Chapter 1

Some Basic Ideas in Analysis

1.1 Functions
In general, a function f is a rule which associates with each t in a set A a
value f (t) from a set B. We write

f : A −→ B

t f (t)
f

A B

The set A is called the “Domain” of the function f and the set B is called
the “Codomain” of the function f . For the domains set A and the codomain
set B we shall be considering the following types of sets:

1. a finite set
{1, 2, 3, · · · , N }

2. an infinite discrete set

{1, 2, 3, · · ·}

1
2 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

or
{0, ±1, ±2, ±3, · · · , }
3. a finite interval on the real line
[0, T ]
(The interval can also be taken as open interval or open on one side
and closed on the other side)
4. a semi infinite interval on the real line, such as
[a, +∞] or A = [−∞, a] where a ∈ R
(Again the intervals above can also be open on the right or left respec-
tively)
5. the infinite interval, that is the domain is the set R of real numbers
(−∞, ∞)

In general the domain and codomain can be any arbitrary nonempty sets.
For example, the domain A itself can be a collection of functions, and the
codomain B a collection of sets, or A can be a collection of subsets of a fixed
set Ω, and B the set R of real numbers.
Corresponding to any function f : A −→ B we associate several types of
sets of points, some of which are described below:
1. The domain set A and the codomain set B
2. The set of all values taken by the function , called the “Range of the
function” denoted by Rf . We have
Rf = {f (a) : a ∈ A}
= {b ∈ B 3 ∃a ∈ A such that f (a) = b}

3. For any point b in B consider the set {b}, consisting of the single element
b. We collect all those points in A whose image under f is b, that is,
we define the set
def
S(b) = {a ∈ A : f (a) = b}
1.1. FUNCTIONS 3

S(b)
b

A B
f

It may happen that S(b) is empty for some b ∈ B since there may not
be any a ∈ A such that f (a) = b. Note that the set of all those b ∈ B
for which S(b) is nonempty the Range of the function f

4. Inverse image of a set under f

We next define the inverse image of a set under f . If E ⊆ B is any
subset of B then we define f (−1) (E) as follows:

def
[
f (−1) (E) = S(b)
b∈E
= {a ∈ A : f (a) ∈ E} (1.1.1)

Thus f (−1) (E) is the set of all those points in A whose image under f
fall in the set E. Note that for some sets E the set f (−1) (E) could be
empty.
4 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

f (−1) (E) E

a
f (a)

A B

We classify the following special types of functions:

“one-one” function:
A function f : A −→ B is said to be a one-one function if for every b ∈ B,
S(b) = φ or S(b) is a singleton. This means that

x, y ∈ A and f (x) = f (y) =⇒ x = y

That is, any element b ∈ Rf has exactly one preimage ain A.

“Onto” function;
A function f : A −→ B is said to be a onto function if S(b) 6= φ for every
b ∈ B. This means that

b ∈ B =⇒ ∃a ∈ A 3 f (a) = f (b)

That is, any element b ∈ B has at least one preimage a ∈ A under f .

1-1 correspondence:
A function f : A −→ B is said to be a 1-1 correspondence between A and
B if it is both one-one and onto, that is S(b) is a singleton for every b ∈ B.
This means that distinct elements in the domain A have distinct images in
the codomain B under f , and every element b in the codomain B has exactly
one preimage in the domain A. We can then use this unique preimage to
define a new function f (−1) : B −→ A as

f (−1) (b) = the unique preimage in A of b under f

1.2. CONVERGENCE OF A SEQUENCE OF REAL NUMBERS 5

Remark 1.1.1 Note that the inverse image under any function f of a set
E, namely f −1 (E) is defined for any subset E of the codomain. On the other
hand the inverse function f −1 : B −→ A is defined if and only if the function
f is one-one and onto

1.2 Convergence of a Sequence of Real Num-

bers
An infinite sequence of real numbers is a function

f : N −→ R

The value of this function at n, namely f [n], is a real number which we call
the nth term of the sequence. We also write this sequence as

f [1], f [2], · · · , f [n], · · ·

or as
{f [n]}n
or sometimes as
{fn }
or as
f1 , f2 , · · · , fn , · · ·
We recall the notion of convergence of such a sequence. We have

Definition 1.2.1 An infinite sequence {fn }n , of real numbers, is said to

converge to a real number f if
for every positive real number ε > 0 there exists a positive integer,
(which may depend on ε), Nε such that
n ≥ Nε =⇒ |fn − f | < ε, that is
n ≥ Nε =⇒ f − ε < fn < f + ε

Remark 1.2.1 This says that the terms of the sequence can be made to
differ from f by less than any tolerance threshold ε if the terms are chosen
sufficiently far off in the sequence.
6 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

Remark 1.2.2 The above definition is useful to verify convergence if we

know the limit f ápriori. Otherwise we can use the above definition to show
that the sequence does NOT converge to a real number as follows:
To show that fn does not converge to a we show the following: There exists
a real number ε > 0 such that for every positive integer N there exists a
positive integer n > N for which |fn − a| ≥ ε

We next introduce the notion of Cauchy sequences. This tries to look at

convergence of a sequence without an ápriori knowledge of the limit. We
have seen above that a sequence of real numbers {fn }n∈N converges to a real
number f if the terms of the sequence differ from f by as small an error as
needed, if we look at sufficiently far off in the sequence. If all the far off
terms differ from f by a small error, then the terms far off themselves must
be differing from each other by small error . Formalizing this idea gives us
the notion of Cauchy sequences of real numbers. We have

Definition 1.2.2 A sequence of real numbers {fn }n is said to be a Cauchy

sequence if
for every positive real number ε > 0 there exists a positive integer Nε ,
(which, in general, will depend on ε), such that
n, m ≥ Nε =⇒ |fn − fm | < ε

Remark 1.2.3 It is easy to see that every convergent sequence of real num-
bers must be a Cauchy sequence. The Completeness axiom of the real num-
bers guarantees that every Cauchy sequence of real numbers converges. (For
a Cauchy sequence, for sufficiently large n, the terms of the sequence give a
good approximation of the limit of the sequence)

We next look at the notion of “bounded sequences”.

Definition 1.2.3 A sequence of real numbers {fn }n of real numbers is said

to be bounded
  
above M fn ≤ M
if there exists a real number such that for every n
below m m ≤ fn

A sequence is said to be bounded if it is both bounded below and bounded

above.
1.2. CONVERGENCE OF A SEQUENCE OF REAL NUMBERS 7

Remark 1.2.4 It is not difficult to verify that every convergent sequence

is bounded. However a bounded sequence need not be convergent as de-
mostrated by the sequence
fn = (−1)n+1
Remark 1.2.5 From the above Remark it follows that a necessary condition
for a sequence to converge is that it is bounded. Hence we shall first analyse
bounded sequences.
We now define two important notions, namely the supremum and infimum
of a bounded sequence
Definition 1.2.4 Let {fn }n be a bounded infinite sequence of real numbers.
A real number M0 is said to be the supremum (or the least upper bound) of
the sequence if
1. M0 is an upper bound, that is, fn ≤ M0 for every n, and
2. M0 is the smallest among all upper bounds, that is, if M is any upper
bound for the sequence then M0 ≤ M
Analogously,
A real number m0 is said to be the infimum (or the greatest lower bound) of
the sequence if
1. m0 is a lower bound, that is, m0 ≤ fn for every n, and
2. m0 is the largest among all lower bounds, that is, if m is any lower
bound for the sequence then m ≤ m0
Remark 1.2.6 The Completeness Axiom guarantees that every sequence
bounded above has a (unique) supremum and that every sequence bounded
below has a (unique) infimum
Remark 1.2.7 We have the following characterization of supremum and
infimum of a sequence of real numbers.
1. M0 = sup fn
n
⇐⇒
1) M0 is an upper bound for the sequence, that is,
f [n] ≤ M0 for every n and
2) for every ε > 0 there exists a positive integer nε such that
M0 − ε < fnε ≤ M0
8 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

2. m0 = inf fn
n
⇐⇒
1) m0 is a lower bound for the sequence , that is,
m0 ≤ fn for every n and
2) for every ε > 0 there exists a positive integer nε such that
m0 ≤ fnε ] < m0 + ε

Remark 1.2.8 We also denote the sup fn as lub(fn ) and inf fn as glb(fn )
n n

1.3 Monotone Sequences

We saw above that not every bounded infinite sequence of real numbers con-
verges. We shall now look at a class of bounded sequences which necessarily
converge. These are called monotone sequences.

Definition 1.3.1 An infinite sequence of real numbers {fn }n∈N of real num-
bers is said to be
 
nondecreasing fn ≤ fn+1
if for all n
nonincreasing fn+1 ≤ fn

Such sequences are called monotone sequences

Remark 1.3.1 A nondecreasing sequence is said to be increasing if fn <

fn+1 for all n and a nonincreasing sequence is said to be decreasing if fn+1 <
fn for all n

Remark 1.3.2 A nondecreasing sequence is always bounded below since f1

is a lower bound for the sequence. Hence in order that a nondecreasing
sequence is bounded it is enough if it is bounded above. Similarly,
A nonincreasing sequence is always bounded above since f1 is an upper bound
for the sequence. Hence in order that a nonincreasing sequence is bounded
it is enough if it is bounded below.

Remark 1.3.3 If a nondecreasing sequence {fn }n∈N is bounded above then

its least upper bound is denoted by sup fn
n
If a nondecreasing sequence {fn }n∈N is not bounded above then we define
its least upper bound as sup fn = +∞
n
1.3. MONOTONE SEQUENCES 9

Similarly,
If a nonincreasing sequence {fn }n∈N is bounded below then its greatest lower
bound is denoted by inf fn
n
If a nonincreasing sequence {fn }n∈N is not bounded below then we define its
greatest lower bound as inf an = −∞
n

We shall now look at the convergence of such monotone bounded sequences.

We shall look at a nondecreasing sequence bounded above and then we can
make an analogous statement for a nonincreasing sequence.
Let {fn }n be a nondecreasing sequence bounded above. Since it is bounded
above it has the least upper bound. Let

M0 = sup fn
n

By the characterization of supremum of a sequence (See Remark 1.2.7) we

have
For every ε > 0 there exists a positive integer Nε such that

M0 − ε < fNε ≤ M0 < M0 + ε (1.3.1)

Since the sequence is nondecreasing we have

n ≥ Nε =⇒ fNε < fn ≤ M0 < M0 + ε (1.3.2)

Combininig the above two equations we see that

For every ε > 0 there exists a positive integer Nε such that

n ≥ Nε =⇒ M0 − ε < fn ≤ M0 < M0 + ε (1.3.3)

Hence the sequence converges to M0 . Hence we get the following:

Theorem 1.3.1 Every nondecreasing sequence of real numbers bounded
above converges to its supremum

Remark 1.3.4 For a nondecreasing sequence not bounded above we say

that its limit is +∞

Analogously we have
Theorem 1.3.2 Every nonincreasing sequence of real numbers bounded be-
low converges to its infimum
10 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

Remark 1.3.5 For a nonincreasing sequence not bounded below we say that
its limit is −∞
We next see how the analysis of convergence of a bounded sequence (which
may or may not be monotone) can be converted to that of the analysis of
bounded monotone sequences.
Let {fn }n∈N be a bounded sequence of real numbers. We define a new
sequence {gn }n∈N as follows:
g1 = sup fk = sup {f1 , f2 , · · · , fn , · · ·} = sup fk (1.3.4)
k k≥1
g2 = sup {f2 , · · · , fn , · · ·} = sup fk (1.3.5)
k≥2

and so on and in general

gn = sup {fn , fn+1 , · · ·} = sup fk (1.3.6)
k≥n

This sequence {gn }n∈N is a nonincreasing sequence which is bounded below

and hence by Theorem 1.3.2 it converges to its infimum. Hence the sequence
{gn }n converges to inf gn . We denote this by lim sup fn . Thus we have
n n→∞

lim sup fn = inf gn

n→∞ n
= inf sup fn
n k≥n

Thus we have
Definition 1.3.2 For any bounded sequence of real numbers {fn }n∈N we
define
lim sup fn = inf sup fn (1.3.7)
n→∞ n k≥n

Analogously we define
lim inf fn = sup inf fn (1.3.8)
n→∞ n k≥n

We can show that in general

lim inf fn ≤ lim sup fn (1.3.9)
n→∞ n→∞

The convergent sequences are precisely those for which the equality holds.
We shall state the following theorem without proof.
1.4. GENERALIZATION TO SETS 11

Theorem 1.3.3 A bounded sequence of real numbers {fn }n∈N converges if

and only if

lim inf fn = lim sup fn (1.3.10)

n→∞ n→∞

1.4 Generalization to Sets

We shall now generalize the ideas of limsup and liminf to a sequence of subsets
of a given set. Let Ω be a nonempty set and let C be a collection of subsets
of Ω.
Upper and Lower Bounds
Definition 1.4.1 A subset K of Ω, (not necessarily a member of the collec-
tion C), is said to be an upper bound for the collection C if

A ⊆ K for every A ∈ C (1.4.1)

Note that clearly Ω is an upper bound for any collection C of subsets of Ω.

Similarly,
Definition 1.4.2 A subset L of Ω, (not necessarily a member of the collec-
tion C), is said to be a lower bound for the collection C if

L ⊆ A for every A ∈ C (1.4.2)

Note that clearly the empty set φ is a lower bound for any collection C of
subsets of Ω.
Supremum and Infimum
Definition 1.4.3 A subset K0 of Ω, (not necessarily a member of the col-
lection C), is said to be supremum (or Least Upper Bound) of the collection
C if
1. K0 is an upper bound for C, that is,

A ⊆ K0 for every A ∈ C

and

2. K0 is the least among upper bounds, that is, if K is any upper bound
for C then K0 ⊆ K
12 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

Note that clearly the union of all the sets in C is the supremum for C. Thus
we have
[
sup C = A (1.4.3)
A∈C

Similarly we have,
Definition 1.4.4 A subset L0 of Ω, (not necessarily a member of the collec-
tion C), is said to be infimum (or Greatest Lower Bound) of the collection C
if
1. L0 is an lower bound for C, that is,

L0 ⊆ A for every A ∈ C

and
2. L0 is the largest among lower bounds, that is, if L is any upper bound
for C then L ⊆ L0
Note that clearly the intersection of all the sets in C is the infimum for C.
Thus we have
\
inf C = A (1.4.4)
A∈C

Sequences of subsets of Ω
Analogously we define the following concepts for infinite sequences of subsets
of a set Ω as follows: Let A1 , A2 , · · · , An , · · · be an infinite sequence of subsets
of Ω. Then clearly
∞
[ ∞
\
sup An = An and inf An = An
n n
n=1 n=1

Let us now define a new sequence of subsets as

∞
[
B1 = sup Ak = Ak
k≥1
k=1
[∞
B2 = sup Ak = Ak
k≥2
k=2
1.4. GENERALIZATION TO SETS 13

and in general
∞
[
Bn = sup Ak = Ak
k≥n
k=n

It is easy to see that Bn+1 ⊆ Bn for any n. Hence Bn is a non increasing

sequence and hence we define its limit as the infimum, that is,
∞
def
\
lim Bn = inf Bn = Bn
n→∞ n
n=1

Substituting for Bn we can write this as

∞
[ ∞ [
\ ∞
lim sup Ak = lim Ak = Ak
n→∞ k≥n n→∞
k=n n=1 k=n

This limit is denoted by lim sup An .

n→∞
Analogously we define the sequence of subsets Cn as
∞
\
Cn = inf Ak = Ak
k≥n
k=n

Then the sequence Cn is nondecreasing and we define its limit as its sup,
that is
∞
def
[
lim Cn = sup Cn = Cn
n→∞ n
n=1

Substituting for Cn we can write this as

∞
\ ∞ [
[ ∞
lim inf Ak = lim Ak = Ak
n→∞ k≥n n→∞
k=n n=1 k=n

This limit is denoted by lim inf An .

n→∞
Thus we have
Definition 1.4.5 If {An }∞
n=1 is any infinite sequence of subsets of a set Ω
then
∞ [ ∞
def
\
lim sup An = Ak (1.4.5)
n→∞
n=1 k=n
∞ \ ∞
def
[
lim inf An = Ak (1.4.6)
n→∞
n=1 k=n
14 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

Note that lim sup An and lim inf An are both subsets of Ω
n→∞ n→∞

Remark 1.4.1 Let us now look at the structure of the set lim sup An . We
n→∞
have
∞ [
\ ∞
x ∈ lim sup An ⇐⇒ x ∈ Ak
n→∞
n=1 k=n

∞
[
⇐⇒ x ∈ Ak for every n
k=n
⇐⇒ for every n there exists a k ≥ n such that x ∈ Ak
⇐⇒ however far we look in the sequence there exists a set
in the sequence to which x belongs
⇐⇒ x is an element of infinitely many of the An sets
Thus we have
lim sup An is the set consisting precisely of those elements that are
n→∞
elements in infinitely many of the An sets
Similarly we can see that
lim inf An is the set consisting precisely of those elements that are
n→∞
elements of all the An sets beyond a certain stage

1.5 Some Simple Tests For Convergence

We shall now look at simple methods by which we can test the convergence
of a sequence of real numbers.
We first observe that if a sequence of real numbers {fn }n∈N is such that the
sequence of real numbers {|fn |}n converges, it does not necessarily mean that
the sequence of real numbers {fn } converges . (For example the sequence
deefined as fn = (−1)n does not converge whereas the sequence of absolure
values |fn | = 1 converges to 1). However, the sequence of real numbers
{fn }n∈N convergence to zero if and only if the sequence of absolute values
{|fn |}n∈N also converges to zero. We have
Proposition 1.5.1 The sequence of real numbers {fn }n∈N converges to zero
⇐⇒
The sequence of real numbers {|fn |}n converges to zero
1.5. SOME SIMPLE TESTS FOR CONVERGENCE 15

We next observe that for a sequence of real numbers {fn }n∈N the following
holds:

Proposition 1.5.2

{fn }n∈N converges to f ⇐⇒ the sequence {fn − f }n∈N converges to 0

In view of the above Proposition 1.5.2 it follows that we need tests for con-
vergence to zero. We shall now look at such simple tests.
We shall first state a useful theorem (without proof):

Theorem 1.5.1 Sandwich Theorem

Suppose {fn }n∈N , {gn }n∈N and {hn }n∈N are sequences of real numbers such
that

1. {gn }n∈N and {hn }n∈N both converge to the same limit `, and

2. gn ≤ fn ≤ hn for all n

Then
The sequence {fn }n∈N also converges to `

Remark 1.5.1 Since convergence depends only on how the tail of the se-
quence behaves, the above theorem holds even if we replace condition (2)
above by the following:
There exists a positive integer N0 such that

gn ≤ fn ≤ hn for all n ≥ N0

Remark 1.5.2 In particular, if {fn }n∈N is a sequence of nonnegative real

numbers, (that is fn ≥ 0 for all n), and {hn }n∈N is a sequence of nonnegative
real numbers converging to zero such that fn ≤ hn for all n then by taking
gn = 0 for all n and applying the Sandwich Theorem 1.5.1 we get that the
sequence {fn }n also converges to zero.

Remark 1.5.3 In the above Remark we can also replace fn ≤ hn byfn ≤ Khn
where K is some positive constant.

We shall therefore look at some simple sequences that converge to zero that
can be used to apply the Sandwich Theorem
16 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

1
1. The sequence of real numbers defined as an = converges
np
to zero if and only if p > 0

Example 1.5.1 We shall now use this to show that the sequence de-
fined as
n2 + 1
fn = 3
n +4
converges to zero.
We have
n2 + 1
fn =
n3 + 4
=⇒
2n2
fn ≤ since 1 ≤ n2 for all n
n3 + 4
2n2 2 1 1
≤ = since ≤ for all n
n3 n n3 + 4 n3

Hence we get
2
0 ≤ fn ≤
n
1
Now by choosing hn = n and using the Remark 1.5.3 we get the se-
quence {fn }n also converges to zero.

2. In general the sequence defined as

P (n)
fn =
Q(n)
where P (n) and Q(n) are polynomials in n, converges to zero
if degree of P (n) < degree of Q(n)

Example 1.5.2 We apply this to the sequence in Example 1.5.1 We

have
P (n) = n2 + 1
Q(n) = n3 + 4
degree of P (n) = 2
degree of Q(n) = 3
1.5. SOME SIMPLE TESTS FOR CONVERGENCE 17

Since degree of P (n) < degree of Q(n) we have that the given se-
quence converges to zero

F (n)
3. More generally, if fn = where F (n) and G(n) are ex-
G(n)
pressions each term of which is a power of n, then the sequence
converges to zero if

highest exponent of n in F (n) < highest exponent of n in G(n)

Example 1.5.3 Consider the sequence defined as

n2 + 1
fn = √ 5
( n) + 1

Then we have

F (n) = n2 + 1
√
G(n) = ( n)5 + 1
highest exponent of n in F (n) = 2
5
highest exponent of n in G(n) =
2

Since

highest exponent of n in F (n) < highest exponent of n in G(n)

the sequence converges to zero

F (n)
4. If an = where F (n) and G(n) are expressions each term
G(n)
of which is a power of n, and if

highest exponent of n in F (n) = highest exponent of n in G(n)

then the sequence converges to the ratio of the coefficients of

the highest exponent terms in the numerator and denominator
respectively
18 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

Example 1.5.4 The sequence defined as

√
3( n)5 − n2 + 1
fn = √ 5
(2 n) − 3n − 1
3
converges to
2
5. A very useful sequence of real numbers that converges is the geometric
sequence. We have
The sequence {r n−1 }n∈N coverges to zero if 0 ≤ r < 1
We can easily see this as follows:
Let ε > 0 (Without loss of generality we assume ε < 1. Let Nε be a
positive integer such that
`n(ε)
Nε >
`n(r)
Then we have
`n(ε)
n ≥ Nε =⇒ n >
`n(r)
=⇒ n`n(r) < `n(ε) (since `n(r) < 0)
=⇒ `n(rn ) < `n(ε)
=⇒ rn < ε
Thus for any ε > 0 we can find a positive integer Nε such that
n ≥ ε =⇒ −ε < 0 ≤ rn < ε
Hence fn = rn−1 converges to zero
Remark 1.5.4 Using Proposition 1.5.2 we can now see that
the sequence of real numbers {r n−1 }n ∈ N converges if |r| < 1

Remark 1.5.5 The sequence fn = r n−1 does not converge if

|r| > 1. When |r| = 1 it converges if r = 1 and does not
converge if r = −1
 n   n 
1 1
Example 1.5.5 The sequences and − both con-
2 n 2 n
verge.
The sequence {(−1)n }n does not converge whereas the sequence {(1)n }n
converges.
The sequence {2n }n dose not converge.
1.6. EXAMPLE 19

1.6 Example
Let us consider the following sequence of real numbers:
3n2 + 2
fn =
4n2 + 3
We first observe that
3n2 + 2
fn =
4n2 + 3
n2 (3 + 2n−2 )
=
n2 (4 + 3n−2 )
3 + 2n−2
=
4 + 3n−2
We look at some of the terms of the sequence
n fn

3.0002
102
4.0003

3.000002
103
4.000003

3.00000002
104
4.00000003

1. It appears from the above table of values of fn that the terms fn are
3
getting closer to as n increases and hence intuitively we see the limit
4
3
as . We shall now prove, using the definition of convergence, that the
4
3
sequence indeed converges to .
4
In order to do this we need to show the following:
We must show that for any given ε > 0 we can find a positive integer
Nε such that  
 3 
n ≥ Nε =⇒ fn −  < ε
4
20 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS
 
Let us first estimate the error fn − 34 . We have

3 3n2 + 2 3
fn − = −
4 4n2 + 3 4
(12n2 + 8) − (12n2 + 9)
=
4(4n2 + 3)
−1
=
4(4n2 + 3)

Thus
3 −1
fn − = (1.6.1)
4 4(4n2 + 3)

Hence we get
3 1
|fn − | =
4 4(4n2 + 3)
1
≤ 2
(4n + 3)
1
≤
4n2
1
≤
n2
 
 3 
Hence fn −  will be < ε if

4
1
< ε, that is if,
n2
1
n> √
ε
1
Hence given any ε > 0 we choose a positive integer Nε > √ . Then
ε
 
 3  1 1
n ≥ Nε =⇒ fn −  < 2 ≤ 2 < ε
4 n Nε
3
Hence the sequence converges to f =
4
1.6. EXAMPLE 21

2. Next we shall verify that the sequence is a Cauchy sequence.

In order to do this we need to show the following:
We must show that for any given ε > 0 we can find a positive integer
Nε such that
m, n ≥ Nε =⇒ |fn − fm | < ε
Let us first estimate |fn − fm |. Without loss of generality let us assume
m > n. We have
3n2 + 2 3m2 + 2
fn − fm = −
4n2 + 3 4m2 + 3
(12n2 m2 + 9n2 + 8m2 + 6) − (12n2 m2 + 8n2 + 9m2 + 6)
=
(4n2 + 3)(4m2 + 3)
n2 − m2
=
(4n2 + 3)(4m2 + 3)

Hence we get

m2 − n2
|fn − fm | =
(4n2 + 3)(4m2 + 3)
m2
≤
(4n2 + 3)(4m2 + 3)
1
≤ 2
4n + 3
1
≤
n2

1
Hence as above given any ε > 0 we choose a positive integer Nε > √ .
ε
Then
1 1
n ≥ Nε =⇒ |fn − fm | < 2 ≤ 2 < ε
n Nε
Thus the given sequence is a Cauchy sequence

3. Next we shall verify whether the sequence is a monotone sequence. The

sequence will be nondecreasing if

fn ≤ fn+1
22 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

This will be true if

3n2 + 2 3(n + 1)2 + 2
≤
4n2 + 3 4(n + 1)2 + 3
if
12n2 (n + 1)2 + 9n2 + 8(n + 1)2 + 6 ≤ 12n2 (n + 1)2 + 8n2 + 9(n + 1)2 + 6
if
n2 ≤ (n + 1)2
which is true. Hence the sequence is nondecreasing.
4. Since the sequence is nondecreasing and bounded it must converge to
its lub. But we have already proved in (1) above that the sequence
3 3
converges to . Hence must be its lub. We shall now directly verify
4 4
3
that is indeed its lub.
4
In order to show that 43 is the lub we must show the following:
(a) it is an upper bound, that is,
3
fn ≤ for all n
4
and
(b) it is the least among all upper bounds, that is,
3
for every ε > 0 there exists a term fN of the sequence such that fN > 4
−ε

We prove (a) as follows:

3 3n2 + 2 3
fn ≤ if 2
≤
4 4n + 3 4
This is true if
12n2 + 8 ≤ 12n2 + 9
3
which is obviously true for all n. Hence is an upper bound.
4
We prove (b) as follows:
We have observed in equation 1.6.1 above that
3 −1
fn − =
4 4(4n2 + 3)
1.7. SEQUENCES OF FUNCTIONS 23

We want to find n such that

3
fn ≥ −ε
4
that is we want to find n such that
3
fn − ≥ −ε
4
Substituting for fn − 34 from equation 1.6.1 we get that we want to find
n such that
−1
≥ −ε
4(4n2 + 3)
that is
1
≤ε
4(4n2 + 3)
1
Now as in (1) we can choose N to be any positive integer > √ and
ε
we will get
3
fN > − ε
4
This proves (b).
3
Hence we see that is the lub of the sequence.
4

1.7 Sequences of Functions

We shall now investigate the idea of convergence of a sequence of functions.
Let
fn : D −→ R, n = 1, 2, 3, · · ·
be a sequence of real valued functions defined on a set D. Without loss of
generality, for illustration, let us assume that the set D is an interval I on
the real line. Hence we have

fn : I −→ R

If we now look at the graphs of these functions, each function gives rise to
a path in the plane and we would like to see if these paths approach some
fixed path as n → ∞.
24 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

Example 1.7.1 Let I = [0, 1], the closed interval in R. Let the sequence of
functions be defined as
x
fn (x) =
n
If we now plot these functions we get a family of straight lines.

fn (x)

f1
1

f2
f3
f10
x
0 1

From the figure above it appears that these curves (straight lines) represent-
ing the functions fn (x) are approaching the zero function

f (x) = 0 for all x

We want to now give a mathematical meaning to this type of convergence of a

sequence of functions. We first introduce the notion of pointwise convergence
which is the first simple way of giving meaning to this convergence.
Consider a sequence of real valued functions

fn : D −→ R n = 1, 2, 3, · · ·

and a real valued function

f : D −→ R
For each fixed x0 ∈ D, the sequence of values {fn (x0 )}n∈N of the values of the
functions at the point x0 form a sequence of real numbers. If this sequence
of real numbers converges to the real number f (x0 ), (namely the value of
the function at the point x0 ) we say that the given sequence of real valued
functions converges to f at the point x0 . Thus we say
A sequence of real valued functions fn : D −→ R converges to the
real valued function f : D −→ R at a point x0 ∈ D, if
1.7. SEQUENCES OF FUNCTIONS 25

for every ε > 0 there exists a positive integer N0 (ε), (which will,
in general, depend not only on ε but also on the point x0 ), such
that
n ≥ N0 (ε) =⇒ |fn (x0 ) − f( x0 )| < ε
If the sequence of functions converges at every point x ∈ D to f (x) we then
say that the sequence of real valued functions fn converges pointwise to the
function f . Hence we have the following definition:

Definition 1.7.1 A sequence of real valued functions fn : D −→ R

is said to converge to a real valued function f : D −→ R if
for every fixed every x ∈ D, for any ε > 0, there exists a positive
integer Nx (ε), (which will, in general, depend on the point x), such
that
n ≥ Nx (ε) =⇒ |fn (x) − f (x)| < ε
We then write
pw (D)
fn −→ f
x
Example 1.7.2 Consider the sequence, fn (x) = on the interval [0, 1], of
n
Example 1.7.1
For x = 0 we have fn (0) = 0 for all n and hence fn (x) converges to 0 at
x = 0.
For any fixed x 6= 0 and 0 < x ≤ 1 we have
x
0 ≤ fn (x) =
n
1
≤
n
−→ 0

Hence this sequence of real valued functions converges pointwise to the zero
function. We have
pw [0,1]
fn −→ 0
We shall see this by using the ε definition of point wise convergence.
For x = 0 we already have fn (0) = 0 for all n. Hence for any ε > 0 we can
take Nε = 1 and we have

n ≥ Nε =⇒ |fn (0)| = 0 < ε

26 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

Hence we have convergence to zero at the point x = 0

For x 6= 0 we have 0 < x ≤ 1 and
x
|fn (x)| =
n
1
≤ since 0 < x ≤ 1
n

1
For any ε > 0 choose Nε to be a positive integer > . Then we have from
ε
above
1
|fn (x)| ≤
n
1
≤ for all n ≥ Nε
Nε
< ε by the choice of Nε

Hence fn (x) converges to zero for all x ∈ [0, 1]. Thus we have
pw [0,1]
fn −→ 0

Example 1.7.3 By a similar argument as in Example 1.7.2 we get

pw [0,∞)
fn −→ 0

Example 1.7.4 Consider the sequence of real valued functions

fn : [0, 1] −→ R

defined as
1
fn (x) =
n+x
We have for any x ∈ [0, 1],
1
0 ≤ |fn (x)| =
n+x
1
≤
n
We have
pw [0,1]
fn −→ 0
1.7. SEQUENCES OF FUNCTIONS 27

By a similar argument we also get

pw [0,A]
fn −→ 0 for any positive real number A

and
pw [0,∞)
fn −→ 0

Let I be an interval in R. Suppose a sequence of real valued functions

fn : I −→ R converge pointwise to a function f : I −→ R and the functions
fn (x) of the sequence all have some nice property, (if not all fn at least all fn
beyond a certain stage). Then will the limit function f (x) also possess this
nice property - in other words does this pointwise limiting property preserves
nice properties of the functions?
Which nice properties are of interest?
For example we would be interested in the answers to the following questions?
Suppose
pw I
fn −→ f
We list below some fundamental questions:

1. If all fn , (if not all fn at least all fn beyond a certain stage), are bounded
functions then will f be bounded - that is, does pointwise convergence
preserve boundedness?

2. If all fn , (if not all fn at least all fn beyond a certain stage), are
continuous functions in I then will f be continuous in I - that is, does
pointwise convergence preserve continuity?

3. If all fn , (if not all fn at least all fn beyond a certain stage), are
differentiable functions in I then will f be differentiable in I, and if so
will the derivative f 0 of the limit function be the limit of the sequence
of derivatives, {fn0 }n∈N - that is, does pointwise convergence preserve
differentiability?

4. If all fn , (if not all fn at least all fn beyond a certain stage), are
integrable over I then will f be integrable over I and is
Z Z
lim fn (x)dx = lim fn (x)dx
n→∞ I I n→∞
28 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

- that is, will the following hold:

Z Z
lim fn (x)dx = f (x)dx
n→∞ I I

that is, does pointwise convergence preserve integrals?

The answer to all these questions is, in general, these properties are not
preserved by pointwise convergence. We shall illustrate this aspect by a few
examples.
Example 1.7.5 Consider the sequence of real valued functions fn : (0, 1) −→ R
defined as
n
fn (x) =
1 + nx
We have, for any x ∈ (0, 1),
n
fn (x) =
1 + nx
1
= 1
n
+x
1
−→
x
Thus we have
pw (0,1)
fn −→ f (x)
where f : (0, 1) −→ R is the function defined as
1
f (x) =
x
We have for any x ∈ (0, 1), for every n,
 
 n 
|fn (x)| = 
  ≤ n since 1 + nx ≥ 1
1 + nx 
Hence, for every n

|fn (x)| ≤ n for every x ∈ (0, 1)

Hence for every n, fn is a bounded function on (0, 1). However the limit
function f (x) is not a bounded function on (0, 1). Thus we see that pointwise
convergence does not, in general, preserved boundedness.
1.7. SEQUENCES OF FUNCTIONS 29

Example 1.7.6 Consider the sequence of real valued functions fn : [0, 1] −→ R

defined as
fn (x) = xn
Clearly, we have
pw [0,1]
fn −→ f (x)
where f : [0, 1] −→ R is the function defined as

0 if 0 ≤ x < 1
f (x) =
1 if x = 1
We see that for every n, fn is a continuous function on [0, 1] but the limit
function f (x) is not continuous at x = 1. Thus we see that pointwise con-
vergence does not, in general, preserve continuity.
Example 1.7.7 In the above example, we see that for every n, fn is a dif-
ferenttiable function on [0, 1] but the limit function f (x) is not differentiable
at x = 1 - it is not even continuous at x = 1. Thus we see that pointwise
convergence does not, in general, preserve differentiability.
Example 1.7.8 In the above example the limit function f was not even
differentiable whereas all the fn were. Suppose the limit function is differ-
entiable, will the derivetaive of the limit be the limit of the derivatives? We
shall now show, in this example, that even this may not happen in general.
Consider the sequence of real valued functions fn : R −→ R defined as
x
fn (x) =
1 + nx2
For x = 0 we have fn (0) = 0 for every n and hence fn converges to 0 at
x = 0.
For x 6= 0 we have
 
 x 
0 ≤ |fn (x)| =  
1 + nx2 
|x|
=  
1 2
n +x
n
1 |x|
=
n 1 + x2
−→ 0 for every x 6= 0
30 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

Hence we have
pw (R)
fn −→ 0
Further we see that for every n, fn is a differentiable function, and
(1 + nx2 )(1) − x(2nx)
fn0 (x) =
(1 + nx2 )2
1 − nx2
=
(1 + nx2 )2
For x = 0 we have
fn (x) − fn (0)
fn0 (0) = lim
x→0 x
fn (x)
= lim (since fn (0) = 0)
x→0 x
1
= lim
x→0 1 + nx2
= 1

Thus

fn0 (0) = 1 for all n

−→ 1

For x 6= 0 we have
1 − nx2
fn0 (x) =
(1 + nx2 )2
 2

n n1 − xn
=
n2 ( n1 + x2 )2
1 2
1 n
− xn
=
n 1 2 2


n
+ x
−→ 0

Thus the derivative sequence fn0 (x) is such that

pw R
fn0 −→ g
1.7. SEQUENCES OF FUNCTIONS 31

where g : R −→ R is the function defined as


1 if x = 0
g(x) =
0 if x 6= 0
On the other hand since the limit function f of the sequence fn is the zero
function we have
f 0 (x) = 0 for every x
Hence f 0 (x) 6= g(x) since f 0 (0) 6= g(0)
Thus we see that even if the derivative sequence converges pointwise to a
function it need not be the derivative of the limit function.
Example 1.7.9 Consider the sequence of real valued functions fn : [0, ∞) −→ R
be the functions whose graph is as shown below:
fn (x)

1
n
n x
0

We have
1

n
if 0 ≤ x ≤ n
fn (x) =
0 if x > n
It is easy to see that
pw [0,∞)
fn −→ 0
For every n the function fn is integrable over [0, ∞) and
Z ∞
fn (x)dx = 1
0
On the other hand the limit function f being the zero function we have
Z ∞
f (x)dx = 0
0
Thus Z ∞ Z ∞
lim fn (x)dx 6= lim fn (x)dx
n−→∞ 0 0 n−→

Thus we see that pointwise convergence does not, ingeneral, preserve integrals
32 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

We shall next introduce different notions of convergence of sequence of func-

tions where some of these defects are taken care of.

1.8 More Notions of Convergence of Sequences

of Real Valued Functions
We shall next introduce other notions of convergence of a sequence of func-
tions. In all these notions of convergence, we get the respective notions based
on how we try to control the error between the functions fn in the sequence
and the limit function f . We shall first introduce the notion of uniform
convergence. We shall begin with the two simple examples.
Example 1.8.1 Let N be a fixed positive integer and let D be the finite set,
D = {1, 2, 3, · · · , N }
Let fn and f be the functions defined as follows;
x
fn (x) = for all x ∈ D
n
f (x) = 0 for all x ∈ D
We have tabulated these functions below :
x = 1 x = 2 = x3 · · · x = N

2 3 N
f1 (x)(x) 1 1 1
··· 1

1 2 3 N
f2 (x) 2 2 2
··· 2

1 2 3 N
f3 (x) 3 3 3
··· 3

.. .. .. .. .. ..
. . . . . .

1 2 3 N
fn (x) n n n
··· n

.. .. .. .. .. ..
. . . . . .
1.8. MORE NOTIONS OF CONVERGENCE OF SEQUENCES OF REAL VALUED FUNCTIONS33

In order to look at the convergence of fn at a point x ∈ D we have to look at

the x-th column, (in the table above), which gives us the sequence of values
f1 (x), f2 (x), f3 (x), · · · , fn (x), · · ·. From the table it follows that

lim fn (x) = 0 for every x ∈ D

n−→∞

Hence the sequence converges to the zero function pointwise, that is

pw (D)
fn −→ 0

Consider any ε > 0. For the sake of illustration let us take ε = 10−4 and we
choose Nε to be a positive integer > xε . Then

x
|fn (x) − f (x)| = < ε = 10−4 if n ≥ Nε > x(10)4
n

Thus we have a single stage Nε beyond which

|fn (x) − f (x)| < 10−4 for all x ∈ D

Thus, in this example we see that there exists a single positive integer Nε
which depends only on ε and not on the point x ∈ D such that the above
inequality holds beyond this Nε .

Example 1.8.2 Consider now the set D = N = {1, 2, 3, · · ·}, the infinite
set of all positive integers, and as in the above example let the sequence
fn : D −→ R be defined as

x
fn (x) = for all x ∈ D
n
34 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

These functions are tabulated below:

x = 1 x = 2 x = 3 ··· x = N ··· ···

2 3 N
f1 (x)(x) 1 1 1
··· n
··· ···

1 2 3 N
f2 (x) 2 2 2
··· 2
··· ···

1 2 3 N
f3 (x) 3 3 3
··· 3
··· ···

.. .. .. .. .. .. .. ..
. . . . . . . .

1 2 3 N
fn (x) n n n
··· n
··· ···

.. .. .. .. .. .. .. ..
. . . . . . . .
From the table it follows that
lim fn (x) = 0 (= f (x)) for all x ∈ D
n−→∞

Hence we get
pw (D)
fn −→ 0
We observe that, in this example, for any ε > 0, we need different Nε at
different points in D to make |fn (x) − f (x)| < ε - and no single positive
integer Nε works for all x ∈ D - whereas in the previous example we could
get one Nε that worked for all points.
We next distinguish these two cases by introducing the notion of uniform
convergence.
Definition 1.8.1 . A sequence of real valued functions, fn : D −→ R,
is said to converge “uniformly” to the function, f : D −→ R on D,
if
For every ε > 0 there exists a positive integer Nε (depending only
on ε) such that
n ≥ Nε =⇒ |fn (x) − f (x)| < ε for all x ∈ D
1.8. MORE NOTIONS OF CONVERGENCE OF SEQUENCES OF REAL VALUED FUNCTIONS35

Thus the sequence in Example 1.8.1 converges uniformly to the zero func-
tion while the sequence in the Example 1.8.2 does not converge to the zero
function uniformly, even though it converges pointwise to the zero function.
If a sequence of functions fn converges uniformly to a function f we then
write,
uc (D)
fn −→ f

Remark 1.8.1 It is easy to see that

uc (D) pw (D)
fn −→ f =⇒ fn −→ f

On the other hand, from Example 1.8.2, we see that

pw (D) uc (D)
fn −→ f 6⇒ fn −→ f

uc (D) uc (D)
Remark 1.8.2 It is easy to see that fn −→ f is equivalent to fn −f −→ 0.
It is, therefore, enough if we know how to test for uniform convergence to
the zero function.

We shall therefore look at a simple test for uniform convergence of a sequence

of functions to the zero function. This is similar to the Sandwich Theorem
(Theorem 1.5.1).
Weierstrass’ Mn Test:
Let fn : D −→ R be a sequence of real valued functions such that
there exists a sequence {Mn } of nonnegative real numbers such
that,

1. For every n, |fn (x)| ≤ Mn for all x ∈ D and

2. Mn −→ 0
uc (D)
Then fn −→ 0

Example 1.8.3 Consider D = [−1, 1] and the sequence fn : D −→ R de-

fined as
x2 + x + 1
fn (x) =
n2 + 1
36 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

We shall use Weierstrass’ Test to show that this sequence converges to 0

uniformly on D. We have
|x2 + x + 1| ≤ |x2 | + |x| + 1
≤ 1 + 1 + 1 for all x ∈ D, and
2
n +1 ≥ n2 and hence
1 1
2
≤
n +1 n2
Thus we have
|x2 + x + 1|
|fn (x)| =
n2 + 1
3
≤
n2
Now if we choose
3
Mn =
n2
and we have Mn −→ 0. Hence all the conditions of Weierstrass’ Test are
uc (D)
satisfied and we get therefore fn −→ 0

Example 1.8.4 Consider the sequence fn : [0, 1] −→ R defined as

x2 + n2 x + x
fn (x) =
n2 + 1
We shall show that this sequence converges uniformly to the function
f : [0, 1] −→ R
defined as f (x) = x. This means that we have to show that the sequence
fn (x) − f (x) converges uniformly to the zero function. We have
 2
 x + n2 x + x


|fn (x) − f (x)| =  − x 
n2 + 1 
 2
 x + n2 x + x − n2 x − x 

=  
n2 + 1 
x2
=
n2 + 1
1
≤ 2
since x ∈ [0, 1]
n +1
1.8. MORE NOTIONS OF CONVERGENCE OF SEQUENCES OF REAL VALUED FUNCTIONS37

1
Choosing Mn = n2 +1
and applying Weierstrass’ test we get

uc (I)
fn − f −→ 0

and hence
uc (I)
fn −→ f

We shall next see that uniform convergence “preserves” “nice” properties of

the functions of the sequence.
Without loss of generality we shall consider that the functions are all defined
in an interval I in R
1. Uniform convergence preserves boundedness
uc (I)
fn : I −→ R be all bounded functions and fn −→ f
=⇒
f is bounded
Proof:
fn bounded =⇒ For every n there exists a positive real number Mn
such that

|fn (x)| ≤ Mn for every x ∈ I

and hence

−Mn ≤ fn (x) ≤ Mn for every x ∈ I (1.8.1)

uc (I)
fn −→ f =⇒
For every ε > 0 there exists a positive integer Nε such that

n ≥ Nε =⇒ |fn (x) − f (x)| < ε for all x ∈ I (1.8.2)

Choosing ε = 1 and n = N1 , (the positive integer corresponding to

ε = 1) we get from equation 1.8.2

|fN1 (x) − f (x)| < 1 for all x ∈ I

=⇒ fN1 (x) − 1 < f (x) < fN1 + 1 for all x ∈ I
=⇒ −MN1 − 1 < f (x) < MN1 + 1 (by equation 1.8.1)
=⇒ f is bounded
38 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

2. Uniform convergence preserves continuity

uc (I)
fn : I −→ R be all continuous functions on I and fn −→ f
=⇒
f is continuous on I
Proof:
We have to show that f (x) is continuous at every point a ∈ I, that is,
we have to show that
for every a ∈ I, the error |f (x) − f (a)| can be made as small as we
please by choosing x sufficiently close to the point a
This means we have to show that, given a ∈ I,
 
for every ε > 0 there exists a δε (a) > 0 such that
(1.8.3)
|x − a| < δ =⇒ |f (x) − f (a)| < ε
We show equation 1.8.3 as follows:
uc (I)
fn −→ f =⇒
for every ε > 0 there exists a positive integer Nε such that

n ≥ Nε =⇒ |fn (x) − f (x)| < ε for every x ∈ I (1.8.4)

This gives

|fNε (x) − f (x)| < ε for every x ∈ I (1.8.5)

Since all the fn (x) are given to be continuous on I we have fNε (x) is
continuous on I
Hence fNε (x) is continuous at x = a
ε
Thus given ε > 0 choose ε1 =
3
By the continuity of fNε at the point a, we get a δε1 (a) > 0 (and hence
δε (a)) such that

|x − a| < δε (a) =⇒ |fNε (x) − fNε (a)| < ε1 (1.8.6)

Then we have

|f (x) − f (a)| = |f (x) − fNε (x) + fNε (x) − fNε (a) + fNε (a) − f (a)|
≤ |f (x) − fNε (x)| + |fNε (x) − fNε (a)| + |fNε (a) − f (a)|
1.8. MORE NOTIONS OF CONVERGENCE OF SEQUENCES OF REAL VALUED FUNCTIONS39

The first and third terms are < ε1 by equation 1.8.5 and the middle
term is < ε1 by equation 1.8.6 if |x − a| < δε .
Thus we have for every ε > 0 there exists a δε (a) such that

|x − a| < δε =⇒ |f (x) − f (a)| < 3 ε1 = ε

Hence f (x) is continuous at every point a ∈ I and hence f is continuous

on I

Example 1.8.5 Consider the sequence given in Example 1.8.3

We have
|x2 + x + 1|
|fn (x)| =
n2 + 1
2
|x | + |x| + 1|
≤
n2 + 1
3
≤ 2
since x ∈ [−1, 1]
n +1
Hence the given sequence is bounded. We have seen in Example 1.8.3
that
uc (I)
fn −→ 0
The limit function,being the zero function, is bounded.
Thus we see that uniform convergence has preserved boundedness.
We also see that all the fn are continuous on I and the limit func-
tion,being the zero function, is continuous on I.
Thus we see that uniform convergence has preserved continuity.

Example 1.8.6 Consider the sequence in Example 1.8.4 We had

uc (I)
fn −→ f where f (x) = x

We have
 2
 x + n2 x + x 

|fn (x)| = 
 
n2 + 1 
2 + n2
≤
n2 + 1
40 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

Hence all the fn are bounded. Further we have for the limit function,
|f (x) = |x| ≤ 1
and hence the limit function is also bounded. Thus we see that uniform
convergence has preserved boundedness.
Further all the fn are continuous functions on I and the limit function
f (x) = x is also continuous on I. Thus we see that uniform convergence
has preserved continuity.

3. We shall next investigate whether uniform convergence preserves dif-

ferentiability. We shall first look at an example.

Example 1.8.7 Consider the sequence fn : R −→ R of Example 1.7.8.

We have
x
fn (x) =
1 + nx2
In Example 1.7.8 we had the following facts:
pw (R)
fn −→ 0 (1.8.7)

1 − nx2
fn0 (x) = (1.8.8)
(1 + nx2 )2

pw (R)
fn0 (x) −→ g(x) (1.8.9)
where

1 if x = 0
g(x) = (1.8.10)
0 if x 6= 0
Now we shall see that the sequence {fn }n converges uniformly to the
zero function f (x) = 0. We have from 1.8.8 that the critical points
1
of fn (x) are at x = ± √ and using second derivative we can show
n
1
that the maximum occurs at x = √ and the minimum occurs at
n
1
x = − √ . The maximum value of fn (x) is
n
 
1 1
fn √ = √
n 2 n
1.8. MORE NOTIONS OF CONVERGENCE OF SEQUENCES OF REAL VALUED FUNCTIONS41

and the minimum value of fn (x) is

 
1 1
fn − √ =− √
n 2 n

Thus we have
1
|fn (x)| ≤ √ for all x ∈ R
2 n
1
Now choosing Mn = √ and applying Weierstrass’ Mn test we see
2 n
that
uc (I)
fn −→ f (x) where f (x) = 0 for all x
pw (R)
Now by 1.8.8 and 1.8.9, fn are all differentiable and fn0 −→ g(x).
But f 0 (x) 6= g(x) at x = 0. Thus we see that even though all the fn
are differentiable and the convergence is uniform and the sequence of
derivatives converge pointwise, the derivative of the limit function f (x)
is not the same as the limit of the derivative functions. Thus preserving
differentiability is not, in general true, under uniform convergence.

We shall state the following theorem (without proof) about preserving

differentiation under uniform convergence:

Theorem 1.8.1 Let fn : I −→ R, f : I −→ R and g : I −→ R

 uc (I)

 i)
 fn −→ f 

ii) fn are all differentiable in I
 uc (I) 
iii) fn0 −→ g
 

=⇒
f is differentiable and f 0 (x) = g(x)

4. Finally we shall see whether uniform convergence preserves integrals.

We shall first look at two examples

Example 1.8.8 Consider the sequence of Example 1.8.4. We had

uc (I)
fn −→ f
42 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

where f (x) = x. We have

1
x2 + n2 x + x
Z Z
fn (x)dx = dx
I 0 n2 + 1
1 n2 1
 
1
= + +
n2 + 1 3 2 2
2
 
1 5 n 1
= + +
n2 + 1 6 2 2
5 2
6 n
= +
n2 + 1 2n2 + 2
1
−→ as n → ∞
2

Thus we have
Z
1
lim fn (x)dx = (1.8.11)
n→∞ I 2
We also have
Z Z 1
1
f (x)dx = xdx = (1.8.12)
I 0 2
Thus we have Z Z
lim fn (x)dx = lim fn (x)dx
n→∞ I I n→∞
Thus in this example, uniform convergence has preserved the integrals

Example 1.8.9 Consider the sequence of Example 1.7.9. We have

seen that the sequence converges pointwise to the zero function. Fur-
ther we have, for every n,
1
|fn (x) ≤ for every x ∈ I
n
where I = [0, ∞).
1
Choosing Mn = , it follows by Weierstrass’ test that
n
uc (I)
fn −→ f
1.8. MORE NOTIONS OF CONVERGENCE OF SEQUENCES OF REAL VALUED FUNCTIONS43

where f is the zero function. We have seen in Example 1.7.9 that

Z Z
fn (x)dx = 1 6−→ 0 = f (x)dx
I I

Thus we see that even though we have uniform convergence of fn to f

the integral of fn over the interval I does not converge to that of f

The difference between the two examples above is that the interval I
in Example 1.8.8 is an infinite interval whereas it is finite in Example
1.8.9. We can prove that in general,

Theorem 1.8.2
 

 i) I is a finite interval 


 


 

uc (I)
 ii) fn −→ f 

 


 

 iii) f is integrable over I 
n

=⇒
 
 i) f is integrable
Z overZ I Z 
 ii) lim fn (x)dx = lim fn (x)dx = f (x)dx 
n→∞ I I n→∞ I

Remark 1.8.3 However, when the interval is an infinite interval, uni-

form convergence may preserve integrals in some special cases.

We shall next look at what it means to say that a sequence of real valued
functions {fn }n∈N does not converge uniformly to the function f . In view of
Remark 1.8.2 it is enough to look at what it means to say that a sequence of
real valued functions {fn }n∈N does not converge uniformly to the function
the zero function. Without loss of generality let us assume D is an interval
I on the real line.
A sequence of functions fn : I −→ R converges uniformly to the zero function
if

for every ε > 0 there exists a positive integer Nε such that
n ≥ Nε =⇒ |fn (x)| < ε for every x ∈ I
44 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

This means that if we sketch the graphs of the functions fn (x) then beyond
the stage n ≥ Nε the graphs of all the functions fn (x) will be trapped inside
the band formed by the lines y = ε and y = −ε

fn (x)

a x
b

−ε
I = [a, b]

The graphs of all the functions fn (x) (for all n beyond a certain stage), must
be trapped in the shaded region, and this must be possible for every ε > 0

Remark 1.8.4 The above observation tells us that the sequence fn will fail
to converge uniformly on I, to the zero function, if there exists an ε > 0 for
which we cannot trap the graph of all the functions fn (x) (beyond a certain
stage) in the ε band around the zero function.
This means that the sequence fn will fail to converge uniformly on I to the
zero function if there exists an ε > 0 such that for every positive integer N we
can find a positive integer n > N and a point xn ∈ I such that |fn (xn )| ≥ ε.
This means that the sequence fn will fail to converge uniformly on I to the
zero function if there exists an ε > 0 such that the graph of fn (x) crosses the
+ε or the −ε barrier infinitely often.
A simple situation where this happens is when the graph of all the functions,
beyond a certain stage in the sequence, cross the ε barrier. This means that
the sequence fn will fail to converge uniformly on I to the zero function if
there exists an ε > 0 and a positive integer N such that, for all n ≥ N , the
1.8. MORE NOTIONS OF CONVERGENCE OF SEQUENCES OF REAL VALUED FUNCTIONS45

graph of fn (x) crosses the +ε or the −ε.

A simple situation where this happens is when there exist a positive real
number k > 0 such that every |fn (x)| attains a value ≥ k at least at one
k
point xn ∈ I. (In such a case we could choose the required ε to be = or
2
any positive real number < k).

Example 1.8.10 Consider the sequence of 1.8.2, namely, fn : N −→ R

defined as
x
fn (x) =
n

x
fn (x) =
n

k=1

n x
1 2 3 4

It is easy to see that the (discrete) graph of the function fn (x) consists of
those points, on the straight line passing through the origin and with slope
1
, corresponding to positive integral values of x and therefore crosses the ε
n
barrier
Now we see that if we take k = 1 then for every n we have xn = n such that
1
fn (xn ) = 1 and hence by the above Remark 1.8.4 since all fn cross the ε =
2
barrier, this sequence of real valued functions cannot converge uniformly on
N to the zero function.
46 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

Example 1.8.11 Let I = [0, ∞). Consider the sequence of real valued
functions {fn }n∈N defined on I as follows:

x2 + x + 1
fn (x) =
n2 + 1
It is easy to see that this sequence converges pointwise to the zero function
on I.
We shall show that this sequence does not converge to 0 uniformly. We shall
use the Remark 1.8.4 for this purpose. We show the following:
Let ε = 1. We shall show that for every n ∈ N we can find an xn ∈ I such
that fn (xn ) ≥ 1. We have for any xn ∈ I,

|x2n + xn + 1|
|fn (xn | =
n2 + 1
xn
≥ (since xn is nonnegative in I)
n2 + 1
Hence if we choose xn = n2 + 1, then xn ∈ I and fn (xn ) ≥ 1. Thus for every
n at least at one point xn ∈ I the function fn (x) crosses the value 1. Hence
the sequence does not converge uniformly to zero on I

We shall next briefly mention other types of convergences. In all these types
of convergences the key is the manner in which the error |fn − f | is quantified
and controlled. In pointwise convergence we controlled the error at each point
separately. In uniform convergence we tried to control the error at all points
simultaneously. Next we shall look at notions of converge that arise by trying
to control the average error (the average being defined by various types of
integrals).
We shall first introduce some terminologies:
A real valued function F : I −→ R defined on an interval I is said to be in
Z
1
L (I) if |F (x)|dx < ∞
I
Z
L2 (I) if |F (x)|2 dx < ∞
I
(and in general for 1 ≤ p < ∞) Z
p
L (I) if |F (x)|p dx < ∞
I
1.8. MORE NOTIONS OF CONVERGENCE OF SEQUENCES OF REAL VALUED FUNCTIONS47

Consider a sequence of real valued functions {fn }n∈N defined on an interval

I in R. Let f be a real valued function defined on I. Then we have the
following notions of convergence:
1. L1 convergence
If fn and f are all in L1 (I) and
Z
lim |fn (x) − f (x)|dx = 0
n→∞ I

we say that the sequence fn converges to f in L1 (I). We denote this

convergence as
L1 (I)
fn −→ f
2. L2 convergence
If fn and f are all in L2 (I) and
Z
lim |fn (x) − f (x)|2 dx = 0
n→∞ I

we say that the sequence fn converges to f in L2 (I). We see that we

can also write this as
sZ
lim |fn (x) − f (x)|2 dx = 0
n→∞ I

We denote this convergence as

L2 (I)
fn −→ f
(This is also called RMS (Root Mean Square) Convergence)
3. Lp convergence (1 ≤ p < ∞)
If fn and f are all in Lp (I) and
Z
lim |fn (x) − f (x)|p dx = 0
n→∞ I

we say that the sequence fn converges to f in Lp (I). We see that we

can also write this as
Z  p1
p
lim |fn (x) − f (x)| dx =0
n→∞ I
48 CHAPTER 1. SOME BASIC IDEAS IN ANALYSIS

We denote this convergence as

Lp (I)
fn −→ f

(This is also called p-mean Convergence)