x √
x→+∞ ( x) (ln x) x

175
15.1.6 Keywords:
L'Hospital's rule, Intermediate-value theorem and Darboux theorem.

15.1.7 References
1. W. Rudin, ''Principles of Mathematical analysis'' Third edition. Me
Grow hill Education(India) Private limited, New Delhi.

2. Tom. M. Apostol, '' Mathematical analysis '' Second Edition. Narosa


Publishing House. New Delhi.

3. Ajit kumar and S. Kumaresan, '' A Basic course in Real analysis'',


CRC press.

176
g
Unit XVI
Derivatives of higher order and differentiation of vector valued
functions

16.1.1 Objectives:
In this unit, we will be studying about higher-order derivatives, Taylor's
theorem and differentiation of vector-valued functions.

16.1.2 Introduction:
The following question lead us to the content of this unit. Can we differen-
tiate a function which is already a differential? Is there any function which
can be differentiated infinitely many times? Can we write any differentiable
function as a polynomial function ? Can we differentiate the complex-valued
and vector-valued functions in the same way as we differentiate real valued
function? At the end of the unit students will be able to answer all these
questions!

16.1.3 Higher-order derivatives:


Let J be an interval. Let f : J → R be differentiable on J. Since the deriva-
tive at c ∈ J is unique, we have a function f 0 : J → R be given by x → f 0 (x).
Let c ∈ J. If f 0 : J → R is differentiable at c, then we say that f has a
second order derivative at c and it is denoted by f 00 (c) or by f (2) (c). If f 00 (c)
exists for all c ∈ J. The number f 00 (c) is called the second derivative of f at c.

Note that to talk of f being twice differentiable at c, it is not enough if


f is differentiable at c. It has to be differentiable in an interval around c.

Assume that we have defined f (n−1) , (n − 1)th order derivative of f on


J. We then say f is n-times differentiable on J if the function f (n−1) is
differentiable on J. If it exists, we denote the derivative of f (n−1) by f (n) .
The number f n (c) is called the nth derivative of f at c. Note that if f 00 exists
then f 0 is continuous. We say f is C 1 on J and write f ∈ C 0 (J) if f is
differentiable and f 0 is continuous. Such a function is said to be continuously

177
differentiable. If f (n) exists on J and if it is continuous on J, we say that f
is n-times continuously differentiable on J and denote it by f ∈ C n (J). We
also say that f is a C n function or a function of class C n .

Theorem 16.1.3.1 (Leibniz formula ) If h = f g is a product of two func-


tions with derivative up to order n, then
n
n
 (k)
h(n) (x) = f (x)g (n−k) (x)
P
k
k=0

Proof: We prove this using induction on n, the order of derivative. We


know from the product rule of derivative

h0 (x) = f (x)g 0 (x) + f 0 (x)g(x)

= 10 f (0) (x)g (1) (x) + 11 f 0 (x)g (0) (x).


 

This means that the result is true for n = 1.


Let n = 2,

h00 (x) = [f (x)g 0 (x) + f 0 (x)g(x)]0

= f (x)g 00 (x) + 2f 0 (x)g 0 (x) + f 00 (x)g(x)

= 20 f (x)g 00 (x) + 21 f 0 (x)g 0 (x) + 22 f 00 (x)g(x).


  

Thus the result is also true for n = 2. Let us assume that the Leibniz rule is
true for n. We shall prove it for n + 1. Let h = f g with f and g both have
derivatives of order n + 1.

hn+1 (x) = (h(n) )0 (x)


n
n
f (x)g (n−k) (x)]0
P  (k)
=[ k
k=0

n
n
[f (x)g (n−k) (x)]0
P  (k)
= k
k=0

n
n
 (k+1)
(x)g (n−k) (x) + f k (x)g (n−k−1) (x)]
P
= k
[f
k=0

178
n n
n n
 
[f (k+1) (x)g (n−k) (x)]+ [f (k) (x)g (n−k+1) (x)]
P P
= k k
k=0 k=0

n n+1
n n
 (k)  (k)
f (x)g (n−k+1) (x) + f (x)g (n−k+1) (x)
P P
= k k−1
k=0 k=1

n
n n
  (n−k+1)
= g (n+1) (x)+ (x)f (k) (x)+f (n+1) (x)
P
[ k
+ k−1
]g
k=1

n
n+1

= g (n+1) (x) + g(n − k + 1)(x)f (k) (x) + f (n+1) (x)
P
k
k=1

n+1
n+1

g (n+1−k) (x)f (k) (x).
P
= k
k=0

Hence the Leibniz rule holds fo n + 1. Thus the proof of the theorem.

Example 97 Let y = x2 ekx . Then using Leibniz rule we can find y n . For
n > 2,

n
 dn (ekx ) 2 n
 dn−1 (ekx ) n
 dn−2 (ekx )
yn = 0
x + 1
2x + 2
2+0
dxn dxn−1 dxn−2
= k n ekx x2 + 2nk n−1 ekx x + n(n − 1)k n−2 ekx .

16.1.4 Taylor’s theorem


We would like to approximate a function f : R → R by means of polynomial
functions. As noted earlier, if f is differentiable at c, then f is approximately
a polynomial (linear) function near c. That is the equation

f (x) = f (c) + f 0 (c)(x − c)

is approximately correct when x − c is small. Taylor’s theorem tells us that,


more generally, f can be approximated by a polynomial of degree n − 1 if
f has a derivative of order n. Moreover, Taylor’s theorem gives a useful
expression for the error made by this approximation

179
Theorem 16.1.4.1 (Taylor’s theorem) Assume that f : [a, b] → R is such
that f (n) is continuous on [a, b] and f (n+1) (x) exists on (a, b). Fix x0 ∈ [a, b].
Then for each x ∈ [a, b] with x 6= x0 , there exists c between x and x0 such
that
n (x − x )n+1 (x − x0 )n+1 (n+1)
0
f k (x0 ) +
P
f (x) = f (x0 ) + f (c).
k=1 k! (n + 1)!
Proof: Define
n (x − t)k
f k (t) + M (x − t)n+1 ,
P
F (t) := f (t) +
k=1 k!
Where M is chosen so that F (x0 ) = F (x) = f (x).
That is,
n (x − x )k
 
1 P 0 k
M= f (x) − f (x0 ) − f (x0 )
(x − x0 )n+1 k=1 k!
and is possible since x 6= x0 . Clearly, F is continuous on [a, b], differentiable
on (a, b) and F (x) = f (x) = F (x0 ). Hence by Rolle’s theorem applied to
the interval [x, x0 ] and [x0 , x], as the case may be, defined by x and x0 , there
(x − c)n (n+1)
exists c ∈ (a, b) such that 0 = F 0 (c) = f (c) − (n + 1)M (x − c)n .
n!
f n+1 (c)
Thus, M = . Hence
(n + 1)!
n (x − x )k (x − x0 )n+1 (n+1)
0
f k (x0 ) +
P
f (x) = F (x0 ) = f (x0 ) + f (c)
k=1 k! (n + 1)!
Hence the proof of the theorem.

Taylor’s theorem will also be obtained naturally as a consequence of a


more general result that is a direct extension of the generalized Mean-value
theorem, which is as follows.

Theorem 16.1.4.2 Let f and g be two functions having finite nth derivatives
f (n) and g (n) in an open interval (a, b) and continuous (n − 1)th derivatives in
the closed interval [a, b]. Assume that c ∈ [a, b]. Then, for every x in [a, b],
x 6= c, there exists a point x, interior to the interval joining x and c such that

180
P f k (c) P g (k) (c)
 n−1
  n−1

k (n) (n) k
f (x) − (x − c) g (x1 ) = f (x1 ) g(x) − (x − c) .
k=0 k! k=0 k!
Proof: For simplicity, assume that c < b and that x > c. Keep x fixed and
define new functions F and G as follows :
n−1f (k) P g k (t)
n−1
(x − t)k , G(t) = g(t) + (x − t)k ,
P
F (t) = f (t) +
k=1 k! k=1 k!

for each t in [c, x]. Then F and G are continuous on the closed interval [c, x]
and have finite derivatives in the open interval (c, x). Then by generalized
mean value theorem, we can write

F 0 (x1 )[G(x) − G(c)] = G0 (x1 )[F (x) − F (c)] where x1 ∈ (c, x).

This reduces to the equation

F 0 (x1 )[g(x) − G(c)] = G0 (x1 )[f (x) − F (c)] where x1 ∈ (c, x)→ (1)

since G(x) = g(x) and F (x) = f (x). If now, we compute the derivative of the
sum defining F (t), keeping in mind that each term of the sum is a product,
we find that all terms cancel but one and we are left with
(x − t)n−1 n
F 0 (t) = f (t).
(n − 1)!
Similarly, we obtain
(x − t)n−1 n
G0 (t) = g (t)
(n − 1)!
if we put t = x1 and substitute the F 0 (x1 ) and G0 (x1 ) in (1), we obtain the
required result.

Note: Special case of the above theorem, when g(x) = (x − c)n , we have
g k (c) = 0 for 0 ≤ k ≤ n − 1 and g n (x) = n!. This theorem then reduces to
Taylor’s theorem.
We now find the Taylor polynomials for some of standard functions.

Example 98 Let f (x) = xn . Let a ∈ R. Let us look at the Taylor expansion


of f (x) about a. Since it has to be a polynomial in powers of (x − a), we
think of the binomial expansion

181
n
n

xn = (x − a + a)n = an + an−k (x − a)k .
P
k
k=1

Observe that the coefficients of (x − a) are precisely f k (a).


k

Example 99 Consider f (x) = ex . Let us find the nth degree Taylor poly-
nomial of f (x) = ex about x = 0, for all n. Hence the nth degree Taylor’s
polynomial of f (x) = ex about x = 0 is given by
x x2 xn
ex = 1 + 1!
+ 2!
+ ··· + n!
as f n (0) = 1 for all n.
Example 100 Let f (x) = sin x. Then, it is easy to compute that

0,
 when n is even
n
f (0) = 1, when n = 4k + 1, k ∈ N

−1, when n = 4k + 3, k ∈ N

Hence the nth degree Taylor’s polynomial of sin x about x = 0 is given by


n f (n) (0)
xk
P
Tn (x) = f (0) +
k=0 k!
x3 x5 p
− . . .  xp! ,
=x− 3!
+ 5!
( (
n, when n is odd 1, when n = 4k + 1
where p = and  = .
n − 1, when n is even −1, when n = 4k + 3

Example 101 Let f (x) := log(1 + x) for x ∈ (−1, ∞). We shall find the
nth degree Taylor’s polynomial of f about x = 0. We have
1 −1 2
f 0 (x) = , f 00 (x) = 2
, f 000 (x) =
1+x (1 + x) (1 + x)3
By induction, we verify that
(−1)n−1 (n − 1)!
f n (x) = ,
(1 + x)n
thus f n (0) = (−1)n−1 (n − 1)!. Hence the nth degree Taylor’s polynomial of
log(1 + x) about x = 0 is given by
n f n (0)
xk .
P
Tn (x) = f (0) +
k=0 k!
x2 x3 x3 x4 xn
=x− 2
+ 3
− 3
− 4
+ ··· ± n
.

182
16.1.5 Differentiation of vector- valued functions.
Till now, we have defined differentiation of a real-valued functions with real
variable. Now, one can ask a natural question that 'How do we define dif-
ferentiation of a vector - valued function ?' In particular 'How do we define
differentiation of a complex valued function ?'

Definition 16.1.5.1 Let f : (a, b) → Rk be a vector-valued function defined


on an open interval (a, b) in R. Then f can be written as f = (f1 , f2 , . . . , fn )
where each fk is a real-valued function defined on (a, b). We say that f is
differentiable at a point c in (a, b) if each component fk is differentiable at c
and we define

f 0 (c) = (f 0 1 (c), f 0 2 (c), . . . , f 0 n (c)).

Using this definition, we can also define differentiation of a complex-valued


function as complex valued functions can be written as sum of two real-valued
functions.

Definition 16.1.5.2 Let f : (a, b) → C be a function. Then f = f1 + if2


where f1 and f2 are real-valued functions. Hence f is differentiable if and
only if both f1 and f2 are differentiable. Also f 0 = f 0 1 + if 0 2 .

There is also equivalent definition for differentiability of vector -valued


functions.

Definition 16.1.5.3 Let f : [a, b] → Rk , For any x ∈ [a, b] form the quotient
f (t) − f (x)
φ(t) = (a < t < b, t 6= x). Here φ : (a, b) − x → Rk . We say
t−x
f is differentiable if lim |φ(t)| exists and f 0 (x) is that point of Rk for which
t→x
f (t) − f (x)
lim − f 0 (x) = 0.
t→x t−x

Note: f 0 is again a vector-valued function.

Remark: In view of these definitions, it is not surprising to find that


many of the theorems on differentiation are also valid for vector-valued func-
tions. For example, if f and g are vector-valued function differentiable at c
and if λ is a real-valued function differentiable at c and we have

183
(f + g)0 (c) = f 0 (c) + g 0 (c),

(λf )0 (c) = λ0 (c)f (c) + λ(c)f 0 (c)

(f g)0 (c) = f 0 (c).g(c) + f (c)g 0 (c)

There is also a chain rule for differentiating composite functions which is


proved in the same way. If f is vector-valued and if U is real valued, then
the composite function g defined by g(x) = f [U (x)] is vector valued. Then
the chain rule states that

g 0 (c) = f 0 [U (c)]U 0 (c),

If the domain of f contains a neighborhood of U (c) and if U 0 (c) and


f 0 [U (c)] both exists.
Note: It is very interesting to note that Mean-value theorem and one of
its consequences, namely L'Hospital's rule doesn’t holds good in general for
differentiation of vector-valued function, which we will be seeing in the two
in the next two example. These example we take complex-valued function as
it is a particular case of vector-valued functions.

Example 102 Define f : R → C as f (x) = eix = cos x + i sin x. This


function is clearly differentiable as both cos x and sin x are differentiable, but
for this function mean-value property doesn’t holds good. As f (2π) − f (0) =
1 − 1 = 0, but f 0 (x) = ieix so that |f 0 (x)| = 1 for all real x. That is, there
f (2π) − f (0)
is no-real number which satisfies = f 0 (x) for some x ∈ (0, 2π).

Hence Mean-value property doesn’t holds good for vector-valued functions.
i
Example 103 Define f, g : (0, 1) → C by f (x) = x and g(x) = x + x2 e x2
clearly lim f (x) = lim g(x) = 0 and Since |eit | = 1 for all real t, we see that
x→0 x→0
f (x) i
= 1. But g 0 (x) = 1 + 2x − 2ix e x2 (0 < x < 1), so that |g 0 (x)| ≥

lim
x→0 g(x)
0 0
2x − 2i − 1 ≥ 2 − 1. Hence | f (x) | = 0 1 ≥ x and so lim f (x) =

x x |g (x)| 2−x
0 g 0 (x) x→0 g 0 (x)
f (x) 0
lim 0 = 0. i. e lim f (x) 6= lim f (x) .
x→0 g (x) x→0 g(x) x→0 g 0 (x)
Hence the L’Hospital’s rule doesn’t holds for complex-valued functions, in
general for vector-valued functions.

184
However , there is a consequence of the mean-valued theorem for vector-
valued theorem, which we will be proving next.

Theorem 16.1.5.1 Suppose f is a continuous mapping of [a, b] into Rk


and f is differentiable in (a, b). Then there exists x ∈ (a, b) such that
|f (b) − f (a)| ≤ (b − a)|f 0 (x)|.

Proof: Let z = f (b) − f (a) and define φ : [a, b] → R as follows φ(t) =


z.f (t)[dot product]. The mean-value theorem can be applied to φ, therefore
we obtain

φ(b) − φ(a) = (b − a)φ0 (x) where x ∈ (a, b).→ (1)

where φ(b) − φ(a) = z.f (b) − z.f (a) = z.[f (b) − f (a)] = z.z = |z|2 and
(b − a)φ0 (x) = (b − a)z > f 0 (x). Hence, by substituting these values in (1),
we obtain

|z|2 = (b − a)|z.f 0 (x)| [ as |z|2 is real, we must have |z.f 0 (x)|]

|z|2 ≤ (b − a)|z||f 0 (z)| by Schwartz inequality

Hence, |z| ≤ (b − a)|f 0 (x)|, where z=f(b)-f(a)


i.e |f (b) − f (a)| ≤ (b − a)|f 0 (x)| which is the required inequality.

16.1.6 Terminal problems:


1. Let f and g be two functions defined and having finite third-order
derivatives f 00 (x) and g 00 (x) for all x ∈ R. If f (x)g(x) = 1 for all
x, show that the following relations hold at those points where the
denominators are not-zero:
f (x) g 0 (x)
(a) + = 0.
f 0 (x) g(x)
f 00 (x) 2f 0 (x) g 00 (x)
(b) 0 − − 0 = 0.
f (x) f (x) g (x)
f 000 (x) f 0 (x)g 00 (x) f 00 (x) g 000 (x)
(c) 0 −3 − 3 − 0 = 0.
f (x) f (x)g 0 (x) f (x) g (x)
2 2
f 000 (x) 3 f 00 (x) g 000 (x) 3 g 00 (x)
 
(d) 0 − = 0 − .
f (x) 2 f 0 (x) g (x) 2 g 0 (x)

185
(e) Show that f and g have the same Schwarzian derivative if g(x) =
[af (x) + b]/[cf (x) + d], where ad − bc 6= 0.

2. If a vector-valued function f is differentiable at c, then prove that

f 0 (c) = lim h1 [f (c + h) − f (c)].


h→0

conversely, if this limit exists, prove that f is differentiable at c.

3. A vector-valued function f is differentiable at each point of (a, b) and


has constant norm ||f ||. Prove that f (t) · f 0 (t) = 0 on (a, b).

4. A vector-valued function f is never zero and has a derivative f 0 which


exists and is continuous on R. If there is a real function λ such that
f 0 (t) = λ(t)f (t) for all t, prove that there is a positive real function u
and a constant vector c such that f (t) = u(t)c for all t.

16.1.7 Keywords:
Higher order derivatives, Leibniz formula and Taylor’s theorem.

16.1.8 References
1. W. Rudin, ''Principles of Mathematical analysis'' Third edition. Me
Grow hill Education(India) Private limited, New Delhi.

2. Tom. M. Apostol, '' Mathematical analysis '' Second Edition. Narosa


Publishing House. New Delhi.

3. Ajit kumar and S. Kumaresan, '' A Basic course in Real analysis'',


CRC press.

186

You might also like