4 Denite Integral

4.1 Denition, Necessary & sucient conditions

Let f : [a, b] R be a bounded real valued function on the closed, bounded interval [a, b].
Also let m, M be the inmum and supremum of f(x) on [a, b], respectively.
Denition 4.1.1. A partition P of [a, b] is an ordered set P = {a = x
0
, x
1
, x
2
, ..., x
n
= b}
such that x
0
< x
1
< ... < x
n
.
Let m
k
and M
k
be the inmum and supremum of f(x) on the subinterval [x
k1
, x
k
],
respectively.
Denition 4.1.2. Lower sum: The Lower sum, denoted with L(P, f) of f(x) with
respect to the partition P is given by
L(P, f) =
n

k=1
m
k
(x
k
x
k1
).
Denition 4.1.3. Upper sum: The Upper sum, denoted with U(P, f) of f(x) with
respect to the partition P is given by
U(P, f) =
n

k=1
M
k
(x
k
x
k1
).
For a given partition P, U(P, f) L(P, f). In fact the same inequality holds for any two
partitions. (see Lemma (4.1.6) below.)
Denition 4.1.4. Renement of a Partition: A partition Q is called a renement
of the partition P if P Q.
The following is a simple observation.
Lemma 4.1.5. If Q is a renement of P, then
L(P, f) L(Q, f) and U(P, f) U(Q, f).
Proof. Let P = {x
0
, x
1
, x
2
, ..., x
k1
, x
k
, ..., x
n
} and Q = {x
0
, x
1
, x
2
, ..., x
k1
, z, x
k
, ..., x
n
}.
1
Then
L(P, f) = m
0
(x
1
x
0
) + ... + m
k
(x
k
x
k1
) + ... + m
n1
(x
n
x
n1
)
m
0
(x
1
x
0
) + ... + m

k
(x
k
z) + m

k
(z x
k1
) + ... + m
n1
(x
n
x
n1
)
= L(Q, f)
where m

k
= inf
[z,x
k
]
f(x) and m

k
= inf
[x
k1
,z]
f(x).
Lemma 4.1.6. If P
1
and P
2
be any two partitions, then
L(P
1
, f) U(P
2
, f).
Proof. Let Q = P
1
P
2
. Then Q is a renement of both P
1
and P
2
. So by Lemma (4.1.8),
L(P
1
, f) L(Q, f) U(Q, f) U(P
2
, f).
Denition 4.1.7. Let P be the collection of all possible partitions of [a, b]. Then the
upper integral of f is

b
a
f = inf{U(P, f) : P P}
and lower integral of f is

b
a
f = sup{L(P, f) : P P}.
An immediate consequence of Lemma (4.1.6) is
Lemma 4.1.8. For a bounded function f : [a, b] IR,

b
a
f

b
a
f.
Denition 4.1.9. Riemann integrability: f : [a, b] R is said to be Riemann inte-
grable if

b
a
f =

b
a
f
and the value of the limit is denoted with

b
a
f(x)dx. We say f R[a, b].
Example 1: Consider f(x) = x on [0, 1] and the sequence of partitions P
n
= {0,
1
n
,
2
n
, ...,
n1
n
,
n
n
}.
2
Then
L(P
n
, f) = 0
1
n
+
1
n

1
n
+ ... +
n 1
n
1
n
=
1
n
2
[1 + 2 + ... + (n 1)]
=
n(n 1)
2n
2
Thus lim
n
L(P
n
, f) =
1
2
. Hence from the denition

1
0
f(x)dx
1
2
. Similarly
U(P
n
, f) =
1
n

1
n
+
2
n

1
n
+ ... +
n
n
1
n
=
1
n
2
[1 + 2 + ... + n]
=
n(n + 1)
2n
2
Hence lim
n
U(P
n
, f) =
1
2
. Again from the denition

1
0
f(x)dx
1
2
.
Example 2: Consider f(x) = x
2
on [0, 1] and the sequence of partitions P
n
= {0,
1
n
,
2
n
, ...,
n1
n
,
n
n
}.
Then
U(P
n
, f) =
1
n
2

1
n
+
(
2
n
)
2

1
n
+ ... +
(
n
n
)
2
1
n
=
1
n
3
[1 + 2
2
+ ... + n
2
]
=
n(n + 1)(2n + 1)
6n
3
Thus lim
n
U(P
n
, f) =
1
3
. Similarly
L(P
n
, f) = 0
1
n
+
(
1
n
)
2

1
n
+ ... +
(
n 1
n
)
2
1
n
=
1
n
3
[1 + 2
2
+ ... + (n 1)
2
]
=
n(n 1)(2n 1)
6n
3
Therefore, lim
n
L(P
n
, f) =
1
3
.
3
Hence from the denition

b
a
f 1/3 and

b
a
f 1/3.
Remark 4.1. In the above two examples

1
0
f =

1
0
f thanks to Lemma 4.1.8
The following example illustrates the non-integrability.
Example 3: On [0, 1], dene f(x) =

1, x Q,
0, x Q.
Let P be a partition of [0, 1]. In any sub interval [x
k1
, x
k
], there exists a rational number
and irrational number. Then the supremum in any subinterval is 1 and inmum is 0.
Therefore, L(P, f) = 0 and U(P, f) = 1. Hence

1
0
f =

1
0
f.
Necessary and sucient condition for integrability
Theorem 4.1.10. A bounded function f R[a, b] if and only if for every > 0, there
exists a partition P

such that
U(P

, f) L(P

, f) < .
Proof. : Let > 0. Then from the denition of upper and lower integral we have

b
a
f

b
a
f U(P

, f) L(P

, f) < ( by hypothesis).
Thus the conclusion follows as > 0 is arbitrary.
: Conversely, since

b
a
f is the inmum, for any > 0, there exists a partition P
1
such
that
U(P
1
, f) <

b
a
f +

2
.
Similarly there exists a partition P
2
such that
L(P
2
, f) >

b
a
f

2
.
4
Let P

= P
1
P
2
. Then P

is a renement of P
1
and P
2
. Hence
U(P

, f) L(P

, f) U(P
1
, f) L(P
2
, f)

b
a
f +

2

b
a
f +

2
= (as f is integrable,

b
a
f =

b
a
f)
This complete the theorem. ///
Now it is easy to see that the functions considered in Example 1 and Example 2 are
integrable. For any > 0, we can nd n (large) and P
n
such that
1
n
< . Then
U(P
n
, f) L(P
n
, f) =
1
2n
2
(n(n + 1) n(n 1)) =
1
n
< .
Similarly one can choose n in Example 2.
Remark 4.2. f : [a, b] R is integrable if and only if there exists a sequence {P
n
} of
partitions of [a, b] such that
lim
n
U(P
n
, f) L(P
n
, f) = 0.
Remark 4.3. Let S(P, f) =
n

i=1
f(
i
)(x
i
x
i1
),
i
[x
i1
, x
i
]. Then we have the follow-
ing
m(b a) L(P, f S(P, f) U(P, f) M(b a).
In fact, one has the following Darboux theorem:
Theorem 4.1.11. Let f f : [a, b] IR be a Riemann integrable function. Then for a
given > 0, there exists > 0 such that for any partition P with P := max
1in
|x
i
x
i1
| <
, we have
|S(P, f)

b
a
f(x)dx| < .
Corollary: If f R[a, b], then for any sequence of partitions {P
n
} with P
n
0, we
have L(P
n
, f)

b
a
f(x)dx and U(P
n
, f)

b
a
f(x)dx.
Remark 4.4. From the above theorem, we note that if there exists a sequence of partition
{P
n
} such that P
n
0 and U(P
n
, f)L(P
n
, f) 0 as n , then f is not integrable.
5
Problem: Show that the function f : [0, 1] IR
f(x) =

1 + x x Q
1 x x Q
is not integrable.
Solution: Consider the sequence of partitions P
n
= {0,
1
n
,
2
n
, ....,
n
n
= 1}. Then
U(P
n
, f) = (1 +
1
n
)
1
n
+ (1 +
2
n
)
1
n
+ .... + (1 +
n
n
)
1
n
= 1 +
1
n
2
(1 + 2 + ... + n)

3
2
as n
Now using the fact that inmum of f on [0,
1
n
] is 1
1
n
, though it is not achieved, we get
L(P
n
, f) = (1
1
n
)
1
n
+ (1
2
n
)
1
n
+ .... + (1
n
n
)
1
n

1
2
as n .
Hence f is not integrable.
Problem: Consider f(x) =
1
x
on [1, b]. Divide the interval in geometric progression and
compute U(P
n
, f) and L(P
n
, f) to show that f R[1, b].
Solution: Let P
n
= {1, r, r
2
, ..., r
n
= b} be a partition on [1, b]. Then
U(P
n
, f) = f(1)(r 1) + f(r)(r
2
r) + .... + f(r
n1
)(r
n
r
n1
)
= (r 1) +
1
r
r(r 1) + .. +
1
r
n1
r
n1
(r 1)
= n(r 1)
= n(b
1
n
1)
Therefore lim
n
U(P
n
, f) = lim
n
b
1
n
1
1
n
= lim
n
b
1
n
ln b(
1
n
2
)
1
n
2
= ln b.
6
Similarly
L(P
n
, f) = f(r)(r 1) + f(r
2
)(r
2
r) + ... + f(r
n
)(r
n
r
n1
)
=
1
r
(r 1) + .. +
1
r
n
r
n1
(r 1)
=
n
r
(b
1
n
1)
= n(1
1
b
1/n
)
=
b
1/n
1
b
1/n
1
n
ln b as n .
Theorem 4.1.12. Suppose f is a continuous function on [a, b]. Then f R[a, b].
Proof. Let > 0. By Theorem 4.1.10, we need to show the existence of a partition P
such that
U(P, f) L(P, f) < .
Since f is continuous on [a, b], this implies f is uniformly continuous on [a, b]. Therefore
there exists > 0 such that
|x y| < |f(x) f(y)| <

(b a)
. (4.1)
Now choose a partition P such that
sup
1kn
|x
k
x
k1
| < . (4.2)
As f is continuous on [a, b] there exist x

k
, x

k
(x
k1
, x
k
) such that m
k
= f(x

k
) and
M
k
= f(x

k
). By (4.2), |x

k
x

k
| < and hence by (4.1) |f(x

k
) f(x

k
)| <

(ba)
. Thus
U(P, f L(P, f) =
n

k=1
(M
k
m
k
)(x
k
x
k1
)
=
n

k=1
(f(x

k
) f(x

k
))(x
k
x
k1
)

(b a)
n

k=1
(x
k
x
k1
) =

(b a)
(b a) = .
Therefore f R[a, b].
7
Integrability and discontinuous functions: We study the eect of discontinuity on
integrability of a function f(x).
Example: Consider the following function f : [0, 1] R.
f(x) =

1, x =
1
2
0, x =
1
2
Clearly U(P, f) = 1 for any partition P. We notice that L(P, f) will be less than 1. We
can try to isolate the point x =
1
2
in a subinterval of small length. Consider the partition
P

= {0,
1
2


2
,
1
2
+

2
, 1}. Then L(P

, f) = (
1
2


2
) + (1
1
2


2
) = 1 . Therefore, for
given > 0 we have U(P

, f) L(P

, f) = . Hence f is integrable.
In fact we have the following theorem.
Theorem 4.1.13. Suppose f : [a, b] R be a bounded function which has nitely many
discontinuities. Then f R[a, b].
Proof follows by constructing suitable partition with sub-intervals of suciently small
length around the discontinuities as observed in the above example.
4.2 Properties of Denite Integral:
Property 1: For a constant c R,

b
a
cf(x)dx = c

b
a
f(x)dx.
Property 2: Let f
1
, f
2
R[a, b] . Then

b
a
(f
1
+ f
2
)(x)dx =

b
a
f
1
(x)dx +

b
a
f
2
(x)dx.
Easy to show that for any partition P,
U(P, f
1
+ f
2
) U(P, f
1
) + U(P, f
2
) (4.3)
L(P, f
1
+ f
2
) L(P, f
1
) + L(P, f
2
) (4.4)
Since f
1
, f
2
are integrable, for > 0 there exists P
1
, P
2
such that
U(P
1
, f
1
) L(P
1
, f
1
) <
U(P
2
, f
2
) L(P
2
, f
2
) <
8
Now taking P = P
1
P
2
, if necessary, we assume
U(P, f
1
) L(P, f
1
) < , U(P, f
1
) L(P, f
2
) < (4.5)
Therefore, using (4.3)-(4.5) we get
U(P, f
1
+ f
2
) L(P, f
1
+ f
2
) U(P, f
1
) + U(P, f
2
) L(P, f
2
) L(P, f
2
)
< + = 2.
Hence, f
1
+ f
2
is integrable.

b
a
(f
1
+ f
2
)(x)dx = lim
n
S(P
n
, f
1
+ f
2
) = lim
n
n

k=1
(f
1
+ f
2
)(
k
)(x
k
x
k1
)
= lim
n
n

k=1
f
1
(
k
)(x
k
x
k1
) + lim
n
n

k=1
f
2
(
k
)(x
k
x
k1
)
=

b
a
f
1
(x)dx +

b
a
f
2
(x)dx
Property 3: If f(x) g(x) on [a, b]. Then

b
a
f(x)dx

b
a
g(x)dx.
First we note that m f(x) M implies m(b a)

b
a
f(x) M(b a). Then
Property 1 and f(x) g(x) imply

b
a
(g f) 0 or

b
a
g(x)dx

b
a
f(x)dx.
Property 4: If f R[a, b] then |f| R[a, b] and |

b
a
f(x)dx|

b
a
|f|(x)dx. Let
m

k
= inf
[x
k1
,x
k
]
|f|(x) and M

k
= sup
[x
k1
,x
k
]
|f|(x). Then we claim
Claim: M
k
m
k
M

k
m

k
Proof of Claim: Note that for x, y [x
i1
, x
i
],
|f|(x) |f|(y) |f(x) f(y)| M
i
(f) m
i
(f).
Now take supremum over x and inmum over y, to conclude the claim.
9
Now since f is integrable, there exists partitions {P
n
} such that lim
n
U(P
n
, f)L(P
n
, f) =
0. i.e.,
lim
n
n

k=1
(M
k
m
k
)(x
k
x
k1
) = 0.
This implies
lim
n
n

k=1
(M

k
m

k
)(x
k
x
k1
) = 0.
Hence |f| is integrable. Note that |f| f |f|. Thus by Property 3 we get

b
a
|f|(x)dx

b
a
f(x)dx

b
a
|f|(x)dx =|

b
a
f(x)dx|

b
a
|f|(x)dx.
Property 5: Let f be bounded on [a, b] and let c (a, b). Then f is integrable on [a, b]
if and only if f is integrable on [a, c] and [c, b]. In this cases

b
a
f(x)dx =

c
a
f(x)dx +

b
c
f(x)dx.
Let f be integrable on [a, b]. For > 0, there exists partition P such that
U(P, f) L(P, f) < . (4.6)
With out loss of generality we can assume that P contain c. (otherwise we can rene
P by adding c and the dierence will be closer than before) Let P
1
= P [a, c] and
P
2
= P [c, b]. Then P
1
and P
2
are partitions on [a, c] and [c, b] respectively. Also by
(4.6), U(P
1
, f) L(P
1
, f) < and U(P
2
, f) L(P
2
, f) < . This implies f is integrable
on [a, c] and [c, b]. Conversely, suppose f is integrable on [a, c] and [c, b]. Then for > 0,
there exists partitions P
1
of [a, c] and P
2
of [c, b] such that U(P
1
, f) L(P
1
, f) <

2
and
U(P
2
, f) L(P
2
, f) <

2
. Now take P = P
1
P
2
. Then U(P, f) L(P, f) < . So by
Remark 4.3, there exists {P
n
} such that

b
a
f(x)dx = lim
n
S(P
n
, f) = lim
n

P
n
f(
k
)(x
k1
x
k
)
=

P
n
[a,c]
f(
k
)(x
k
x
k1
) +

P
n
[c,b]
f(
k
)(x
k
x
k1
)

c
a
f(x)dx +

b
c
f(x)dx
10
Example: Consider the following function f : [0, 1] R.
f(x) =

1 x =
1
n
for some n N, n 2
0 x =
1
n
Then f is Riemann integrable.
Solution: Let > 0. Choose N such that
1
N
<

2
. Note that f(x) has only nitely many
discontinuities in [
1
N
, 1] say
1
,
2
, ...,
r
. Dene the partition P

as
P

= {0,
1
N
,
1

4r
,
1
+

4r
, ...,
r


4r
,
r
+

4r
, 1}.
Since
r
is the last discontinuity, f = 0 in [
r
+

4r
, 1]. Now L(P

, f) = 0 and
U(P

, f) = 1
1
N
+

2r
+

2r
+ ... +

2r
+ 0 (1
r


4r
)
=
1
N
+

2
< .
Alternatively, using the approach of isolating the discontinuous points, we may take the
Partition with length of

2
k
at each discontinuous points x
k
, k = 1, 2, 3, 4, ....... Then using
the fact that

1
2
k
converges, one can prove that f is integrable.
Example: Consider the following function f : [0, 1] R.
f(x) =

0 x Q
sin
1
x
x Q
Then f is not Riemann integrable.
Solution: We will show that f is not integrable on a sub interval of [0, 1]. Consider the
f on the subinterval I
1
= [
2

, 1]. Clearly L(P, f) = 0 for any partition P of I

1
because
f(x) 0 in the sub interval [
2

, 1]. Let M
k
be the Supremum of f on subintervals [x
k1
, x
k
]
of [
2

, 1]. Also the minimum of M

k
s is sin 1. Therefore,
U(P, f) =
n

k=1
f(
k
)(x
k
x
k1
) > (1
2

) sin 1.
11
Hence U(P, f) L(P, f) can not be made less than for any < (1
2

) sin 1. ///
Mean Value Theorem
Theorem 4.2.1. Let f(x) be a continuous function on [a, b]. Then there exists [a, b]
such that

b
a
f(x)dx = f()(b a).
Proof. Let m = min
x[a,b]
f(x) and M = max
x[a,b]
f(x). Then by Property 3, we have
m(b a)

b
a
f M(b a),
i.e.
m
1
(b a)

b
a
f M.
Now since f(x) is continuous, it attains all values between its maximum and minimum
values. Therefore there exists [a, b] such that f() =
1
(b a)

b
a
f.
Fundamental Theorem
Theorem 4.2.2. Let f(x) be a continuous function on [a, b] and let (x) =

x
a
f(s)ds.
Then is dierentiable and

(x) = f(x).
Proof. As
(x + x) (x)
x
=
1
x

x+x
x
f(s)ds, By Mean value theorem, there exists
[x, x + x] such that

x+x
x
f(s)ds = xf().
Therefore lim
x0
(x + x) (x)
x
= lim
x0
f(). Since f is continuous, lim
x0
f() =
f( lim
x0
) = f(x). Thus

(x) = f(x). ///

Remark 4.5. It is always not true that

b
a
f

(x)dx = f(b) f(a).

For example, take f(x) = x
2
sin
1
x
for x = 0 and f(0) = 0. Then f is dierentiable on
[0, 1]. Here the derivatives at the end points are the left/right derivatives. It is easy to
check that f

(x) = 2xsin
1
x
2

2
x
cos
1
x
2
for x (0, 1) and f

(0) = 0. Therefore f

is not
12
bounded and so not integrable.
Denition 4.2.3. A function F(x) is called anti-derivative of f(x), if F

(x) = f(x).
Second Fundamental Theorem:
Theorem 4.2.4. Suppose F(x) is an anti- derivative of continuous function f(x). Then

b
a
f(x)dx = F(b) F(a).
Proof. By First fundamental theorem, we have
d
dx

x
a
f(s)ds = f(x).
Also F

(x) = f(x). Hence

x
a
f(s)ds = F(x) + c for some constant c R. Taking x = a,
we get c = F(a). Now taking x = b we get

b
a
f(x)dx = F(b) F(a). ///
Change of Variable formula
Theorem 4.2.5. Let u(t), u

(t) be continuous on [a, b] and f is a continuous function on

the interval u([a, b]). Then

b
a
f(u(x))u

(x)dx =

u(b)
u(a)
f(y)dy.
Proof. Note that f([a, b]) is a closed and bounded interval. Since f is continuous, it has
primitive F. i.e., F(x) =

x
a
f(t)dt. Then by chain rule of dierentiation,
d
dt
F(u(t)) =
F

(u(t))u

(t). i.e., F(u(t)) is the primitive of f(u(t))u

(t) and by Newton-Leibnitz formula,

we get

b
a
f(u(t))u

(t)dt = F(u(b)) F(u(a)).

On the other hand, for any two points in u([a, b]), we have (by Newton-Leibnitz formula)

B
A
f(y)dy = F(B) F(A).
Hence B = u(b) and A = u(a).
Example: Evaluate

1
0
x

1 + x
2
dx.
13
Taking u = 1 + x
2
, we get u

= 2x and u(0) = 1, u(1) = 2. Then

1
0
x

1 + x
2
dx =
1
2

2
1

udu =
1
3
u
2
3
2
u=1
=
1
3
(2
2
3
1).
References
1. Understanding Analysis, Abbott,S.
2. Elementary Analysis: The Theory of Calculus, K. A. Ross.
3. Calculus, G. B. Thomas and R. L. Finney, Pearson .
14