Riemann Integration

Paper: Analysis-III

Lesson: Riemann Integration

Author Name: Mrs. Rajni Gupta and Deepti Kaur

College/ Department: Department of Mathematics,

Maitreyi College,
University of Delhi and Department of Mathematics,
University of Delhi

Institute of Lifelong Learning, University of Delhi

Table of Contents

RIEMANN INTEGRATION

1. Learning Outcomes
2. Introduction
3. Darboux Integral
4. Refinement of Partitions and Integrability Theorems
5. Riemann Sums
6. Properties of Integral
7. Classes of Integrable Functions
8. Solved Problems
9. Summary
10. Exercises
11. References

2
1. Learning Outcomes
This chapter contains a careful development of the Riemann integral of real valued functions
defined on a closed bounded interval of R . We believe that at the end of this chapter, the
reader shall be able to understand the basic properties of integrable functions. We will
establish the Riemann integrability of several important classes of functions and we shall
see that not all functions are Riemann integrable. The reader will also be able to interpret
the integral of a function on a closed bounded interval as the signed area of the
corresponding region. The fact that the integral is equal to the geometric area of the region
between graph of the function and the x -axis can be used to compute definite integral of
simple functions.

2. Introduction
The main idea of defining the Riemann integral is to find the area of planar region under a
curve.

Motivation: Find the area A of the region enclosed by a circle of radius r . For an arbitrary
n , consider n equal inscribed and superscribed triangles. We divide (partition) the circular
disc into ‘ n ’ equal inscribed and circumscribed triangles as shown in the following Figures
1(a) and 1(b), respectively.

Figure 1(a) Figure 1(b)

It is expected that

Sum of areas of all the inscribed triangles  A  Sum of areas of all the circumscribed
triangles.

3
Let us first focus on the inscribed triangles. Consider the OAB as shown in Figure 2 below.

Figure 2

Here,

OP  r cos( / n)

AP  r sin( / n)

1 
Area of OAB  2    OP  AP   r 2 cos( / n)sin( / n) .
2 
So, sum of the areas of all the inscribed triangles is n r 2 cos( / n)sin( / n) .
Next, we calculate the areas of the circumscribed triangles. Consider the OAB as shown
in Figure 3 below.

4
 Figure 3

OP  r
AP  r tan( / n)

1 
Area of OAB  2    OP  AP   r 2 tan( / n) .
2 
So, sum of the areas of all the circumscribed triangles is n r 2 tan( / n) .
But A lies between these two areas, so that we have

  
n r 2 cos sin  A  n r 2 tan . (1)
n n n
It is observed that as n increases, we are able to cover the area of the circle more
precisely. So, for large n , we would be able to minimize the error between the areas of the
inscribed (circumscribed) triangles and the area of the circle.
Consider

   sin( / n) 
lim n r 2 cos sin  lim n r 2 cos   r2
n  n n n  n  /n n
and

 tan( / n) 
lim n r 2 tan  lim n r 2   r2
n  n n   /n n
Hence, if n is very very large, i.e., when the partition becomes smaller and smaller, both
 r2 and hence we obtain A   r .
2
left and right hand sides of the inequality (1) approaches

This idea will be used to define the partition of an interval and to find the area of the region
under the graph of a function.

Now, suppose f is a nonnegative bounded function defined on the interval [a, b] . We first

subdivide the interval into a finite number of subintervals. Then, the area of the region
under the graph of f lies between the total area of the inscribed rectangles (shown by blue

color in Figure 4(a) below) and total area of superscribed rectangles (shown by yellow color
in Figure 4(b) below).

5
 Figure 4(a)

Figure 4(b)

If the total area of the inscribed and superscribed rectangles converge to the same limit as
we make the partition of [a, b] finer and finer, then the area of the region under the graph

of f can be defined as this limit and f is said to be integrable.

Prerequisites: We expect that the reader has studied real analysis and is familiar with the
basic concepts such as boundedness, supremum, infimum, sequences, limit, continuity,
uniform continuity, differentiability etc.

Throughout this chapter, f is a bounded function defined on a closed bounded interval

[a, b] ( a, b  R and a  b ).

3. Darboux Integral
Definition 3.1 A partition of [a, b] is a finite ordered set P  {t0 , t1 ,, tn } of points in [a, b]
such that

a  t0  t1    tn1  tn  b .

6
In this chapter, we will denote the partition P by the notation P  {a  t0  t1    tn  b} .
Next, we see some examples of partition.

The simplest partition of [a, b] is P  {a, b} .

 1 
The partition P  0, ,1 of [0,1] subdivides the interval [0,1] in two subintervals each of
 2 
length 1/ 2 .

 1 1 3 
The partition P  0, , , ,1 of [0,1] subdivides the interval [0,1] into four subintervals
 4 2 4 
each of length 1/ 4 .

In general, the partition P  {a  t0  t1    tn  b} of [a, b] , where

k
tk  a  b  a  , k  0,1,..., n
n

subdivides the interval [a, b] into ‘ n ’ subintervals each of length (b  a) / n . It is clear that

as n becomes large, the subdivision of [a, b] corresponding to the above partition P

becomes finer.

REMARK
It is not necessary to have partition of [a, b] into subintervals of equal length. For

 1 1 1 1 
instance, the partition P  0, , , , ,1 subdivides the interval [0,1] into subintervals
 5 4 3 2 
of unequal lengths.

Definition 3.2 The mesh of a partition P is the maximum length of the subintervals

forming P . For P  {a  t0  t1    tn  b} , mesh( P) = max{tk  tk 1 : k  1, 2,..., n} .

For S  [a, b] , denote

m( f , S )  inf{ f ( x) : x  S} ,

M ( f , S )  sup{ f ( x) : x  S} .

7
These infimum and supremum are finite real numbers since f is bounded on [a, b] .

Since, for the partition P  {a  t0  t1    tn  b} , the subintervals [tk 1 , tk ]  [a, b],

k  1, 2,..., n , so by the infimum and supremum property, we have the following remark.

Remark 3.1 Given a partition P  {a  t0  t1    tn  b} of [a, b] , it is clear that

m( f ,[a, b])  m( f ,[tk 1 , tk ])  M ( f ,[tk 1, tk ])  M ( f ,[a, b]) for all k  1,..., n .

Definition 3.3 The lower Darboux sum L( f , P) of f with respect to the partition

P  {a  t0  t1    tn  b} is

given by

n
L( f , P)   m( f ,[tk 1 , tk ])(tk  tk 1 ) ,
k 1

Likewise, the upper Darboux sum U ( f , P) of f with respect to the same partition P is

defined as

n
U ( f , P)   M ( f ,[tk 1 , tk ])(tk  tk 1 )
k 1

Geometrically, the lower Darboux sum is the sum of the areas of the inscribed rectangles as
shown in Figure 5(a) and the upper Darboux sum is the sum of the areas of the
circumscribed rectangles as illustrated in Figure 5(b).

8
 Figure 5(a) Figure 5(b)

From Figures 5(a) and 5(b), intuitively it is clear that

L( f , P)  U ( f , P) .

It is natural to expect the following result.

Proposition 3.1 Let f :[a, b]  R be a bounded function. Then, for any partition

P  {a  t0  t1    tn  b} of [a, b]

m( f ,[a, b])(b  a)  L( f , P)  U ( f , P)  M ( f ,[a, b])(b  a) (2)

Proof: For each subinterval [tk 1 , tk ], k  1,..., n of [a, b] , tk  tk 1  0 , therefore from Remark
3.1, for all k  1,..., n , we have

m( f ,[a, b])(tk  tk 1 )  m( f ,[tk 1, tk ])(tk  tk 1 )  M ( f ,[tk 1, tk ])(tk  tk 1 )  M ( f ,[a, b])(tk  tk 1) .

This implies

n n n n

 m( f ,[a, b])(t
k 1
k  tk 1 )   m( f ,[tk 1 , tk ])(tk  tk 1 )   M ( f ,[tk 1 , tk ])(tk  tk 1 )   M ( f ,[a, b])(tk  tk 1 )
k 1 k 1 k 1

which gives

n n
m( f ,[a, b]) (tk  tk 1 )  L( f , P)  U ( f , P)  M ( f ,[a, b]) (tk  tk 1 )
k 1 k 1

and hence

m( f ,[a, b])(b  a)  L( f , P)  U ( f , P)  M ( f ,[a, b])(b  a) .

Geometrically, for a bounded function f on [a, b] , the inequality (2) can be viewed as

9
 Figure 6(a) Figure 6(b)

Figure 6(c) Figure 6(d)

It is clear from Figures 6(a), 6(b), 6(c) and 6(d) that

m( f ,[a, b])(b  a)  L( f , P)  U ( f , P)  M ( f ,[a, b])(b  a) .

10
NOTE
1. The inequality (2) is true for any partition P of [a, b] .

2. It follows from Proposition 3.1 that the sets L( f , P) : P is a partition of [a, b] and
U ( f , P) : P is a partition of [a, b] are bounded and since these are bounded sets,

we can meaningfully talk about their infimum and supremum.

Motivation: The natural question comes to mind that what is the significance of taking
the infimum and supremum of the sets L( f , P) : P is a partition of [a, b] and

U ( f , P) : P is a partition of [a, b] and more specifically, for which set out of these two, we

should take the infimum and for which set, supremum should be taken?

In the forthcoming discussion, we elaborate these ideas which would lay the foundation for
defining the integral of a bounded function on [a, b] .

Consider the following Figure 7, in which we have taken lower Darboux sum corresponding

to a partition P  {a  t0  t1    tn  b} of [a, b] .

Figure 7

Let us insert a point z between t0 and t1 . The below Figure 8 illustrates what happens to
the first rectangle after inserting z .

11
 Figure 8
We observe that the area of the first rectangle increases after inserting the point z . So, if
we increase the number of points in a partition P of [a, b] to get a new partition Q of [a, b]

, the lower sum corresponding to the partition Q is expected to be more than or equal to

the lower sum corresponding to the partition P and further note that L( f , Q) which is the

sum of the areas of inscribed rectangles associated with partition Q is more closer to the

area under the curve of function f than L( f , P) .

Therefore, we look for partitions P of [a, b] with respect to which the lower sums are as

large as possible in order to get the best approximation of the area under the curve of the
function f and hence, we must take supremum of all the lower sums corresponding to all

partitions of [a, b] . So, we have the following definition.

Definition 3.4 The lower Darboux integral of f is defined to be

L( f )  sup L( f , P) : P is a partition of [a, b] .

Now, let us consider the following Figure 9, in which we have taken the upper sum

corresponding to a partition P  {a  t0  t1    tn  b} of [a, b] .

12
 Figure 9

Let us again insert a point z between t0 and t1 . The Figure 10 below illustrates what
happens to the first rectangle after inserting z .

Figure 10

We observe that the area of the first rectangle decreases after inserting the point z . So, if
we increase the number of points in a partition P of [a, b] to get a new partition Q of [a, b]

, the upper sum corresponding to the partition Q is expected to be less than or equal to the

upper sum corresponding to the partition P and further note that U ( f , Q) which is the sum

of the areas of circumscribed rectangles associated with partition Q is more closer to the

area under the curve of function f than U ( f , P) .

Therefore, we look for partitions P of [a, b] with respect to which the upper sums are as

small as possible in order to get the best approximation of the area under the curve of

13
function f and hence, we must take infimum of all the upper sums corresponding to
different partitions of [a, b] . So, we have the next definition.

Definition 3.5 The upper Darboux integral of f is defined as

U ( f )  inf{U ( f , P) : P is a partition of [a, b]}

It is clear from (2) that L( f ) and U ( f ) are real numbers.

It is verified in the next section that L( f )  U ( f ) .

The above explanation shows that if f :[a, b]  R is a nonnegative bounded function and

A is the area of the region below the graph of f and between the lines x  a , x  b and
above the x -axis, then

L( f )  A  U ( f ) .

Hence, if L( f )  U ( f ) , then this common value is suitable to define the area A . The next

definition is motivated by this reasoning.

Definition 3.6 A bounded function f defined on [a, b] is Darboux integrable (integrable) if

b b
L( f )  U ( f ) . In this case, we write 
a
f or  f ( x)dx
a
for this common value:

b b


a
f   f ( x)dx  L( f )  U ( f )
a

Example 3.1 Show that the constant function f ( x)  c is integrable over any closed
b
interval [a, b] . What is f?
a

For any partition P  {a  t0  t1    tn  b} of [a, b] , we have

m( f ,[tk 1 , tk ])  c  M ( f ,[tk 1, tk ]) , for all k  1, 2,..., n . Therefore,

n n n
L( f , P)   m( f ,[tk 1 , tk ])(tk  tk 1 )   c(tk  tk 1 )  c (tk  tk 1 )  c(b  a)
k 1 k 1 k 1

14
and

n n n
U ( f , P)   M ( f ,[tk 1 , tk ])(tk  tk 1 )   c(tk  tk 1 )  c  (tk  tk 1 )  c(b  a) .
k 1 k 1 k 1

Hence, L( f )  c(b  a)  U ( f ) , which implies that the constant function f ( x)  c is

b
integrable and f
a
 c(b  a) .

Example 3.2 Show that the Dirichlet function f defined by

0, when x is irrational

f ( x)  
1, when x is rational

is not integrable on any interval [0, b] , where b  0.

Let P  {0  t0  t1    tn  b} be a partition of [0, b] . Since each subinterval

[tk 1 , tk ], k  1, 2,..., n contains rationals and irrationals, therefore, for all k  1, 2,..., n,

m( f ,[tk 1 , tk ])  0 and M( f ,[tk 1, tk ])  1 , so that

n n
U ( f , P)   M ( f ,[tk 1 , tk ])(tk  tk 1 )  1 (tk  tk 1 )  b
k 1 k 1

and

n n
L( f , P)   m( f ,[tk 1 , tk ])(tk  tk 1 )   0 (tk  tk 1 )  0
k 1 k 1

Therefore, it follows that U ( f )  b and L( f )  0 , so f is not integrable on [0, b] .

NOTE
Though we have defined Riemann integral for bounded functions but the above example
shows that if a function is bounded then it is not necessary that is integrable.

Example 3.2 Show that the function f defined by

15
 0, if 0 < x  1
f ( x)  
1, if x  0

1
is integrable on [0,1] and f
0
 0.

Let P  {0  t0  t1    tn  1} be a partition of [0,1] . Since f ( x)  0 for x  0 , we have

m( f ,[tk 1 , tk ])  0  M ( f ,[tk 1, tk ]) , for k  2,3,..., n . The first subinterval in the partition is

[0  t0 , t1 ] , where 0  t1  1 . For this subinterval, m( f ,[t0 , t1 ])  0 and M ( f ,[t0 , t1 ])  1 , as

f (0)  1 and f ( x)  0 , for 0  x  t1 . Hence, it follows that

n
L( f , P)   m( f ,[tk 1 , tk ])  tk  tk 1   0
k 1

and

n
U ( f , P)   M ( f ,[tk 1 , tk ])  tk  tk 1   M ( f ,[t0 , t1 ])  t1  t0   t1 .
k 1

Thus,

L( f )  sup L( f , P) : P is a partition of [0,1]  0

and

U ( f )  inf U ( f , P) : P is a partition of [0,1]

 inf t1 : 0  t1  1  0 .

1
Hence, L( f )  U ( f ) , proving that f is integrable on [0,1] and f
0
 L( f )  U ( f )  0 .

NOTE
The above example shows that even a discontinuous function is integrable.

16
4. Refinement of Partitions and Integrability Theorems

Definition 4.1 A partition Q is said to be a refinement of a partition P if P  Q , i.e. if

every point of P is a point of Q . If P1 and P2 are two partitions, then P1  P2 is called the
common refinement of P1 and P2 .

Lemma 4.1 If P and Q are two partitions of [a, b] and P  Q , then

L( f , Q)  L( f , P) and U ( f , Q)  U ( f , P) , i.e. under refinement, lower sums grow and

upper sum shrinks.

Proof: To prove L( f , Q)  L( f , P) , we first assume that the refinement Q has only a single

point, say z , more than P . We have seen geometrically in Figure 8 that the result holds in

this case. If P  {a  t0  t1    tn  b} , then Q  {a  t0  t1    ti 1  z  ti  tn  b} , for

some i {1, 2,, n}. Now

i 1 n
L( f , P)   m( f ,[tk 1 , tk ])  tk  tk 1   m( f ,[ti 1 , ti ])  ti  ti 1    m( f ,[t k 1 , tk ])  tk  tk 1 
k 1 k i 1

and

i 1
L( f , Q)   m( f ,[tk 1 , tk ])  tk  tk 1   m( f ,[ti 1 , z ])  z  ti 1   m( f ,[ z , ti ]) ti  z 
k 1

n
  m( f ,[tk 1 , tk ])  tk  tk 1  .
k i 1

The lower Darboux sums for P and Q are same except for the terms which involves ti 1 or

ti . We have

L( f , Q)  L( f , P)  m( f ,[ti 1, z ])( z  ti 1 )  m( f ,[ z, ti ])(ti  z)  m( f ,[ti 1, ti ])(ti  ti 1) (3)

and

m( f ,[ti 1 , ti ])(ti  ti 1 )  m( f ,[ti 1, ti ])(ti  z  z  ti 1 )

17
  m( f ,[ti 1 , ti ])(ti  z)  m( f ,[ti 1 , ti ])( z  ti 1 ) . (4)

Since [ti 1 , z ], [ z, ti ]  [ti 1 , ti ] , so by infimum property, m( f ,[ti 1 , ti ])  m( f ,[ti 1 , z]) and

m( f ,[ti 1 , ti ])  m( f ,[ z, ti ]) . Hence, from (4), it follows that

m( f ,[ti 1 , ti ])(ti  ti 1 )  m( f ,[ z, ti ])(ti  z)  m( f ,[ti 1, z])( z  ti 1 ) (5)

Thus, from (3) and (5), it follows that L( f , P)  L( f , Q) , when Q contains just one point

more than P . Using induction, we have L( f , P)  L( f , Q) if Q contains r points more than

P and a similar argument holds for the upper Darboux sums.

Lemma 4.2 If f ( x)  M , x [a, b] and Q is a refinement of partition P having r points

more than P , then

L( f , P)  L( f , Q)  L( f , P)  2rM mesh(P)

U ( f , P)  U ( f , Q)  U ( f , P)  2rM mesh(P)

Proof: Firstly, assume that the refinement Q has only a single point, say z , more than P .

Proceeding as in the previous lemma, we see that

L( f , Q)  L( f , P)  m( f ,[ti 1, z])( z  ti 1 )  m( f ,[ z, ti ])(ti  z)  m( f ,[ti 1, ti ])(ti  z  z  ti 1)

  m( f ,[ti 1 , z ])  m( f ,[ti 1 , ti ])  ( z  ti 1 )   m( f ,[ z, ti ])  m( f ,[ti 1, ti ])  (ti  z )

Also, as M  f ( x)  M , x [tk 1 , tk ] , we have

M  m( f ,[ti 1 , ti ])  m( f ,[ti 1, z])  M , (since [ti 1 , z ]  [ti 1 , ti ] )

which implies

0  m( f ,[ti 1 , z])  m( f ,[ti 1, ti ])  2M .

In a similar manner, we obtain 0  m( f ,[ z, ti ])  m( f ,[ti 1, ti ])  2M . Therefore,

0  L( f , Q)  L( f , P)  2M (ti  ti 1 )  2M mesh(P) , when Q contains just one point more

than P. Now let us add a point z * to the partition Q to obtain a partition Q* , i.e., Q* has

18
exactly one point more than Q. Then, by the above argument, we get,

L(Q* , f )  L(Q, f )  2M mesh(P) and hence we obtain L(Q* , f )  L( P, f )  4M mesh(P) . By

induction argument, we obtain L( f , Q)  L( f , P)  2rM mesh(P) , when Q contains r points

more than P. Similarly, the other result for the upper Darboux sums may be proved.

Lemma 4.3 If P and Q are any two partitions of [a, b] , then L( f , P)  U ( f , Q) .

Proof: The set P  Q , the common refinement of P and Q is a partition of [a, b] . Since

P  P  Q and Q  P  Q , from Lemma 4.1, we obtain

L( f , P)  L( f , P  Q)  U ( f , P  Q)  U ( f , Q) .

Theorem 4.1 For any bounded function f on [a, b] , L( f )  U ( f ) .

Proof: Fix a partition P of [a, b] . From Lemma 4.3, it follows that L( f , P)  U ( f , Q), for all

partitions Q of [a, b] . It shows that L( f , P ) is a lower bound of the set

U ( f , Q) : Q is a partition of [a,b] . As U ( f )  inf{U ( f , Q) : Q is a partition of [a, b] }, so

L( f , P)  U ( f ) . This shows that U( f ) is an upper bound of the set

L( f , P) : P is a partition of [a,b] . As L( f )  sup{L( f , P) : P is a partition of [a, b] }, so

L( f )  U ( f ) .

Example 4.1 Show that the identity function defined by f ( x)  x for all x [a, b] is
b
integrable on [a, b] and find f?
a

Let P  {a  t0  t1    tn  b} be a partition of [a, b] . Since m( f ,[tk 1 , tk ])  tk 1 and

M ( f ,[tk 1 , tk ])  tk , for k  1,..., n , so we have

n n
U ( f , P)   tk  tk  tk 1  and L( f , P)   tk 1  tk  tk 1  , which gives the relations
k 1 k 1

n
U ( f , P)  L( f , P)    tk  tk 1 
2
(6)
k 1

and

19
 n
U ( f , P)  L( f , P)    tk 2  tk 12  . (7)
k 1

Hence, from (6) and (7), we obtain

b2  a 2 1 n
   tk  tk 1 
2
U ( f , P) 
2 2 k 1

and

b2  a 2 1 n
   tk  tk 1  .
2
L( f , P ) 
2 2 k 1

Consider  0 and let Pn  {a  t0  t1    tn  b} be a partition of [a, b] into n equal

parts, such that

k
tk  a  b  a  , k  0,1,..., n ,
n

where the natural number n is so chosen that (b  a)2 / n   .

Then,

2
 (k  1) (b  a) 2 (b  a) 2
 b  a    2 n 
n n
k
 t  tk 1     a  b  a   a 
2
k  .
k 1 k 1  n n  n n

Therefore, we obtain

U ( f )  inf U ( f , P) : P is a partition of [a, b]

b2  a 2 
 U ( f , Pn )  
2 2

and

L( f )  sup L( f , P) : P is a partition of [a, b]

b2  a 2 
 L( f , Pn )   .
2 2

20
 (b 2  a 2 ) (b 2  a 2 )
Since  0 is arbitrary, it follows that U( f )  and L( f )  .
2 2

(b2  a 2 )
But from Theorem 4.1, L( f )  U ( f ) , which shows that L( f )   U ( f ) . Hence, f
2
b
(b 2  a 2 )
is integrable on [a, b] and 
a
f 
2
.

REMARK
The above example shows even for a simple function, it is not straightforward to
determine whether a bounded function is integrable on [a, b] using Definition 3.6 and if

the function is integrable, it may be even more complex to compute its integral. We have
a simple condition for this reason, which will not only enable us to determine large classes
of integrable functions but also help in obtaining numerous significant properties of the
integral.

In the next two theorems, we establish the important Cauchy criterion of integrability.

Theorem 4.2 A bounded function f on [a, b] is integrable if and only if for each   0,
there exists a partition P of [a, b] such that

U ( f , P)  L( f , P)   (8)

Proof: We first assume that f is integrable on [a, b] . As

L( f )  sup L( f , P) : P is a partition of [a, b]

and

U ( f )  inf U ( f , P) : P is a partition of [a, b] ,

therefore, given any   0 , there exist partitions Q and R of [a, b] such that

 
L( f , Q )  L ( f )  and U( f , R)  U ( f ) 
2 2

21
Let P be the common refinement of Q and R. Then, by Lemma 4.1,

L( f , Q)  L( f , P) and U ( f , P)  U ( f , R) . Therefore,

U ( f , P)  L( f , P)  U ( f , R)  L( f , Q)

 
U( f )    L( f )  U ( f )  L( f )  
2 2

Since L( f )  U ( f ) , so condition (8) holds.

Conversely, assume that for each   0, condition (8) holds for some partition P . From

Theorem 4.1, L( f )  U ( f ) . Since U ( f )  inf U ( f , P) : P is a partition of [a, b]  U ( f , P)

and L( f , P)  L( f )  sup{L( f , P) : P is a partition of [a, b]} , so we have

0  U ( f )  L( f )  U ( f , P)  L( f , P)   .

The above inequality is true for each   0 , so we must have L( f )  U ( f ) and hence f is

integrable on [a, b] .

Theorem 4.3 A bounded function f on [a, b] is integrable if and only if for each   0,
there exists a  0 such that U ( f , P)  L( f , P)   , for all partitions P of [a, b] having

mesh(P) <  .

Proof: If for each   0, there exists a  0 such that U ( f , P)  L( f , P)   , for all

partitions P of [a, b] having mesh(P) <  , then Theorem 4.2 implies that f is integrable

on [a, b] .

Conversely, assume that f is integrable on [a, b] and consider   0. Then, by Theorem

4.2 there exists a partition P1  {a  z0  z1    zm  b} of [a, b] such that


U ( f , P1 )  L( f , P1 )  . (9)
2

Since f is bounded on [a, b] , so there exists M  0 such that f ( x)  M , x [a, b] .


Choose   , where m is the number of subintervals in partition P1 .
8mM

22
Let P  {a  t0  t1    tn  b} be a partition of [a, b] having mesh( P) <  .

Let Q  P  P1 , then Q has at most m points which are not in P . It follows from Lemma
4.2 that


L( f , Q)  L( f , P)  2mM mesh(P)  2mM   .
4

Since P1  Q , so we have L( f , P1 )  L( f , Q) . Hence


L( f , P1 )  L( f , P)  L( f , Q)  L( f , P)  . (10)
4

Similarly,


U ( f , P)  U ( f , P1 )  . (11)
4

Therefore, from (10) and (11), it follows that

 
U ( f , P)  L( f , P)  U ( f , P1 )    L( f , P1 )
4 4


 U ( f , P1 )  L( f , P1 ) 
2

 
   . (Using (9))
2 2

The next theorem provides a way to prove the existence of an integral and evaluate it.

Theorem 4.4 Let f be a bounded function on [a, b] . Suppose there exists a sequence ( Pn )
of partitions of [a, b] such that

limU ( f , Pn )  L( f , Pn )  0 . (12)
n

b
Then f is integrable on [a, b] and  f ( x)dx  limU ( f , P )  lim L( f , P ) .
a
n 
n
n 
n

23
Proof: By condition (12), for given   0 , there is a natural number n such that

U ( f , Pn )  L( f , Pn )   , so Theorem 4.2 implies that f is integrable on [a, b] and

b

 f ( x)dx  L( f )  U ( f ) . Further, since U ( f )  U ( f , P )

a
n and L( f , Pn )  L( f ) , we have

0  U ( f , Pn )  U ( f )  U ( f , Pn )  L( f )  U ( f , Pn )  L( f , Pn ) (13)

Since the limit of the right hand side in (13) is zero as n   , the Squeeze Theorem for
limits implies

b
limU ( f , Pn )  U ( f )   f ( x)dx .
n 
a

It follows that

lim L( f , Pn )  limU ( f , Pn )  lim U ( f , Pn )  L( f , Pn ) 

n n  n 

b b
  f ( x)dx  0   f ( x)dx .
a a

Example 4.2 Show that f ( x)  x 2 is integrable on the interval [0, b] , where b  0 .

kb
We consider the partition Pn  {0  t0  t1    tn  b} by choosing tk  , k  0,1, n,
n
obtained by dividing the interval [0, b] into n equal parts. Since f is increasing on [0, b] ,
m( f ,[tk 1 , tk ])  tk21 and M ( f ,[tk 1 , tk ])  tk2 , for each k  1, 2,..., n . We have

n
U ( f , Pn )   M ( f ,[tk 1 , tk ])(tk  tk 1 )
k 1

n
  tk2 (tk  tk 1 )
k 1

n
k 2b 2 b b 3 n 2
 2
  3 k
k 1 n n n k 1

24
 b3 n(n  1)(2n  1)
 
n3 6

b3  1  1
 1   2  
6  n  n

b3
So, lim U ( f , Pn )  . For the partition Pn , we have
n  3

n
L( f , Pn )   m( f ,[tk 1 , tk ])(tk  tk 1 )
k 1

n
  tk21 (tk  tk 1 )
k 1

n
(k  1)2 b2 b b3 n
 2
  3  (k  1)2
k 1 n n n k 1

b3 (n  1)n(2n  1)
  ,
n3 6

b3  1  1
 1   2  
6  n  n

b3
so, it follows that lim L( f , Pn )  . Hence by Theorem 4.4, f ( x)  x is integrable on [0, b]
2
n  3
b
b3
and  x dx  
2

a
3

5. Riemann Sums

Definition 5.1 Let P  {a  t0  t1    tn  b} be a partition of [a, b] . Then Riemann sum

of f corresponding to a partition P denoted by S ( f , P) is defined as

n
S ( f , P)   f ( xk )(tk  tk 1 ),
k 1

25
where xk [tk 1 , tk ] , k  1, 2,..., n . As xk can be any point of the interval [tk 1 , tk ], k  1, 2,..., n ,
so Riemann sum S ( f , P) of f corresponding to P is not unique.

Figure 11: A Riemann Sum

It is evident from Figure 11 that the accuracy of the Riemann sum approximation appears to
improve if the rectangles gets thinner. In this sense, we take the limit of these
approximating Riemann sums as the width of the subintervals of the partitions approaches
zero. This limit, if it exists is the Riemann’s definition of integrability.

Definition 5.2 A function f is said to be Riemann integrable on [a, b] if to each   0,

there exists a  0 and a number I such that

S ( f , P)  I   ,

for every Riemann sum S ( f , P) of f corresponding to every partition P having

mesh( P)   . The number I is said to be the Riemann integral of f on [a, b] .

Example 5.1 Show that the function f ( x)  x 2 is integrable on the interval [0, b] , where
b  0 by using Riemann sum.

26
Let the partition Pn  {0  t0  t1    tn  b} divide [0, b] into n equal subintervals each of
kb
length 1/ n so that mesh( P)  0 as n  , where tk  , k  1, 2,..., n . Let
n
xk  tk [tk 1 , tk ], k  1, 2,..., n . Then, the Riemann sum for this choice of xk is

n
S ( f , Pn )   f ( xk )(tk  tk 1 )
k 1

n
  f (tk )(tk  tk 1 )
k 1

2
 kb 
n
    (tk  tk 1 )
k 1  n 

b3 n 2 b3 n(n  1)(2n  1)
 3 k  3 
n k 1 n 6

b3  1  1
 1   2  
6  n  n

b3
Therefore, lim S ( f , Pn )  lim S ( f , Pn )  and as the limit exists, the function is
mesh(P)0 n  3
b
b3
integrable and a
f 
3
.

Now, we have two definitions of integrability. The equivalence of these two definitions is
established in the next theorem.

Theorem 5.1 A bounded function f on [a, b] is Riemann integrable if and only if it is

Darboux integrable and the values of the integrals are same.

Proof: We first assume that f is Darboux integrable in the sense of Definition 3.6.
Consider   0. By Theorem 4.3, there exists a  0 such that for every partition P of
[a, b] having mesh(P) <  , U ( f , P)  L( f , P)   . Clearly,

L( f , P)  S ( f , P)  U ( f , P) (14)

27
 n
for a Riemann sum S ( f , P)   f ( x )(t
k 1
k k  tk 1 ) , xk [tk , tk 1 ] , k  1, 2,..., n of f associated

with a partition P  {a  t0  t1    tn  b} having mesh(P) <  . Since

b
U ( f , P)  L( f , P)    L( f )     f ( x)dx   (15)
a

and

b
L( f , P)  U ( f , P)    U ( f )     f ( x)dx   (16)
a

From (14), (15) and (16), it follows that

b
S ( f , P)   f ( x)dx   ,
a

b
which implies that f is Riemann integral on [a, b] and the value of the integral is  f ( x)dx .
a

Conversely, assume that f is Riemann integrable in the sense of Definition 5.2 and let 
be any positive number. Choose a partition P  {a  t0  t1    tn  b} of [a, b] with

mesh(P) <  and select xk [tk , tk 1 ] such that

f ( xk )  m[ f ,[tk 1 , tk ]   , for each k  1, 2,..., n .

For this choice of xk , the Riemann sum S ( f , P) satisfy

n
S ( f , P)   f ( xk )(tk  tk 1 )
k 1

n
   m[ f ,[tk 1 , tk ]    tk  tk 1 
k 1

n n
  m[ f ,[tk 1 , tk ]  tk  tk 1      tk  tk 1 
k 1 k 1

 L( f , P)   (b  a)

28
Since S ( f , P)  I   , so the following inequalities hold

L( f )  L( f , P)  S ( f , P)   (b  a)  I     (b  a) .

As  0 is arbitrary, we get L( f )  I . Analogously, it can be shown that U ( f )  I . But

L( f )  U ( f ) , so it follows that L( f )  U ( f )  I and hence f is Darboux integrable and

b

 f ( x)dx  I .
a

In view of the equivalence of the two definitions of integral, we say f is integrable on [a, b]

if and only if it is Darboux integrable on [a, b] according to Definition 3.6 or it is Riemann

integrable on [a, b] according to Definition 5.2.

6. Properties of Integral

Theorem 6.1 If f and g are integrable functions on [a, b] and c be any real number,
then

b b
(i) Linearity: c f is integrable on [a, b] and cf
a
 c f .
a

b b b
(ii) Additivity: f  g is integrable on [a, b] and  ( f  g )   f   g.
a a a

b b
(iii) Monotonicity: f ( x)  g ( x), x [a, b] implies a
f   g.
a

Proof: (i) The result is trivially true for c  0. Consider a partition

P  {a  t0  t1    tn  b} of [a, b].

First, let c  0 . Then,

M (cf ,[tk 1 , tk ])  sup (cf )( x) : x [tk 1, tk ]

 sup c f ( x) : x [tk 1 , tk ]

 c sup  f ( x) : x [tk 1 , tk ]

29
  cM ( f ,[tk 1 , tk ]) , k  1, 2,..., n ,

which implies U (c f , P)  cU ( f , P) . Consequently, U (c f )  cU ( f ) and similarly,

L(c f )  cL( f ) . Since f is integrable on [a, b] , it follows that

L(c f )  cL( f )  cU ( f )  U (c f ) ,

b b
hence, c f is integable on [a, b] and c f
a
U (c f )  cU ( f )  c  f .
a

Next, we suppose that c  1. Then,

M ( f ,[tk 1 , tk ])  sup  f ( x) : x [tk 1, tk ]

  inf  f ( x) : x [tk 1, tk ]

 m( f ,[tk 1 , tk ]) , k  1, 2,..., n ,

which leads to U ( f , P)   L( f , P) and thus U ( f )   L( f ) . Replacing f with  f ,

we get U ( f )   L( f ) . Because f is integrable on [a, b] , we have

U ( f )   L( f )  U ( f )  L( f ) ,

b b
which shows  f is integrable on [a, b] and   f U ( f )  U ( f )   f .
a a

If c  0 , then c  0 , so from above we get c f is integrable on [a, b] and

consequently, (c f )  c f is integrable on [a, b] . Also, in this case

b b b b b b

 c f   (c) f   (c f )   (c f )  (c) f  c  f .

a a a a a a

(ii) Consider   0 . Since f and g are integrable on [a, b] , so there exist partitions Q

and R of [a, b] such that


U ( f , Q)  L( f , Q)  ,
2

30
 
U ( g , R )  L( g , R ) 
2

Let P  Q  R be a common refinement of Q and R . Then


U ( f , P )  L( f , P )  U ( f , Q )  L ( f , Q )  (17)
2
Similarly,


U ( g , P )  L( g , P )  (18)
2
For any subset S of [a, b] , m( f  g , S )  m( f , S )  m( g , S ). This implies

L( f  g , P)  L( f , P)  L( g , P) (19)

Similarly,

U ( f  g , P)  U ( f , P)  U ( g , P) (20)

Using (17), (18), (19) and (20), we get

 
U ( f  g , P )  L( f  g , P )  U ( f , P )  U ( g , P )  L ( f , P )  L ( g , P )    .
2 2

Hence, f  g is integrable on [a, b] .

Since

 ( f  g )  U ( f  g )  U ( f  g , P)
a

 U ( f , P)  U ( g , P) (From (20))

 
 L( f , P )   L( g , P )  (From (17) and (18))
2 2

 L( f , P)  L( g , P)  

 L( f )  L( g )  

31
 b b
  f   g  .
a a

As  0 is arbitrary, it follows that

b b b

 ( f  g)   f   g .
a a a

Also,

 ( f  g )  L( f  g )  L( f  g , P )
a

 L( f , P)  L( g , P) (From (19))

 
 U ( f , P)   U ( g , P)  (From (17) and (18))
2 2

 U ( f , P)  U ( g , P)  

 U ( f ) U (g)  

b b
  f   g  .
a a

As  0 is arbitrary, it follows that

b b b

 ( f  g)   f   g .
a a a

b b b
Therefore,  ( f  g )   f   g.
a a a

(iii) Let h  g  f  0 . Since f and g are integrable on [a, b] , so h is integrable on

n
[a, b] and L(h, P)   m(h,[tk 1 , tk ])(tk  tk 1 )  0 for all partitions P of [a, b] . Also,
k 1

as h is integrable on [a, b] , so

32
 b

 h  L(h)  sup L(h, P) : P is a partition of [a, b  L(h, P)  0.

a

By applying (i) and (ii), we have

b b b b b b

 h    g  ( f )    g    f   g   f ,
a a a a a a

b b
which implies a
f g .
a

The first and second results in the above theorem enables to form certain algebraic
combinations of integrable functions.

Theorem 6.2 (Sandwich Rule) Let f , g :[a, b]  R be integrable functions. A function

b b
h is such that f ( x)  h( x)  g ( x), x [a, b] . If 
a
f dx   g dx , then h is integrable on
a

b b
[a, b] and  h dx  f dx .
a a

Proof: Let P  {a  t0  t1    tn  b} be a partition of [ a , b] . Since

f ( x)  h( x)  g ( x), x [a, b] , therefore, for all k  1, 2,..., n , we have

m( f ,[tk 1 , tk ])  m(h,[tk 1 , tk ])  m( g ,[tk 1, tk ]) ,

which implies

L( f , P)  L(h, P)  L( g , P) .

Analogously, we obtain

U ( f , P)  U (h, P)  U ( g , P) .

Now, L( f , P)  L(h, P)  sup{L(h, P) : P is a partition of [a, b]}  L(h) , which shows L(h)

is an upper bound of the set {L( f , P) : P is a partition of [a, b]} , hence

L( f )  L(h) (21)

33
and similarly, we have

U (h)  U ( g ) . (22)

b
But f and g are integrable on [a, b] , therefore,  f dx  L( f )  U ( f )
a
and

 g dx  L( g )  U ( g ) . From (21) and (22), it follows that

a

b b

 f dx  L(h)  U (h)   gdx .

a a

b b b
Also,  f dx  g dx ,
a a
therefore, we obtain L(h)  U (h)   f dx . This shows h is
a

b b
integrable on [a, b] and  h dx  f dx .
a a

The next result shows that the integral is an additive function of the interval over which
the function is integrated, i.e., if the interval [a, b] is divided into two subintervals, then

the integral of f :[a, b]  R is the sum of the integrals of the restrictions of f to the

two subintervals.

Theorem 6.3 (Domain Additivity Theorem) Let f be a bounded function on [a, b] .

Then f is integrable on [a, b] if and only it is integrable on [a, c] and [c, b] , where

c  (a, b) . In either case,

b c b


a
f  f  f
a c

Proof: Consider   0. If f is integrable on [a, b] , there exists a partition P of [a, b]

such that U ( f , P)  L( f , P)   . Let Q  P {c} . Then L( f , P)  L( f , Q) and

U ( f , Q)  U ( f , P) . Hence, U ( f , Q)  L( f , Q)  U ( f , P)  L( f , P)   . Let P1  Q  [a, c]

be a partition of [a, c] and P2  Q  [c, b] be a partition of [c, b] . Also,

U ( f , Q)  U ( f , P1 )  U ( f , P2 )

34
and

L( f , Q)  L( f , P1 )  L( f , P2 ) .

Therefore,

U ( f , Q)  L( f , Q)

 U ( f , P1 )  U ( f , P2 )  L( f , P1 )  L( f , P2 )  [U ( f , P1 )  L( f , P1)]  [U ( f , P2 )  L( f , P2 )]   ,

which implies

U ( f , P1 )  L( f , P1 )  

and

U ( f , P2 )  L( f , P2 )  

This shows f is integrable on [a, c] and [c, b] .

Conversely, suppose f is integrable on [a, c] and [c, b] . Then there exist partitions P1

and P2 of [a, c] and [c, b] , respectively such that U ( f , P1 )  L( f , P1 )  and
2

U ( f , P2 )  L( f , P2 )  .
2

The set P  P1  P2 is a partition of [a, b] and

U ( f , P)  U ( f , P1 )  U ( f , P2 ) , L( f , P)  L( f , P1 )  L( f , P2 ) .

Hence

 
U ( f , P)  L( f , P)  U ( f , P1 )  L( f , P1 )  U ( f , P2 )  L( f , P2 )    ,
2 2

implying that f is integrable on [a, b] .

We have
b c b


a
f  U ( f , P)  L( f , P)   = L( f , P1 )  L( f , P2 )    
a
f   f  .
c
Since  0 is

b c b
arbitrary, it follows that  f   f   f . To obtain the other inequality, we note that
a a c

35
b c b

 f  L( f , P)  U ( f , P)   = U ( f , P )  U ( f , P )     f   f   ,which leads to
a
1 2
a c

b c b b c b


a
f   f   f . Hence,
a c

a
f  f  f .
a c

Corollary 6.1 (Subinterval Property) If f is integrable on [a, b] and if [c, d ]  [a, b] ,

then the restriction of f to [c, d ] is integrable on [c, d ] .

Proof: Since f is integrable on [a, b] and c  [a, b] , it follows from Theorem 6.3 that its

restriction to [c, b] is integrable on [c, b] . Also, as d  [c, b] , then again Theorem 6.3

shows that the restriction of f to [c, d ] is integrable on [c, d ] .

Corollary 6.2 If f is integrable on [a, b] and if a  c0  c1    cm  b , then the

restrictions of f to each of the subintervals [ci 1 , ci ], i  1, 2,..., m are integrable and

b m ci

 f dx   
a i 1 ci 1
f dx .

Value Addition: Remark

If f :[a, b]  R is integrable over [c, b] for every c  (a, b) then f may not be
integrable over [a, b] . For example, consider f :[0,1]  R defined by

0, if x  0

f ( x)   1
 , 0  x 1
x
For every c  (0,1) , f is continuous on [c,1] and hence integrable over [c,1] but f

is not integrable over [0,1] .

Lemma 6.1 For a bounded function f on [a, b] ,

 
M ( f , S )  m( f , S )  sup f ( x1 )  f ( x2 ) : x1, x2  S ,

where S is a subset of [a, b] .

36
Proof: M ( f , S )  m( f , S ) is an upper bound of the set  f ( x )  f ( x ) : x , x  S . Let
1 2 1 2

  0 be a real number. Since m( f , S ) and M ( f , S ) are the infimum and supremum,


respectively of f in S , so there exist x3 and x4 in S such that f ( x3 )  M ( f , S )  and
2

f ( x4 )  m( f , S )  . Therefore, f ( x3 )  f ( x4 )  M ( f , S )  m( f , S )   , which implies that
2
any number less than M ( f , S )  m( f , S ) is not an upper bound of the set

 f ( x )  f ( x ) : x , x  S , hence the result.

1 2 1 2

b b
Theorem 6.4 If f is integrable on [a, b] , then f is integrable on [a, b] and 
a
f  f .
a

Proof: Consider   0. Since f is integrable on [a, b] , so there exist a partition

P  {a  t0  t1    tn  b} of [ a , b] such that U ( f , P)  L( f , P)   . For all

x1 , x2 [tk 1 , tk ], k  1, 2,..., n , we have

f ( x1 )  f ( x2 )  f ( x1 )  f ( x2 )  f ( x1 )  f ( x2 )

 M ( f ,[tk 1 , tk ])  m( f ,[tk 1 , tk ]) (using Lemma 6.1)

which gives

M ( f ,[tk 1 , tk ])  m( f ,[tk 1 , tk ])  sup  f ( x )  f ( x ) : x , x [t

1 2 1 2 k 1 
, tk ] (by Lemma 6.1)

 M ( f ,[tk 1 , tk ])  m( f ,[tk 1 , tk ])

Now,

n
U ( f , P)  L( f , P)    M ( f ,[tk 1 , tk ])  m( f ,[tk 1 , tk ])  tk  tk 1 
k 1

n
   M ( f ,[tk 1 , tk ])  m( f ,[tk 1 , tk ])  tk  tk 1 
k 1

 U ( f , P)  L( f , P)   ,

37
proving that f is integrable on [a, b] .

b b b b b
Since f  max( f ,  f ) , so by Theorem 6.1 (iii) ,  f 
a a
f and   f    f   f . This
a a a

b b
implies  f 
a a
f .

Note
Converse of the above theorem may not be true. For instance, consider the function
f :[a, b]  R defined as

1, when x is rational

f ( x)  
 1, when x is irrational
For this function, L( f )  (b  a) and U ( f )  b  a , implying f is not integrable on
b
[a, b] . But f ( x)  1,  x [a, b] is integrable on [a, b] and 
a
f dx  b  a .

Value Addition: Remark

f f
If f :[a, b]  R be a bounded function, then f  f   f  , where f   is the
2
f f
positive part of f and f  is the negative part of f . We see that f is
2
integrable if and only if f  and f  are integrable and
b b b


a
f dx  f  dx   f  dx Area(R f  ) Area(R f  ) ,
a a

where R f   ( x, y)  R 2 : a  x  b, 0  y  f ( x) and

R f   ( x, y)  R 2 : a  x  b, f ( x)  y  0 . This suggests that the integral of f on

[a,b] can be interpreted as the signed area of the corresponding region.

For example, in the following Figure 12, A1 , A2 and A3 represents the areas of the
individual parts and the integral of f over [a, b] is then equal to the region
between the graph of f and the x -axis with the understanding that the areas

38
 below the x -axis are assigned negative values.

Figure 12

Theorem 6.5 If f is integrable on [a, b] , then f 2 is integrable on [a, b] .

Proof: Consider   0. If f is integrable, then f is bounded on [a, b] with

f ( x)  B, x [a, b] for some B  0 . Since f is integrable on [a, b] , there exist a

partition P  {a  t0  t1    tn  b} of [a, b] such that


U ( f , P )  L( f , P )  .
2B

For all x, y [tk 1 , tk ], k  1, 2,..., n , we have

f 2 ( x)  f 2 ( y )  f ( x)  f ( y ) f ( x)  f ( y )

 2 B f ( x)  f ( y )

 2B  M ( f ,[tk 1 , tk ])  m( f ,[tk 1, tk ])  (By Lemma 6.1)

Applying Lemma 6.1, we get

M ( f 2 ,[tk 1 , tk ])  m( f 2 ,[tk 1, tk ])  2 B  M ( f ,[tk 1, tk ])  m( f ,[tk 1, tk ])  .

39
Hence,

n
U ( f 2 , P)  L( f 2 , P)   M ( f 2 ,[tk 1 , tk ])  m( f 2 ,[tk 1 , tk ])
k 1

n
 2 B M ( f ,[tk 1 , tk ])  m( f ,[tk 1 , tk ])
k 1


 2 B U ( f , P)  L( f , P)   2 B. 
2B

Therefore, f 2 is integrable on [a, b] .

Theorem 6.6 If f and g are integrable on [a, b] , then their product f g is integrable

on [a, b] .

Proof: Since the integral is linear, then we see from the identity,

1
 f  g    f  g   ,
2 2
fg 
4 

that f g is integrable using Theorems 6.1 and 6.5.

Note
Example of bounded functions f , g :[a, b]  R that are not integrable but f  g

and f g are integrable. Consider f , g :[a, b]  R defined as

 1, when x is rational
f ( x)  
1, when x is irrational
1, when x is rational
g ( x)  
 1, when x is irrational
Then, ( f  g )( x)  0 and ( f g )( x)  1, x [a, b] are integrable but neither f nor

g are integrable on [a, b] .

40
Theorem 6.7 If f and g are integrable on [a, b] and there exists a number  0
such that g ( x)   , x [a, b] , then f / g is integrable on [a, b] .

Proof: f and g being integrable are bounded on [a, b] , so there exists B0

such that f ( x)  B and g ( x)  B , x [a, b] which implies

f / g  ( x)  B /  , x [a, b] .

Hence f / g is bounded on [a, b] . Let P  {a  t0  t1    tn  b} be a partition of [a, b] .

For all x, y [tk 1 , tk ], k  1, 2,..., n , we have

f ( x) f ( y )
f / g  ( x)   f / g  ( y )  
g ( x) g ( y )

g ( y )[ f ( x)  f ( y )]  f ( y)[ g ( x)  g ( y)

g ( x) g ( y )

 ( B /  2 ) f ( x)  f ( y )  ( B /  2 ) g ( x)  g ( y )

 ( B /  2 )  M ( f ,[tk 1 , tk ])  m( f ,[tk 1, tk ])   ( B /  2 )  M ( g ,[tk 1, tk ])  m( g,[tk 1, tk ])  (By

Lemma 6.1).

Applying Lemma 6.1, we get

M ( f / g ,[tk 1 , tk ])  m( f / g ,[tk 1, tk ])

 ( B /  2 )  M ( f ,[tk 1 , tk ])  m( f ,[tk 1, tk ])   ( B /  2 )  M ( g ,[tk 1, tk ])  m( g,[tk 1, tk ])  (23)

Since f and g are integrable on [a, b] , there exists a  0 such that for all partitions

P of [a, b] having mesh( P)   ,

U ( f , P)  L( f , P)    2 / 2B,

U ( g , P)  L( g , P)    2 / 2B.

It follows from (23) that for any partition P having mesh( P)   ,

41
 U ( f / g , P)  L( f / g , P)  ( B /  2 ) U ( f , P   L( f , P)   ( B /  2 ) U ( g , P   L( g , P) 

 ( B /  2 )(  2 / 2B)  ( B /  2 )(  2 / 2B)   ,

proving that f / g is integrable on [a, b] .

Value Addition: Remark

A composition of continuous functions is continuous and a composition of
differentiable functions is differentiable but a composition of integrable functions
need not be integrable. Consider the function f :[0,1]  R defined by

0, for x  0
f ( x)  
1, for 0  x  1
and consider the Thomae’s function g :[0,1]  R defined by

0, when x  [0,1] is irrational

1, when x  0

g ( x)   p
1/q, when x is a non zero rational nummber and x  , p, q  N
q

 are relatively prime.
Then f and g are integrable on [0,1] (proved in Section 7). The composite

functions gof and fog are both defined on [0,1] and ( gof )( x)  1, x [0,1]
whereas

0, if x Q
( fog )( x)  
1, if x Q

The function gof is integrable on [0,1] whereas fog is not integrable on [0,1] .

Theorem 6.8 (Special Composition Integrability) Let f :[a, b]  R be integrable and

g :[c, d ]  R be continuous where f ([a, b])  [c, d ] . Then gof :[a, b]  R is integrable.

Proof: Consider  0 
and suppose r  sup g (t ) : c  t  d . Choose     0 such that


  . Since g is continuous on [c, d ] , it is uniformly continuous on [c, d ] , as such
b  a  2r
there exists a  0 such that     and

42
 g ( p)  g (q)    whenever p q , p, q [c, d ] (24)

Since f is integrable on [a, b] , so there exists a partition P  {a  t0  t1  ...  tn  b} of

[a, b] such that

U ( f , P)  L( f , P)   2 (25)

We claim that for this partition P

n
U ( gof , P)  L( gof , P)    M ( gof ,[tk 1 , tk ])  m( gof ,[tk 1 , tk ])  tk  tk 1    .
k 1

Let A  k : M ( f ,[tk 1 , tk ])  m( f ,[tk 1 , tk ])    and B  k : M ( f ,[tk 1 , tk ])  m( f ,[tk 1 , tk ])    .

Thus, if x, y [tk 1 , tk ], k  A , we have f ( x)  f ( y)  M ( f ,[tk 1 , tk ])  m( f ,[tk 1, tk ])   . So,

from (24), it follows that ( gof )( x)  ( gof )( y)   , x, y  A . Therefore,

sup  ( gof )( x)  ( gof )( y) : x, y  A   , that is, M ( gof ,[tk 1 , tk ])  m( gof ,[tk 1, tk ])    and so

  M ( gof ,[t
k A
k 1 , tk ])  m( gof ,[tk 1 , tk ])  tk  tk 1      tk  tk 1    (b  a) .
k A
(26)

If k  B , then M ( f ,[tk 1 , tk ])  m( f ,[tk 1, tk ])   and so

1
 t
kB
k  tk 1     M ( f ,[t
 kB k 1 , tk ])  m( f ,[tk 1 , tk ])  tk  tk 1 

1 1
 U ( f , P)  L( f , P)   2     (using (25))
 

Since M ( gof ,[tk 1 , tk ])  m( gof ,[tk 1, tk ])  2r, k  1, 2,..., n , we have

  M ( gof ,[t
kB
k 1 , tk ])  m( gof ,[tk 1 , tk ])  tk  tk 1   2r  tk  tk 1   2r 
kB
(27)

Thus, from (26) and (27), we obtain

U ( gof , P)  L( gof , P)

43
   M ( gof ,[tk 1 , tk ])  m( gof ,[tk 1 , tk ])  tk  tk 1     M ( gof ,[tk 1 , tk ])  m( gof ,[tk 1, tk ]) tk  tk 1 
kA kB

  (b  a)  2r   (b  a  2r )   

and hence, gof is integrable on [a, b] .

7. Classes of Integrable Functions

In this section, we will establish the integrability of several important classes of functions.

Theorem 7.1 Every monotonic function f on [a, b] is integrable.

Proof: We will prove the theorem for an increasing function f on [a, b] (the proof for the

decreasing case follows similarly). Then f is bounded on [ a , b] as

f (a)  f ( x)  f (b), x [a, b] . Consider  0 and choose a partition


P  {a  t0  t1    tn  b} of [ a , b] with mesh(P) < . Since f is
f (b)  f (a)  1
monotonically increasing, therefore

m( f ,[tk 1 , tk ])  f (tk 1 ), M ( f ,[tk 1 , tk ])  f (tk ), k  1, 2,..., n.

n
U ( f , P)  L( f , P)    M ( f ,[tk 1 , tk ])  m( f ,[tk 1 , tk ])  tk  tk 1 
k 1

n
   f (tk )  f (tk 1 )  mesh(P)
k 1

n

   f (tk )  f (tk 1 ) 
k 1 f (b)  f (a)  1

 n
   f (t )  f (t ) 
f (b)  f (a)  1 k 1
k k 1


  f (b)  f (a)   
f (b)  f (a)  1

Hence, f is integrable on [a, b] .

Value Addition: Remark

44
 Monotonicity is a sufficient condition for integrability. It is not necessary that every
integrable function is monotone. For instance, consider the function f :[2, 2]  R

defined as f ( x)  x 2 . Then f being continuous is integrable on [2, 2] (by the next

Theorem 7.2) but f is not monotone on [2, 2] as it is decreasing on [2, 0] and
increasing on [0, 2] .

Theorem 7.2 Every continuous function f on [a, b] is integrable.

Proof: Consider   0. Since every continuous function on a closed bounded interval is

bounded and uniformly continuous on that interval, so there exists a  0 such that


f ( x)  f ( y )  , if x  y   and x, y [a, b] (28)
ba

Choose a partition P  {a  t0  t1    tn  b} of [a, b] having mesh( P)   . As f assumes

its maximum and minimum on each subinterval [tk 1 , tk ] , k  1, 2,..., n , therefore, for each

k  1, 2,..., n , there exist xk , yk [tk 1 , tk ] such that f ( xk )  M ( f ,[tk 1 , tk ]) and

f ( yk )  m( f ,[tk 1, tk ]) . Also,

xk  yk  tk  tk 1  mesh( P)   .

So, by (28)


M ( f ,[tk 1 , tk ])  m( f ,[tk 1 , tk ])  f ( xk )  f ( yk )  , for each k  1, 2,..., n.
ba
Therefore,
n
U ( f , P)  L( f , P)    M ( f ,[tk 1 , tk ])  m( f ,[tk 1 , tk ])  tk  tk 1 
k 1

 n


ba
 (t k 1
k  tk 1 ) 
ba
(b  a)   .

This proves that f is integrable on [a, b] .

Value Addition: Remark

45
 Continuity is only a sufficient condition for integrability. It is not necessary that every
integrable function is continuous. For instance, consider the function f :[0, 2]  R

defined as f ( x)   x  , the greatest integer function. Then f is integrable as it is

monotonically increasing on [0, 2] (by Theorem 7.1) but f is not continuous in [0, 2] .

Example 7.1 Suppose f is a continuous function on [a, b] and f ( x)  0, x [a, b] . Prove

b
that if  f dx  0 , then
a
f ( x)  0 , x [a, b] .

We assume that f ( x0 )  0 for some x0  (a, b) . Since f is continuous at x0 , so

f ( x0 )
corresponding to   0 , there exists a   0 such that
2

f ( x)  f ( x0 )   whenever x  x0   ,

i.e., f ( x0 )    f ( x)  f ( x0 )   whenever x  ( x0   , x0   ) ,

f ( x0 ) f ( x0 )
which implies, f ( x)  f ( x0 )   whenever x  ( x0   , x0   )
2 2

b x0  x0  b


a
f dx  
a
f dx  
x0 
f dx  
x0 
f dx

x0  x0  b
 
x0 
f dx (since f ( x)  0  
a
f dx  0 and 
x0 
f dx  0 )

x0 
f ( x0 )

x0 
 2
dx ,

f ( x0 )
  2   f ( x0 )  0 .
2

Hence our assumption is wrong and f ( x)  0, x  (a, b) .

46
 f (a)
Suppose that f (a)  0 , then by the continuity of f , corresponding to   0 , there
2
exists a 1  0 such that

f ( x)  f ( a )   whenever 0  x  a  1

f (a)
 f ( x)  whenever x  (a, a  1 ) .
2

a 1 a 1 a 1

b b
f (a) 1 f (a)
 f dx  
a a
f dx  
a
f dx  a
f dx 
2 
a
dx 
2
0.
1

Hence our assumption is wrong and f (a)  0 . Similarly, it can be shown that f (b)  0 .

Value Addition: Remark

The above result may fail if f is not continuous on [a, b] , i.e., there are integrable
b
functions on [a, b] such that f ( x)  0 for all x [a, b] and  f dx  0,
a
but f ( x)  0 for

some x [a, b] . For example, consider the function f :[a, b]  R defined as

0, x  c
f ( x)  
1, x  c
b
where a  c  b . Here, f ( x)  0 for all x [a, b] and  f dx  0,
a
but f (c)  1  0 . Note

that f is discontinuous at c . Thus, the continuity hypothesis in the preceding example

cannot be dropped.

Example 7.2 Show that if f is a continuous real valued function on [a, b] satisfying
b

 f ( x) g ( x)dx  0
a
for every continuous function g on [a, b] , then f ( x)  0 for all x [a, b] .

47
 b
Let f be a continuous function on [a, b] satisfying  f ( x) g ( x)dx  0
a
for every continuous

  f ( x)  dx  0 . Also,
2
function g on [a, b] . By taking g  f , we obtain f 2 is a continuous
a

 f ( x) 
2
nonnegative function on [a, b] . Therefore, Example 7.1 implies  0, x [a, b] , from

which we get f ( x)  0 for all x [a, b] .

Theorem 7.3 A bounded function f which has only a finite number of points discontinuity

in [a, b] is integrable.

Proof: Consider   0 . Firstly, we assume that f has only one point of discontinuity, say

c1 [a, b] .

Case (i): c1  a and c1  b , i.e. c1  (a, b) . Let m  inf{ f ( x) : x [a, b]} and

M  sup{ f ( x) : x [a, b]} . Since f breaks at c1 , therefore m  M . Choose a1 and b1 in


[a, b] such that a1  c1  b1 and b1  a1  . Now, f is continuous on the
3( M  m)
subintervals [a, a1 ] and [b1 , b] and therefore integrable on these subintervals. There exists
partitions P1  {a  r0  r1    rn  a1} of [a, a1 ] and P2  {b1  s0  s1    sm  b} of [b1 , b]

such that U ( f , Pi )  L( f , Pi )  , i  1, 2 . Let P denote the partition of [a, b] obtained from
3
the points of P1 followed by the points of P2 , i.e. ,

P  {a  r0  r1    rn  a1  b1  s0  s1    sm  b} then

U ( f , P)  L( f , P)  U ( f , P1 )  L( f , P1 )  (M1  m1 )(b1  a1 )  U ( f , P2 )  L( f , P2 ) ,

where m1  inf{ f ( x) : x [a1 , b1 ]} and M1  sup{ f ( x) : x [a1 , b1 ]} . As M1  m1  M  m ,

  
therefore U ( f , P )  L( f , P )   ( M  m)    which shows f is integrable on
3 3( M  m) 3
[ a , b] .

48
 
Case (ii): Suppose c1  a . Choose a point t  (a, b) such that t  a  . The function
2( M  m)
f being continuous in [t , b] is integrable in [t , b] , therefore there exists a partition

P1  {t  t0  t1  ...  tn  b} of [t , b] such that U ( f , P1 )  L( f , P1 )  . For the partition
2
P  {a  t  t0  t1  ...  tn  b} of [a, b] ,

U ( f , P)  L( f , P)  U ( f , P1 )  L( f , P1 )  ( M1  m1 )(t  a) ,

where m1  inf{ f ( x) : x [a, t ]} and M1  sup{ f ( x) : x [a, t ]} . As M1  m1  M  m , therefore

 
U ( f , P )  L( f , P )   (M  m)   which shows f is integrable on [a, b] .
2 2( M  m)

Case (ii): Suppose c1  b . This case can be proved in a similar manner as we proved Case
(ii).

In the general case involving n points of discontinuities, c1 , c2 ,..., cn in [a, b] , the

integrability of f can be proved with the help of induction.

Value Addition: Remark

It follows that for a bounded function f to be non integrable on [a, b] , it must have

infinite number of points of discontinuity. But there are integrable functions with infinite
number of points of discontinuity. For instance, the ‘infinite-steps function’
f :[0,1]  R defined by

1 1 1
 when  x  , n  1, 2,...
f ( x)   n n 1 n

0 when x  0
is monotonically increasing on [0,1] and hence integrable, although it has an infinite

number of points of discontinuity.

Value Addition: Remark

49
 Example of an integrable function that is neither monotone nor continuous. Consider
f :[2, 2]  R defined by

 x2 ,  2  x  1
f ( x)  
2, 1 x  2
is a bounded function which has only one point of discontinuity at x  1 and therefore,
integrable on [2, 2] but is neither monotone nor continuous.

Theorem 7.4 If a function f is bounded in [a, b] and the set of the points of discontinuity

has a finite number of limit points, then f is integrable on [a, b] .

Proof: Consider   0. Firstly, we assume that S , the set of points of discontinuity of f

has only one limit point, say c1 [a, b] .

Case (i): c1  a and c1  b , i.e. c1  (a, b) . Let m  inf{ f ( x) : x [a, b]} and

M  sup{ f ( x) : x [a, b]} . Since f has discontinuity, therefore m  M . Choose a1 and b1 in


[a, b] such that a1  c1  b1 and b1  a1  . Since c1 is the only limit point of S ,
3( M  m)
therefore, only finite number of points of discontinuity of f lies in [a, a1 ] and [b1 , b] . So, by

Theorem 7.3, f is integrable on [a, a1 ] and [b1 , b] . Therefore, there exist partitions P1 of

[a, a1 ] and P2 of [b1 , b] such that U ( f , Pi )  L( f , Pi )  , i  1, 2 . Rest of the proof is same as
3
in Case (i) of Theorem 7.3.


Case (ii): Suppose c1  a . Choose a point t  (a, b) such that t  a  . Since ‘ a ’ is
2( M  m)
the only limit point of S , therefore f has only finite number of points of discontinuity in
[t , b] , so by Theorem 7.3, f is integrable in [t , b] . Rest of the proof is same as in Case (ii)
of Theorem 7.3.

Case (ii): Suppose c1  b . This case can be proved in a similar manner as in Case (ii).

50
In the general case involving n number of limit points of discontinuities, c1 , c2 ,..., cn in [a, b] ,
the integrability of f can be proved with the help of induction.

Example 7.3 Consider Thomae’s function f :[0,1]  R defined by


0, when x  [0,1] is irrational


f ( x)  1, when x  0
 p
1/q, when x is a non zero rational nummber and x  , p, q  N are relatively prime.

 q
1
Then f is integrable in [0,1] and  f dx  0 .
0

Here, f is bounded on [0,1] . Consider   0. There exists a largest integer q N such

1  p 1 
that  , so there are only a finite number of rational numbers for which  .
q 2 q q 2

Enclose such rational numbers in order in non-overlapping closed intervals [t1  0, t1 ],


[t2 , t2 ], ... [tn , tn  1] , sum of whose lengths is less than . In the remaining intervals
2

[t0  0, t1 ], [t1 , t2 ], ... , [tn 1 , tn ] , at each x, f ( x )  . Consider the partition
2

P  {0  t0  t1  t1  t2  t2  ...  tn  tn  1} of [0,1] . Since each subinterval contains

irrational number, therefore m( f ,[tk  , tk  ])  0 and m( f ,[tk 1 , tk  ])  0 , for each k  1, 2,..., n


which gives L( f , P)  0 . As f ( x)  ,  x  [tk 1 , tk  ], k  1, 2,..., n , therefore
2

M ( f ,[tk 1 , tk  ])  , k  1, 2,..., n . Also f ( x)  1, x [0,1] , hence
2

M ( f ,[tk , tk ])  1, k  1, 2,..., n . Therefore,

n n
U ( f , P)   M ( f ,[tk  , tk  ])(tk  tk )   M ( f ,[tk 1 , tk ])(tk  tk 1 )
k 1 k 1

n
 n
  
 1.(tk   tk )   (tk  tk 1 )   (1  )  
k 1 k 1 2 2 2 2

51
Hence, U ( f , P)  L( f , P)    0   , implying that f is integrable on [a, b] . In fact,
1

 f dx  sup{L( f , P) : P is a partition of [a,b]}=0.

0

In this example, the function is discontinuous at every rational number and is not monotone
but is integrable, nevertheless.

Example 7.4 Show that the function f defined as follows:

 1 1 1
 when  x  n 1 , n  1, 2,...
f ( x)   a n 1 a n
a

0 when x  0

1
a
where a  1 is integrable on [0,1] and  f dx  1  a , although it has an infinite number of
0

points of discontinuity.

 1
1, when a  x  1

 1 , when 1  x  1
a a2 a

Here, f ( x)  
 1 1 1
 n 1 , when n  x  n 1
a a a


0, when x  0

1 1
Thus, f is bounded on [0,1] and has infinite number of points of discontinuity 0, , ,...
a a2
having only one limit point, 0 and hence is integrable on [0,1] .

1 1 1/ a 1/ a 2

 f dx  
0 1/ a
f dx   2
f dx   3
f dx  
1/ a 1/ a

1 1/ a 1/ a 2
1 1
  1dx   a
dx   2 dx  
a
1/ a 1/ a 2 1/ a3

52
  1 1 1 1  1 1  1
 1      2    2  3  2  
 a a a a a a a

 1
1   a
a
  
 1  1 a
1  2 
 a 

8. Solved Problems
1. Suppose that f :[a, b]  R and that f ( x)  0 except for a finite number of points
b
c1 ,..., cn in [a, b] . Prove that f is integrable on [a, b] and  f dx  0 .
a

Solution: The function f has a finite number of discontinuities at c1 ,..., cn , so by

Theorem 7.3, it is integrable on [a, b] .

Let S  {c1 ,..., cn } . Then S is a finite subset of [a, b] such that f ( x)  0, x  S . Consider

a partition P  {a  t0  t1    tn  b} of [a, b] . Then, for each k  1,..., n , there are

points zk [tk 1 , tk ] such that zk  S . Therefore, f ( zk )  0 and hence, we have

m( f ,[tk 1 , tk ])  0 , for each k  1,..., n . This implies

n
L( f , P)   m( f ,[tk 1 , tk ])(tk  tk 1 )  0 .
k 1

Thus, L( f )  sup{L( f , P) : P is a partition of [a, b]}  0 . Since f is integrable on [a, b] ,

b
so by Theorem 6.4, f is integrable on [a, b] and a
f dx  L( f )  0 . Also, by Theorem

6.4, we have

b b

 f dx  
a a
f dx  0

53
 b b
and hence, 
a
f dx  0 , which shows  f dx  0 .
a

2. Let f be integrable on [a, b] and suppose that g is a function on [a, b] such that
g ( x)  f ( x) except for finitely many x in [a, b] . Show that g is integrable and
b b

 gdx  f dx .
a a

Solution: Consider h :[a, b]  R defined by

h( x)  ( g  f )( x), x [a, b]

Then, h( x)  0 except for a finite number of points in [a, b] . By the previous problem, h
b
is integrable on [a, b] and  h dx  0 . Also,
a
g  f  h and both f and h are integrable

on [a, b] , so g is integrable on [a, b] . We have

b b b b
0   h dx   ( g  f )dx   g dx   f dx
a a a a

b b
and hence,  gdx  f dx .
a a

3. Show that
 /2
2 x 2 2
(i)
9
 
 /6 sin x
dx 
9
2
 dx
(ii)
2
 
0 10  6sin x
 .

 
Solution: (i) For x , we have
6 2

 
sin  sin x  sin
6 2

1
  sin x  1
2

54
 1
1 2
sin x

x   
and therefore, we have x   2 x, x   ,  . All these three functions being
sin x 6 2
  
continuous are integrable on
 6 , 2  and

 /2  /2  /2
x
 /6 xdx /6 sin x dx /6 2 xdx

 /2
12 2  x 2 2 
     dx    
2  4 36   /6 sin x  4 36 

which leads to

 /2
2 x 2 2
9 /6 sin x
 dx  .
9

1
(ii) Let f ( x)  , x  [0, 2 ] . Then f being continuous is integrable on
10  6sin x
1 1 1
[0, 2 ] . Further, m  min{ f ( x) : x  [0, 2 ]}   
10  6 max(sin x) 16 4

1 1 1
and M  max{ f ( x) : x [0, 2 ]}    . By using the
10  6 min(sin x) 4 2
property:

b
m(b  a)   f dx  M (b  a) ,
a

2
1 1
we get,  2  0    f dx   2  0  and hence we obtain
4 0
2

2

2
 
0
f dx   .

55
9. Summary
In this chapter, we motivate and formulate the definition of Darboux integral [Definition
3.6] by noting that for a nonnegative function, the lower Darboux sum is the sum of the
areas of the inscribed rectangles and the upper Darboux sum is the sum of the areas of the
circumscribing rectangles. We give useful criteria for integrability and it is revealed that
integrability is really equivalent to the existence of partitions whose lower and upper
Darboux sums are arbitrarily close together. Next, we show that the integral of a function
can be approximated by certain sums using its values at arbitrary chosen points. This
approach yields Riemann’s definition of integral [Definition 5.2] which is a little different
from Definition 3.6 but we have established in Theorem 5.1 that the two definitions are
equivalent. Further, we have investigated a number of important algebraic and order
properties of integrable functions. We have established the integrability of several important
classes of functions. However, it is seen that there are functions that are not Riemann
integrable. Several examples are provided for understanding of the concepts.

10. Exercises
Question 1. Find the upper and lower Darboux integrals for f ( x)  x3 on [0, b] , where
b  0 . Is f integrable on [0, b] ?
Question 2. Let f , g :[a, b]  R be integrable. Show that the functions max( f , g ) and
min( f , g ) are integrable on [a, b] .

1 1
[Hint: max( f , g )   f  g  f  g  and min( f , g )   f  g  f  g  ].
2 2

Question 3. Let f :[a, b]  R . Show that f is not Riemann integrable over [a, b] if and

only if there exists 0  0 such that for every natural number n, there exist partitions Pn

1 1
and Qn with mesh( Pn )  and mesh(Qn )  such that S ( f , Pn )  S ( f , Qn )   0 .
n n

Question 4. If f is integrable on [0,1] , show that

1
1 1 2  n 
lim  f   f    f      f ( x)dx .
n  n
 n n  n  0

56
Question 5. Prove that f :[a, b]  R is integrable if and only if for every   0 , there exist
functions  and  which are integrable on [a, b] satisfying

 ( x)  f ( x)   ( x), x [a, b]

b
and such that       x  dx   .
a

Question 6. Let f :[0,3]  R be defined by

0, for 0  x  1

f ( x)  2, for 1  x  2
1, for 2  x  3


Show that f is neither monotonic nor continuous on [0,3] but is integrable on [0,3] and
3
compute  f ( x)dx .
0

Question 7. Prove that f :[a, b]  R defined by

 x, when x is rational
f ( x)  
 x, when x is irrational

is not integrable but f is integrable over [a, b] .

Question 8. If f is nonnegative integrable function on [a, b] , show that f is integrable

on [a, b] .

Question 9. Let f :[0,1]  R be defined by

k , for x  1/ k (k  N )
f ( x)  
0, elsewhere

Show that f is not integrable over [0,1] .

Question 10. Let f :[0,1]  R be the broken-line function given by

57
 0, for x  0 or x  1/ n for some n  N
f ( x)  
1, elsewhere

Show that f is integrable over [0,1] .

11. References
[1] Tom M. Apostol, Mathematical Analysis: A Modern Approach to Advanced Calculus,
Addison-Wesley Publishing Company, 1974.
[2] W. Rudin, Principles of Mathematical Analysis (Third Edition), McGraw-Hill, 1976.
[3] R.G. Bartle and D.R. Sherbert, Introduction to Real Analysis (Third Edition), John
Wiley and Sons (Asia) Pvt. Ltd., Singapore, 2002.
[4] K.A. Ross, Elementary Analysis: The Theory of Calculus, Undergraduate Texts in
Mathematics, Springer (SIE), Indian Reprint, 2004.
[5] Sudhir R. Ghorpade and Balmohan V. Limaye, A Course in Calculus and Real
Analysis, Undergraduate Texts in Mathematics, Springer (SIE), Indian Reprint, 2006.
[6] Stephen Abbott, Understanding Analysis, Springer, Indian Reprint, 2008.

58