Chapter 14 Improper Integral and Lebesgue Integral

Introduction.
The concept of Riemann integral of a function presupposes that the function is
defined on a bounded subset of R , usually a connected interval and the function is
necessarily bounded on this bounded and connected interval. Since the function is
bounded, by arbitrarily assigning values at the boundary or end points of the
connected interval, we may always assume that the domain of the function is a closed
and bounded interval, say [a, b]. Lebesgue's Theorem then asserts that a function f :
[a, b] R is Riemann integrable if and only if it is continuous except on a set of
measure zero in [a, b]. Hence we have two obvious situations when a function is not
Riemann integrable. The first is when f is discontinuous on a set of positive
measure even though f is bounded. The second is when either f is not bounded or
the domain is not a bounded subset of R. We shall be concern with this later case.
Obviously this presents infinite possibilities and situations. We shall describe the
more manageable ones here.
We shall first consider functions on unbounded domains. Domains are assumed to be
connected.
Suppose f :[0, ) R is such that for any b > 0, the restriction of f to [0, b] is
Riemann integrable. Cauchy's definition of the improper integral is defined to be
.
0
+
f =
bd+
lim
0
b
f (x)dx
We shall see later that this definition is equivalent to the Lebesgue integral L
[0,)
f
if f is non-negative and the limit exists. We say f is improperly integrable if the
limit exists and is finite.
bd+
lim
0
b
f (x)dx


f(x)
x
y
Note that by Lebesgue'sTheorem f is integrable on [0, b] if and only if f is
continuous except on a set of measure zero in [0, b]. Thus if f is Riemann integrable
on [0, b] for any b > 0, then f is continuous on [0, ) except on a set of measure zero.
We can deduce this as follows. Take b to be any integer n and let Sn to be the set of
measure zero in [0, b] on which f is discontinuous. Then the set S on which f is
discontinuous is a countable union {Sn : n P} and is a set of measure zero, since
each Sn is of measure zero and the measure of S, (S) (Sn) = 0, where is the
Lebesgue measure. This means f is continuous almost everywhere on [0, ) and so f
is (Lebesgue) measurable on [0, ).
Similarly, for function f :(-, 0] R such that f is Riemann integrable on [a, 0] for
any a < 0, we can define the improper integral.
.

0
f =
ad
lim
a
0
f (x)dx
And if further more f is non-negative and if the limit exists, then the improper
integral is also the Lebesgue integral.
Plainly, for non-positive function f satisfying either f is Riemann integrable on [0,
b] for any b > 0 or f is Riemann integrable on [a, 0] for any a < 0, the improper
integral or , if it exists is also the Lebesgue integral.
0
+
f

0
f
For non-negative function f defined on the whole of R, which is the only interval
unbounded on both sides, we can define the improper integral as
,

f =
td
lim
t
t
f
if f is Riemann integrable on [t, t] for each t > 0.
We would also like to use the convergence theorems in Lebesgue integration theory
to deduce statements about improper Riemann integrals. The notion of Lebesgue
integrability is actually a notion about absolute integrability, meaning f is Lebesgue
integrable if and only if | f | is Lebesgue integrable. Thus if a function f is such that
the improper integral of f exists but the improper integral of | f | does not exist, then
it is not necessary that the improper Riemann integral is equal to the Lebesgue
integral. For instance, take the function f defined by f (x) = sin(x)/x for x 0 and
f (0) = 1, the improper Riemann integral exists and equals /2. But the
0
+
f
improper integral does not exist. We can deduce this as follows. First note
0
+
| f |
that for each integer k > 0. Thus we
(k1)v
kv sin(x)
x
dx m
1
kv

(k1)v
kv
sin(x) dx =
2
kv
get,
.
0
nv
f (x =
E
k=1
n

(k1)v
kv sin(x)
x
dx m
E
k=1
n
2
kv
And so since is divergent, does not exist. It follows that f is not
E
k=1

2
kv

0
+
| f |
Lebesgue integrable.

14.1 Improper Integral on Unbounded Domain
Definition 1.
1. Suppose f :[a, ) R is such that f is Riemann integrable on every closed
sub-interval of [a, ). We define the improper Riemann integral
.
a
+
f (x)dx =
td+
lim
a
t
f (x)dx
2. If f : (, b] R is Riemann integrable on every closed sub-interval of (,
b], then we define the improper integral
.

b
f (x)dx =
td
lim
t
b
f (x)dx
Chapter 14 Improper Integral and Lebesgue Integral
2
Ng Tze Beng
3. If f : R R is such that f is Riemann integrable on every closed sub-interval of
R, then we define the improper integral,

+
f (x)dx =
td
lim
t
c
f (x)dx +
td+
lim
c
t
f (x)dx
for some c in R.
In each case if the limit exists, then we say the improper integral is convergent,
otherwise it is divergent.
Plainly, if exists, then exists for any r a. Similar conclusion for the
a
+
f
r
+
f
other two types of improper integral is valid.
In general, convergence of the integral (respectively ,
a
+
f (x)dx

b
f (x)dx
) does not imply the convergence of (respectively

+
f (x)dx

+
f (x) dx
, ).

b
f (x) dx

+
f (x) dx
If (respectively , ) exists, then we say the

+
f (x) dx

b
f (x) dx

+
f (x) dx
improper integral is absolutely convergent. Absolute convergence amounts to the
function f being Lebesgue integrable.
First we have a convergence theorem.
Theorem 2. Suppose f :[a, ) R is Riemann integrable on every closed
sub-interval of [a, ). Then the improper integral exists if and only if
a
+
f (x)dx
for every > 0, there exists a number N > 0 (depending on ) such that for all s and t
> N,
.
s
t
f (x)dx < c
Proof. Suppose exists. Then given any > 0 there exsits N > 0 such I =
a
+
f (x)dx
that
for r > N. I
a
r
f (x)dx <
c
2
Therefore,
s
t
f (x)dx =
a
t
f (x)dx
a
s
f (x)dx =
a
t
f (x)dx I + I
a
s
f (x)dx
[ I
a
t
f (x)dx + I
a
s
f (x)dx <
c
2
+
c
2
= c
for r, s > N.
Conversely, assume the condition is satisfied, that is, for every > 0, there exists a
number N > 0 such that for all s and t > N, .
s
t
f (x)dx < c
Let (an ) be a sequence in [a, ) such that an . Then there exists an integer M > 0
such that n M an > N. Hence the sequence

a
an
f (x)dx
satisfies that . n, m m Mu
a
an
f (x)dx
a
am
f (x)dx =
an
am
f (x)dx < c
Therefore, is convergent and so by the Cauchy Princoiple of
a
an
f (x)dx
Convertgence, it is convergent. Suppose . We shall show that
a
an
f (x)dx d A
. We shall show that for any sequence (bn ) in [a, ) such that bn ,
a

f (x)dx = A
. Suppose . For any > 0, there exists an integer N
1
a
bn
f (x)dx d A
a
bn
f (x)dx d B
such that
Chapter 14 Improper Integral and Lebesgue Integral
3
Ng Tze Beng
. ------------------------- (1) n m N
1
u
a
bn
f (x)dx B <
c
3
Since , there exists an integer N
2
such that
a
an
f (x)dx d A
. ------------------------- (2) n m N
2
u
a
an
f (x)dx A <
c
3
By assumption, there exists a number N > 0 such that for all s and t > N,

s
t
f (x)dx <
c
3
Since an and bn . and there exists an integer M > 0 such that
n M an , bn > N.
Therefore, . --------------------------- (3) n m Mu
an
bn
f (x)dx <
c
3
Therefore, for n max(N
1
, N
2
, M),
B A = B
a
bn
f (x)dx +
an
bn
f (x)dx +
a
an
f (x)dx A
[ B
a
bn
f (x)dx +
an
bn
f (x)dx +
a
an
f (x)dx A
. <
c
3
+
c
3
+
c
3
= c
Since > 0 is arbitrary, B = A. This shows that is convergent. This
a

f (x)dx
completes the proof.
We have a similar result for the integral

b
f (x)dx
Theorem 3. Suppose f : ( , b] R is Riemann integrable on every closed
sub-interval of ( , b]. Then the improper integral exists if and only if

b
f (x)dx
for every > 0, there exists a number N < 0 such that for all s and t < N,

s
t
f (x)dx < c
The proof of Theorem 3 is similar to that of Theorem 2.
Theorem 2 and Theorem 3 then gives a convergence criterion for integrals of the type
.

+
f (x)dx
For non-negative function f : R R we have the following results concerning
convergence.
Theorem 4. If f : R R is Riemann integrable on every closed sub-interval of R
and is non-negative, then
.

+
f (x)dx =
td
lim
t
t
f (x)dx
Proof. Suppose is convergent. Then are

+
f (x)dx
td
lim
t
0
f (x)dx and
td
lim
0
t
f (x)dx
convergent. Therefore,

td
lim
t
t
f (x)dx =
td
lim
t
0
f (x)dx +
td
lim
0
t
f (x)dx
=
td
lim
t
0
f (x)dx +
td
lim
0
t
f (x)dx
. =

+
f (x)dx
Conversely, suppose is convergent. Then since f is non-negative,
td
lim
t
t
f (x)dx
is an increasing function on [0, ). In particular , g(t) =
0
t
f (x)dx
. g(t) =
0
t
f (x)dx [
t
t
f (x)dx [
td
lim
t
t
f (x)dx
Chapter 14 Improper Integral and Lebesgue Integral
4
Ng Tze Beng
Hence, g is bounded above and so by the
td
lim
0
t
f (x)dx = sup{g(t) : t c [0, )}
completeness of R. It can be shown similarly that is an increasing h(t) =
t
0
f (x)dx
function bounded above by . Thus
td
lim
t
t
f (x)dx
td
lim
t
0
f (x)dx = sup{h(t) : t c [0, )}.
This means is convergent. By the above discussion,

+
f (x)dx
. This completes the proof.

+
f (x)dx =
td
lim
t
t
f (x)dx
Suppose f : R R is Riemann integrable on every closed sub-interval of R. We can
define the principal value of to be . Note that though

+
f (x)dx
td
lim
t
t
f (x)dx
may be divergent its principal value may exist. For example, take f (x) =

+
f (x)dx
x. Then plainly, is divergent but its principal value is 0 as .

+
f (x)dx
t
t
f (x)dx = 0
Remark. Theorem 4 simply states that for a non-negative function f : R R, which
is Riemann integrable on every closed sub-interval of R, the improper integral
is equal to its principal value. The proof of Theorem 4 also shows that if

+
f (x)dx
we know that is convergent, then , the improper integral, is

+
f (x)dx

+
f (x)dx
equal to its principal value. Thus very often we just need to compute the principal
value in order to obtain the improper integral once convergence is ascertained.
Example 5.
(1) Let f (x) = 1/(1+ x
2
) . Then for any t > 0, .
0
t
f (x)dx = [tan
1
(x)]
0
t
= tan
1
(t)
Thus, . Also and so
0

f (x)dx =
td
lim tan
1
(t) =
v
2

t
t
f (x)dx = [tan
1
(x)]
t
t
= 2 tan
1
(t)
.

f (x)dx =
td
lim 2tan
1
(t) = v
(2)

xe
x
dx .
by integration by parts

0
t
xe
x
dx = [e
x
x]
0
t
+

0
t
e
x
dx
= [e
x
x e
x
]
0
t
= e
t
t e
t
+ 0 + 1 = 1
t
e
t

1
e
t
.
Therefore,
.

xe
x
dx =
td+
lim

0
t
xe
x
dx =
td+
lim [1
t
e
t

1
e
t
]
Now by L' Hpital's Rule. Thus
td+
lim
t
e
t
=
td+
lim
1
e
t
= 0

0

xe
x
dx = 1 0 0 = 1.
(3) is divergent since

0
+
xdx
td+
lim

0
t
xdx =
td+
lim
t
2
2
= +.
Definition 6. If is convergent, then we say is absolutely
a
+
f (x) dx
a
+
f (x)dx
convergent. Similarly, if is convergent, then we say is

b
f (x) dx

b
f (x)dx
absolutely convergent. Likewise, if is convergent, then we say

f (x) dx
is absolutely convergent.

f (x)dx
Theorem 7. If respectively, , is convergent,
a
+
f (x) dx

b
f (x) dx

f (x) dx
then respectively, , is convergent.
a
+
f (x)dx

b
f (x)dx

f (x)dx
Proof. Suppose is convergent. Then by Theorem 2, for every > 0,
a
+
f (x) dx
there exists a number N > 0 such that for all s and t > N,
Chapter 14 Improper Integral and Lebesgue Integral
5
Ng Tze Beng
.
s
t
f (x) dx < c
It follows that for all s and t > N, . Hence, by Theorem
s
t
f (x)dx [
s
t
f (x) dx < c
2, is convergent. For the case of and , the proof
a
+
f (x)dx

b
f (x) dx

f (x) dx
proceeds in the same way.
Definition 8. If respectively, , is convergent
a
+
f (x)dx

b
f (x)dx

f (x)dx
and respectively, , is divergent, then we say
a
+
f (x) dx

b
f (x) dx

f (x) dx
respectively, , is conditionally convergent.
a
+
f (x)dx

b
f (x)dx

f (x)dx
Example 9. Let . Then . We f (x) =





sin(x)
x
, if x ! 0
1, if x = 0

0

f (x)dx =
0
sin(x)
x
dx
can establish that is convergent as follows.
0
sin(x)
x
dx
For t > s > 0,

s
t sin(x)
x
dx =
cos(x)
x
s
t

s
t cos(x)
x
2
dx =
cos(s)
s

cos(t)
t

s
t cos(x)
x
2
dx
by integration by parts. Therefore,
. --------------------------- (1).
s
t sin(x)
x
dx [
1
s
+
1
t
+
s
t
1
x
2
dx =
2
s
For any > 0, there exists a number K > 0 such that 2/K< . It follows that for
t > s > K,
.
s
t sin(x)
x
dx [
2
s
<
2
K
< c
Therefore, by Theorem 2, is convergent. Actually .
0
sin(x)
x
dx
0
sin(x)
x
dx =
v
2
However, is divergent. To see this, note that
0
sin(x)
x
dx

(k1)v
kv
sin(x)
x
dx m
1
kv

(k1)v
kv
sin(x) dx =
2
kv
and so .
0
nv
sin(x)
x
dx m
2
v
E
k=1
n
1
k
Therefore, since is divergent, it follows that is divergent. Hence
E
k=1
n
1
k

0

sin(x)
x
dx
is conditionally convergent.
0
sin(x)
x
dx
For non-negative function f we have also the following comparison test for
convergence. Indeed we have used the following result in the proof of Theorem 4
implicitly.
Theorem 10. If exists and 0 f (x) k g(x) for x a and f :[a, ) R
a

g(x)dx
is Riemann integrable on every closed sub-interval of [a, ), then exists.
a

f (x)dx
Proof. By hypothesis, is an increasing function on [a, ) and h(t) =
a
t
f (x)dx
bounded above by . Let K = sup{ h(t) : t [a, )}. Then given any k
a

g (x)dx
> 0, there exists a t
0
in [a, )} such that K < h(t
0
) K. Now given any
Chapter 14 Improper Integral and Lebesgue Integral
6
Ng Tze Beng
sequence (an) in [a, ) such that an , there exists an integer N such that n N
an > t
0
. Therefore,
n N h(an) h(t
0
) K < h(t
0
) h(an) K | h(an) K| < .
This shows that for any sequence (an) in [a, ) such that an , h(an) K.
Therefore, by definition, . This means exists.
td
lim h(t) =
td
lim
a
t
f (x)dx = K
a

f (x)dx
Note that we have used the completeness property of R by using the supremum of
{h(t) : t [a, )}. The following is a technical observation which may come in
handy.
Lemma 11. If for every > 0, there exists a number R such that | f (x) f (y) | <
whenever x, y > R, then f (x) tends to a limit as x tends to .
Proof. Take any sequence (an) such that an . Then there exists an integer N such
that n N an > R. Thus | f (an) f (am) | < . This means ( f (an)) is a Cauchy
sequence. Consequently ( f (an)) is convergent. Suppose f (an) L. We shall show
that for any sequence (bn) such that bn , f (bn) L. By assumption, there exists a
number R
1
such that | f (x) f (y) | < /2 whenever x, y > R
1
. Plainly, there exists an
integer M such that n M an , bn > R
1
. Since f (an) L, there exists an integer M
1
such that
n M
1
| f (bn) L | < /2.
Therefore, n max (M ,
1
) | f (bn) L | | f (bn) f (an) + f (an) L |
| f (bn) f (an) | + | f (an) L < /2 + /2 = .
This shows that f (bn) L. Hence
xd
limf (x) = L.
Theorem 12. The following is equivalent to ,

f (x)dx = S.
Given any > 0, there exists a number R > 0 such that . u, v m R u
u
v
f S < c
Proof.
(i) Suppose , Then given any > 0 there exists a number R > 0 such

f (x)dx = S.
that
and that and . u, m R u
u
u
f S <
c
2

v
u
f <
c
2

v
u
f <
c
2
Therefore, for u, v R, we have that if v > u,

u
v
f S =
u
u
f S +
u
v
f [
u
u
f S +
u
v
f <
c
2
+
c
2
= c
and if v < u,
.
u
v
f =
u
v
f +
v
v
f S [
u
v
f +
v
v
f S <
c
2
+
c
2
= c
Hence, for u, v R, .
u
v
f S < c
(ii) Conversely, suppose for any > 0, there exists a number R > 0 such that
. u, v m R u
u
v
f S < c
Then for any x, y > R, .
x
y
f =
R
y
f S + S
R
x
f [
R
y
f S + S
R
x
f < 2c
Chapter 14 Improper Integral and Lebesgue Integral
7
Ng Tze Beng
It follows by Theorem 2 that is convergent. We deduce in the same way
0

f (x)dx
that is convergent. Hence is convergent. Let .

0
f (x)dx

f (x)dx

f (x)dx = S

Then by part (i) there exists u R such that . But .

u
u
f S

< c
u
u
f S < c
Hence
. S

S = S

u
u
f +
u
u
f S [
u
u
f S

+
u
u
f S < c + c = 2c
Since is arbitrary, S' = S.
We next have the following convergence criterion that we shall use later,
Theorem 13. Suppose there exist real numbers R, M > 0 such that for all |x| > R,
x
2
| f (x) | M and f is Riemann integrable on any closed interval in R. Then
is absolutely convergent.

f (x)dx
Proof. f is Riemann integrable on [0, R]. For x > R > 0,

| f (x) | M /x
2
.
Since is convergent, by Theorem 10, is convergent. It follows
R

1
x
2
dx
R

f (x) dx
that is absolutely convergent and therefore convergent (Theorem 7). We
R

f (x)dx
show similarly that is absolutely convergent. Therefore,

R
f (x)dx

0

f (x) dx =
0
R
f (x) dx +
R

f (x) dx
is convergent. Likewise is convergent.

0
f (x) dx =

R
f (x) dx +
R
0
f (x) dx
Therefore, is absolutely convergent.

f (x)dx
Remark. Similarly result holds for or .
a

f (x)dx

b
f (x)dx
Example 14. converges for all p.
1

e
x
x
p
dx
Note that as x . Therefore, there exists a number R > 0 x
2
e
x
x
p
=
x
p+2
e
x
d 0
such that for x > R. Since is continuous on [1, R] and so x
2
e
x
x
p
[ 1 e
x
x
p
=
x
p+2
e
x
is Riemann integrable on [1, R], it follows from Theorem 13 that . is
1

e
x
x
p
dx
convergent.
14.2 Improper Integrals on Bounded Domain, Part 1
Definition 15
(1) Suppose f is continuous on (a, b] and that , then
xd a
+
lim f (x) = !
if the limit exists.

a
b
f (x)dx =
td a
+
lim

a
b
f (x)dx ,
(2) Suppose f is continuous on [a, b) and that , then
xd b

lim f (x) = !
if the limit exists.

a
b
f (x)dx =
tdb

lim

a
t
f (x)dx ,
(3) Suppose f is continuous on where a < c < b and that [a, b] {c} ,
Then if the limits
xdc
lim f (x) = + .

a
b
f (x)dx =
tdc

lim

a
c
f (x)dx +
sdc
+
lim

s
b
f (x)dx ,
exist.
Chapter 14 Improper Integral and Lebesgue Integral
8
Ng Tze Beng
Example 16.
(1) . Using integration by parts,

0
1
x ln(x)dx =
td0
+
lim

t
1
x ln(x)dx

t
1
xln(x)dx = [
x
2
2
ln(x)]
t
1

t
1 x
2
2
$
1
x
dx = [
x
2
2
ln(x)]
t
1

t
1 x
2
dx
. = [
x
2
2
ln(x)
1
4
x
2
]
t
1
=
1
4

t
2
2
ln(t) +
1
4
t
2

But
td0
+
lim t
2
ln(t) =
td0
+
lim
ln(t)
1
t
2
=
td0
+
lim
1
t

2
t
3
=
td0
+
lim
t
2
2
= 0 .
Thus, .

t
1
xln(x)dx =
1
4

1
2
t
2
ln(t) +
1
4
t
2
d
1
4
as t d0
+
Therefore

0
1
xln(x)dx =
1
4
.
(2)

1
1 1
x
1
3
dx

0
1 1
x
1
3
dx =
td0
+
lim

t
1 1
x
1
3
dx =
td0
+
lim [
3
2
x
2
3
]
t
1
=
td0
+
lim
3
2

3
2
t
2
3
=
3
2
.
Similarly,

1
0 1
x
1
3
dx =
td0

lim

1
t 1
x
1
3
dx =
td0

lim [
3
2
x
2
3
]
1
t
=
td0

lim
3
2
t
2
3
1 =
3
2
.
Therefore,

1
1 1
x
1
3
dx =
td0
+
lim

t
1 1
x
1
3
dx
td0

lim

1
t 1
x
1
3
dx=
3
2
+
3
2
= 3 .
(3)

0
2 1
(x1)
2
dx =
td1

lim

0
t 1
(x1)
2
dx +
td1
+
lim

t
2 1
(x1)
2
dx
=
td1

lim [
1
x1
]
0
t
+
td1
+
lim [
1
x1
]
t
2
.
But both limits do not exist, therefore the integral is divergent.

0
2 1
(x1)
2
dx
In a later section we shall consider convergence criterion and relation with Lebesgue
integral.
14.3 Lebesgue Measure and Lebesgue Integral
In this section we give a review of Lebesgue theory.
Definition 17. For any function f , define the following associated functions:
f
+
(x) = ( | f (x) | + f (x) )/2 0,
f

(x) = ( | f (x) | f (x) )/2 0.

Then and f
+
(x) =



f (x) if f (x) m 0
0 if f (x) < 0
f

(x) =



f (x) if f (x) [ 0
0 if f (x) > 0
In particular, f (x) = f
+
(x) f

(x) and | f (x) | = f

+
(x) + f

(x).
Here we shall present the definition of measurable function and the definition of
(Lebesgue) measure on R. | f | is measurable need not imply that f is measurable.
For instance, if we assume the existence of non-measurable set E, then if we define f
(x) = 1 for x in E and f (x) = 1 for x not in E, then | f | is measurable but f is not. In
view of this , we shall always assume or state the requirement that f be measurable.
Lebesgue Measure.
Chapter 14 Improper Integral and Lebesgue Integral
9
Ng Tze Beng
Suppose I is an interval, define (I) to be (b a) if I is bounded and a and b are end
points of I , with a < b. Define (I) to be if I is unbounded.
Let be the family of all countable collections of disjoint open intervals. Define a
function on into the extended non-negative real numbers
z
&
: I dR
+
= [0, ]
by . z
&
(,) =
E
Ic,
z(I)
Suppose E is a subset of R. Define the family of countable cover of E by disjoint
open intervals to be C (E) = { : is a countable collection of disjoint open intervals
covering E}.
Define the Lebesgue outer measure of E by
*(E) = inf {*() : C (E) }.
Thus *(E) is either finite or .
Definition 18. A subset E of R is said to be Lebesgue measurable if and only if for
all subset X of R,
*(X) = *(XE) + *(X \ E),
equivalently if for all X R,
*(X) *(XE) + *(X \ E).
If E is measurable, the Lebesgue measure of E, (E) is defined to be the outer
measure *(E).
Properties 19.
1. and R are measurable,
2. If E is measurable, then its complement R \ E is also measurable.
3. Any open subset of R is measurable and so any closed subset of R is also
measurable. Hence any interval is measurable.
4. Countable union of measurable subsets is measurable and so countable intersection
of measurable subsets is also measurable.
5. If *(E) = 0, then E and any subset of E is also measurable.
6. If E
1
, E
2
, is a countable collection of disjoint measurable sets, then
. u(4{E
n
: n = 1, 2, }) =
E
n=1

u(E
n
)
Measurable Function
Definition 20.
A function f : E R is measurable if and only if for any open subset U of R, f
-1
(U)
is measurable.
Hence it is necessary that E be measurable for f to be measurable.
The following is an immediate consequence of the definition.
Properties 21.
Chapter 14 Improper Integral and Lebesgue Integral
10
Ng Tze Beng
(1) Suppose E is measurable. Then every continuous function f : E R is
measurable.
(2) Suppose f : E R is measurable and f (E) X and g : X R is continuous.
Then the composite function g ) f : E R is measurable.
(3) Suppose f : E R and g : E R are measurable, then f + g and f g are
measurable.
Lemma 22. If f : E R is measurable, then | f |, f
+
and f

are measurable.
Proof. If f is measurable, then f $ f is measurable by Property 21 (3) and by Property
(2), | f | = ( f $ f ) is measurable. Thus by Property 21 (3), f
+
= ( | f | + f )/2 is
measurable. Likewise, f

= ( | f | f )/2 is measurable.
Simple Function and Lebesgue Integral
Definition 23.
A function s : E R is said to be simple if s is measurable and s takes on only a
finite number of values.
For any subset A of R, the characteristic function of A, A is the function defined by
. x
A
(x) =



1 if x c A
0 if x " A
Thus if s : E R is simple, s(E) = {
1
,
2
, , n } and Ai = s
-1
(i), i =1, , n, then
. s =
E
i=1
n
c
i
,
A
i
Note that each Ai is measurable since s is measurable.
For a non-negative simple function s : E R
+
= [0, ), we define the Lebesgue
integral of s to be
.
E
sdu =
E
i=1
n
c
i
u(A
i
)
where , s(E) = {
1
,
2
, , n } and Ai = s
-1
(i), i =1, , n. Hence s =
E
i=1
n
c
i
,
A
i
is either finite or infinite.
E
sdu
If the Lebesgue measure of E, (E) is finite, we can define for a simple function
s : E R (not necessarily non-negative) , the Lebesgue integral of s to be
,
E
sdu =
E
i=1
n
c
i
u(A
i
)
where , s(E) = {
1
,
2
, , n } and Ai = s
-1
(i), i =1, , n. s =
E
i=1
n
c
i
,
A
i
Note that we have then is finite since each (Ai) is finite.
E
sdu
We say the simple function s is Lebesgue integrable if its Lebesgue integral is finite.
Thus a simple function defined on a set E of finite measure is Lebesgue integrable.
Definition 24.
Suppose (E) is finite, i.e., (E) < . Suppose f : E R is a bounded function.
Chapter 14 Improper Integral and Lebesgue Integral
11
Ng Tze Beng
We say f is Lebesgue integrable if and only if given > 0 there exists simple
functions u and v on E such that
u f v
and . Following Darboux we define the Lebesgue integral
E
(v u)du < c
.
E
f du = sup{
E
du : is a simple function on E such that [ f }
Notice the Lebesgue integral of u and v are analogous to the lower Darboux and
upper Darboux sums for a bounded function defined by step functions defined for a
partition of a finite interval.
Theorem 25.
Suppose (E) is finite, i.e., (E) < . Suppose f : E R is a bounded function.
Then f is Lebesgue integrable if and only if f is measurable if and only if | f | is
Lebesgue integrable. In particular, when f is Lebesgue integrable, f
+
and f

are
Lebesgue integrable and .
E
f du =
E
f
+
du
E
f

du
Proof. Suppose (E) is finite, i.e., (E) < . Suppose f : E R is a bounded
function. We can define the lower Lebesgue integral to be
L
E
f du = sup{
E
du : is a simple function on E such that [ f }
and the upper Lebesgue integral to be
. U
E
f du = inf{
E
du : is a simple function on E such that f [ }
The condition in Definition 24 is equivalent to that the lower and upper Lebesgue
integrals are the same. It can be shown that f is measurable if and only if the lower
and upper Lebesgue integrals are the same. We omit the details here as it is a little
tedious. (For a reference see Proposition 3, Chapter 4, Royden's "Real Analysis",
Third Edition.)
Note that f = f
+
f

. Since f is measurable, f
+
and f

are measurable and so

their lower and upper Lebesgue integrals are the same. It follows that
.
E
f du =
E
f
+
du
E
f

du
Now we consider unbounded functions.
Definition 26. Lebesgue integral of non-negative function.
Suppose f : E R
+
is a non-negative function and E is measurable. Define the
Lebesgue integral of f to be
.
E
f du = sup{
E
s du : s is a non-negative simple function on E such that 0[ s [ f }
We say f is Lebesgue integrable (or summable) if the integral is finite.
E
f du
The following is well known.
Lemma 27. Suppose f : E R
+
is a non-negative measurable function. Then there
exists an increasing sequence of non-negative measurable simple functions {sn : E
R
+
} such that s
1
s
2
f and ( sn ) converges pointwise to f .
Chapter 14 Improper Integral and Lebesgue Integral
12
Ng Tze Beng
Theorem 28. Lebesgue Monotone Convergence Theorem.
Suppose { f n : E R
+
} is a sequence of non-negative increasing measurable
functions tending pointwise to f : E R
+
. Then f is measurable and
.
E
f
n
du ~
E
f du

In view of Theorem 28 we have
Corollary 29. Suppose f : E R
+
is a non-negative measurable function. Then
there exists an increasing sequence of non-negative measurable simple functions {sn :
E R
+
} such that s n f , ( sn
) converges pointwise to f and .
E
s
n
du ~
E
f du
In some sense Theorem 25 motivates the next definition.
Definition 30. Lebesgue integral of a real valued functions.
Suppose f : E R is a measurable function. We say f is Lebesgue integrable if and
only if . Note that then we have .
E
| f |du <
E
f
+
du,
E
f

du [
E
| f |du <
We define the Lebesgue integral of f to be .
E
f du =
E
f
+
du
E
f

du
Theorem 31. Lebesgue Dominated Convergence Theorem.
Suppose { f n : E R } is a sequence of measurable functions tending pointwise
(almost everywhere) to f : E R . Suppose further that there exists a Lebesgue
integrable function g : E R such that | f n | g for all n in P and for almost all x in E
(i.e., | f n | g for all n in P except perhaps on a set of measure zero). Then f n and f
are Lebesgue integrable and .
E
f
n
du d
E
f du
Remark. A reference for Lebesgue Monotone Convergence Theorem and Lebesgue
Dominated Convergence Theorem is Royden's "Real Analysis, Pearson Prentice Hall
Third Edition", Page 87 Theorem 10, page 91 Theorem 16. Lemma 27 is well known
and is often stated without proof (see for example, Proposition 7 in Chapter 11 page
260 of Royden's "Real Analysis".) One can produce the required simple functions by
an appropriate dissection of the subinterval [0, n] in [0, ) for each n and use the
pre-image of the subintervals of the dissection to define the simple function for each
n.
For bounded function defined on a bounded interval the following is well known.
Theorem 32. Lebesgue.
Suppose f :[a, b] R is a bounded function on a closed and bounded interval [a, b].

If f is Riemann integrable, then f is Lebesgue integrable, hence measurable and the
Lebesgue integral.
,
[a,b]
f du =
a
b
f (x)dx
where the right integral is the Riemann integral of f on [a, b].
The following is a characterization of Riemann integrable functions in terms of
continuity.
Chapter 14 Improper Integral and Lebesgue Integral
13
Ng Tze Beng
Theorem 33. Lebesgue.
Suppose f :[a, b] R is a bounded function on a closed and bounded interval [a, b].
f is Riemann integrable if and only if f is continuous except perhaps on a set of
measure zero, i.e., f is continuous almost everywhere.
14. 4 Improper Integral and Lebesgue integral.
Theorem 34. Suppose f :[a, ) R
+
is a non-negative function which is Riemann
integrable on every closed sub-interval of [a, ). Then f is measurable. f is
Lebesgue integrable if and only if the improper (Riemann) integral
a
+
f (x)dx
exists. When f is Lebesgue integrable, the Lebesgue integral
.
[a,)
f du =
a
+
f (x)dx
Proof.
Since f is Riemann integrable on every sub interval of [a, ), for each integer n in
P, define f n :[a, ) R
+
by . Then ( f n ) is a f
n
(x) =



f (x), a [ x [ a + n
0, x > a + n
monotone increasing sequence converging pointwise to f. Each f n is Riemann
integrable on [a, a + n] and 0 on (a+n, ). Therefore, f n restricted to [a, a + n] is
measurable by Theorem 32 and so for any open U not containing 0,
is measurable. For any open U containing 0, f
n
1
(U) =
(
f
n
x
[a,a+n]
)
1
(U)
which is a union of measurable sets and so f
n
1
(U) =
(
f
n
x
[a,a+n]
)
1
(U) 4(a + n, )
is measurable. This shows that f n is measurable. By definition 26 and the fact that
[a,) is a disjoint union of [a,a+n] and (a+n, ) the Lebesgue integral

[a,)
f
n
du =
[a,a+n]
f
n
du +
[a+n,)
f
n
du =
[a,a+n]
f
n
du + 0 =
a
a+n
f (x)dx
since the Lebesgue integral by Theorem 32.
[a,a+n]
f
n
du =
a
a+n
f (x)dx
By the Lebesgue Monotone Convergence Theorem (Theorem 28), f is measurable
and
.
[a,)
f
n
du d
[a,)
f du
That is, .
[a,)
f du =
nd
lim
a
a+n
f (x)dx =
td
lim
a
t
f (x)dx
f is Lebesgue integrable if and only if is finite if and only if the sequence
[a,)
f du
has a limit if and only if the improper integral
[a,)
f
n
du =
a
a+n
f (x)dx
exists.
a
+
f (x)dx
Thus when f is Lebesgue integrable, .
[a,)
f du =
a
+
f (x)dx
Similarly we can prove the following analogous results.
Theorem 35. Suppose f : ( , a] R
+
is a non-negative function which is
Riemann integrable on every closed sub-interval of ( , a] Then f is measurable.
f is Lebesgue integrable if and only if the improper (Riemann) integral

a
f (x)dx
exists. When f is Lebesgue integrable, the Lebesgue integral
.
(,a]
f du =

a
f (x)dx (=
td
lim
t
a
f (x)dx)

Chapter 14 Improper Integral and Lebesgue Integral
14
Ng Tze Beng
Theorem 36. Suppose f : ( , ) R
+
is a non-negative function which is
Riemann integrable on every closed sub-interval of ( , ) Then f is measurable.
f is Lebesgue integrable if and only if the improper (Riemann) integral

f (x)dx
exists. When f is Lebesgue integrable, the Lebesgue integral
.
(,)
f du =

f (x)dx(=
td
lim
t
t
f (x)dx)

Remark. The proof of Theorem 36 is the same as for Theorem 34 by defining f n to
be equal to f on [n, n] and 0 elsewhere and making use of the fact that the improper
(Riemann) integral

+
f (x)dx =
td
lim
t
t
f (x)dx
from Theorem 4.
Theorem 37. Suppose f :[a, ) R is Riemann integrable on every closed
sub-interval of [a, ). Then f is Lebesgue integrable if and only if the improper
(Riemann) integral exists. When f is Lebesgue integrable, the
a
+
f (x) dx
Lebesgue integral
,
[a,)
f du =
a
+
f (x)dx
where the integral on the right is the improper Riemann integral.
Proof. By Definition 30, f is Lebesgue integrable if and only if | f | is Lebesgue
integrable if and only if the improper integral exists by Theorem 34. If f
a
+
f (x) dx
is Lebesgue integrable,
.
[a,)
f du =
[a,)
f
+
du
[a,)
f

du
Note that | f | is Lebesgue integrable implies both f
+
and f

are Lebesgue integrable.

By Theorem 34, and .
[a,)
f
+
du =
a
+
f
+
(x)dx
[a,)
f

du =
a
+
f

(x)dx
Therefore, when f is Lebesgue integrable,

[a,)
f du =
[a,)
f
+
du
[a,)
f

du
=
a

f
+
(x)dx
a

f

(x)dx
, =
a

(
f
+
(x) f

(x)
)
dx =
a

f (x)dx
where all the improper integrals are convergent.
The following result is similar to Theorem 37 and is proved in exactly the same way.
Theorem 38. Suppose f :(, a] R is Riemann integrable on every closed
sub-interval of (, a]. Then f is Lebesgue integrable if and only if the improper
(Riemann) integral exists. Moreover when f is Lebesgue integrable,

a
f (x) dx
the Lebesgue integral
,
(,a]
f du =

a
f (x)dx
where the integral on the right is the improper Riemann integral.
In view of Theorem 37 we can phrase absolute convergence of improper integral with
Lebesgue integrability.
Chapter 14 Improper Integral and Lebesgue Integral
15
Ng Tze Beng
Definition 39. In view of Definition 6 of absolute convergence, if f :[a, ) R is
Riemann integrable on every closed sub-interval of [a, ), then is
a
+
f (x)dx
absolutely convergent if and only if is convergent if and only if f is
a
+
f (x) dx
Lebesgue integrable on [a, ). If is conditionally convergent (see
a
+
f (x)dx
Definition 8), then this means is convergent but is divergent.
a
+
f (x)dx
a
+
f (x) dx
Hence conditional convergent implies non Lebesgue integrability.
Similarly, if f : (, a] R is Riemann integrable on every closed sub-interval of
(, a], then is absolutely convergent if and only if is

a
f (x)dx

a
f (x) dx
convergent if and only if f is Lebesgue integrable on (, a]. is

a
f (x)dx
conditionally convergent (see Definition 8), means that exists but

a
f (x)dx
is divergent or equivalently f is not Lebesgue integrable but the

a
f (x) dx
improper integral is convergent.
Recall from Definition 8 that for f : (, ) R which is Riemann integrable on
every closed sub-interval of (, ), the improper integral converges

f (x)dx
absolutely if and only if for some a in R (hence any a) the improper integrals
and converge absolutely if and only if f is Lebesgue integrable
a

f (x)dx

a
f (x)dx
on (, a] and on [a, ) if and only if f is Lebesgue integrable. Hence absolute
convergence of f is equivalent to f being Lebesgue integrable.
converges conditionally if either one of the improper integrals or

f (x)dx
a

f (x)dx
converges conditionally which implies that f is not Lebesgue integrable

a
f (x)dx
on R.
Example 40.
Let . Then . We have f (x) =





sin(x)
x
, if x ! 0
1, if x = 0

0

f (x)dx =
0
sin(x)
x
dx =
v
2
shown in Example 9 that f is conditionally convergent hence f is not Lebesgue
integrable on [0, ).
14.5 Improper Integrals on Bounded Domain, Part 2
If f :[a, b] R is unbounded, then f is not Riemann integrable. We shall consider
the case when f has only one singularity at one of the end points of [a, b].
Definition 41. Suppose f :[a, b] R is such that f is Riemann integrable on any
subinterval in (a, b].
(1) Suppose there is a sequence of points (xn ) in [a, b] such that | f (xn) | as xn
a. Then if the limit

tda
+
lim
t
b
f (x)dx
exists, this is defined to be the improper Riemann integral of f on [a, b].
(2) Suppose there is a sequence of points ( yn ) in [a, b] such that | f (yn) | as yn
b. Then if the limit

tdb

lim
a
t
f (x)dx
Chapter 14 Improper Integral and Lebesgue Integral
16
Ng Tze Beng
exists, this is defined to be the improper Riemann integral of f on [a, b].
(3) Suppose there is a sequence of points (xn ) in [a, b] such that | f (xn) | as xn
a and a sequence of points ( y
n
) in [a, b] such that | f (yn) | as yn b. If for some
c in (a, b) the improper integrals, and exist, then the improper
a
c
f (x)dx
c
b
f (x)dx
integral
.
a
b
f (x)dx =
a
c
f (x)dx +
c
b
f (x)dx
We have a convergence criterion similar to Theorem 2.
Theorem 42. Suppose f :[a, b] R
+
is such that f is Riemann integrable on any
subinterval in (a, b]. Then the improper integral exists if and only if for
a
b
f (x)dx
every > 0, there exists a number k > a (depending on ) such that for all a < s < t <
k,
.
s
t
f (x)dx < c
The proof is similar to Theorem 2.
We now consider non-negative function f : [a, b] R
+
, Suppose f is measurable.
What is the relation between Lebesgue integral on [a, b] and improper Cauchy
Riemann integral? It will be very similar to the case of unbounded domain.
Theorem 43. Suppose f :[a, b] R
+
is such that f is Riemann integrable on any
subinterval in (a, b). Suppose f is either unbounded at a or at b or even at both
points. Then f is Lebesgue integrable on [a, b] if and only if the limit of the
sequence of Riemann integrals , where f n = min( f , n), exists.
a
b
f
n
(x)dx
Proof. For each integer n in P, f n :[a, b] R
+
is defined by
. f
n
(x) =



f (x), if f (x) [ n
n, if f (x) > n
where f n = min( f , n). Take a decreasing sequence (xn) in (a, (a+b)/a ) such that xn
a and an increasing sequence (yn) in ((a+b)/a, b) such that and yn b. Then f is
Riemann integrable on [xn , yn] and so is continuous except on a set An of f |
[xn ,yn]
measure zero. Therefore, f is continuous except perhaps on the set E = {An : n in
P} whose measure is also zero since each An has measure zero. Since f n = min( f , n)
= ( f + n | f n|)/2 , f n is continuous except perhaps on a set of measure zero.
Since each f n is plainly bounded, f n is Riemann integrable and so is measurable and
Lebesgue integrable by Theorem 32. Plainly ( f n ) is an increasing sequence and
so is a sequence of increasing non-negative measurable functions converging
pointwise to f . Therefore, by the Lebesgue Monotone Convergence Theorem
(Theorem 28), f is measurable and the Lebesgue integrals,
.
[a,b]
f
n
du ~
[a,b]
f du
Since is equal to the Riemann integral ,
[a,b]
f
n
du
a
b
f
n
(x)dx
Chapter 14 Improper Integral and Lebesgue Integral
17
Ng Tze Beng
.
a
b
f
n
(x)dx ~
[a,b]
f du
Thus f is Lebesgue integrable if and only if the limit exists. This
nd
lim
a
b
f
n
(x)dx
completes the proof.
Now we shall apply the above proof to a different sequence of functions.
Theorem 44. Suppose f :[a, b] R
+
is such that f is Riemann integrable on any
subinterval in (a, b]. Suppose there is a strictly decreasing sequence (xn) in (a, b)
such that xn a but f (xn ) , that is, f is unbounded in any neighbourhood
containing a. Then f is Lebesgue integrable on [a, b] if and only if the improper
Cauchy Riemann integral exists.
a
b
f (x)dx
Proof. For each integer n in P, define f n :[a, b] R
+
by
. f
n
(x) =





f (x), if x m x
n
0, if a < x < x
n
f (a) if x = a
Then since f is Riemann integrable on [xn ,b], f n is Riemann integrable for each n in
P. Therefore, by Theorem 32, f n is measurable and Lebesgue integrable for each n
in P. Obviously, ( f n ) is an increasing sequence of non-negative functions
converging pointwise to f . Therefore, by the Lebesgue Monotone Convergence
Theorem, f is measurable and
.
[a,b]
f
n
du ~
[a,b]
f du
Now the Lebesgue integral . Hence
[a,b]
f
n
du =
a
b
f
n
(x)dx =
xn
b
f
n
(x)dx
.
xn
b
f
n
(x)dx ~
[a,b]
f du
If f is Lebesgue Integrable, then the limit exists for any sequence
nd
lim
xn
b
f
n
(x)dx
(xn) in (a, b) such that xn a
+
. That is to say, the improper Cauchy Riemann
integral exists. On the other hand, if the improper Riemann integral exists,
a
b
f (x)dx
then for any sequence xn a , and so and f is
xn
b
f
n
(x)dx ~
[a,b]
f du
[a,b]
f du <
Lebesgue integrable.
Theorem 45. Suppose f :[a, b] R is such that f is Riemann integrable on any
subinterval in (a, b]. Suppose there is a strictly decreasing sequence (xn) in (a, b)
such that xn a but | f (xn ) | , that is, | f | is unbounded in any neighbourhood
containing a. Then f is Lebesgue integrable on [a, b] if and only if the improper
Cauchy Riemann integral exists. Moreover, when f is Lebesgue
a
b
| f (x)|dx
integrable, the Lebesgue integral is equal to the improper Cauchy Riemann
[a,b]
f du
integral, .
a
b
f x)dx =
tda
+
lim
t
b
f (x)dx
Proof. By Definition 30, f is Lebesgue integrable if and only if | f | is Lebesgue
integrable if and only if both f
+
and f

are Lebesgue integrable. By Theorem 44,
the Lebesgue integral = the improper Cauchy Riemann
[a, b]
f
+
du =
tda
+
lim
t
b
f
+
(x)dx
integral . Similarly the Lebesgue integral
a
b
f
+
(x)dx
. In particular, if the improper Riemann
[a, b]
f

du =
tda
+
lim
t
b
f

(x)dx =
a
b
f

(x)dx
Chapter 14 Improper Integral and Lebesgue Integral
18
Ng Tze Beng
integral exists, then the improper integrals and
a
b
f (x) dx
a
b
f
+
(x)dx
a
b
f

(x)dx
exist (by the Comparison Test) and the Lebesgue integral

[a, b]
f du =
[a, b]
f
+
du +
[a, b]
f
+
du
=
a
b
f
+
(x)dx +
a
b
f

(x)dx =
a
b
f (x) dx
the sum of the improper Cauchy Riemann integrals. Moreover the Lebesgue integral
too exists and
[a, b]
f du
.
[a, b]
f du =
[a, b]
f
+
du
[a, b]
f
+
du =
a
b
f
+
(x)dx
a
b
f

(x)dx
Conversely, suppose that f is Lebesgue integrable, then | f | is Lebesgue integrable
and so both f
+
and f

are Lebesgue integrable. Therefore, by Theorem 44, both
improper Cauchy Riemann integrals and exist and are finite.
a
b
f
+
(x)dx
a
b
f

(x)dx
It follows that

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

(x)dx
exists. Moreover,
a
b
f (x)dx =
a
b
f
+
(x)dx
a
b
f

(x)dx
. =
[a, b]
f
+
du
[a, b]
f

du =
[a, b]
f du
We have a similar result when the function is unbounded in any neighbourhood
containing b.
Theorem 46. Suppose f :[a, b] R is such that f is Riemann integrable on any
subinterval in [a, b). Suppose there is a strictly increasing sequence (xn) in (a, b)
such that xn b but | f ( xn ) | , that is, | f | is unbounded in any neighbourhood
containing a. Then f is Lebesgue integrable on [a, b] if and only if the improper
Cauchy Riemann integral exists. Moreover, when f is Lebesgue
a
b
| f (x)|dx
integrable, the Lebesgue integral is equal to the improper Cauchy Riemann
[a,b]
f du
integral, .
a
b
f (x)dx =
tdb

lim
a
t
f (x)dx
The proof of Theorem 46 is exactly the same as for Theorem 45.
Definition 47.
Suppose one of the following is satisfied.
(1) There is a sequence of points (xn ) in [a, b] such that | f (xn) | as xn a.
(2) There is a sequence of points (yn ) in [a, b] such that | f (yn) | as yn b,
(3) There is a sequence of points (xn ) in [a, b] such that | f (xn) | as xn a and a
sequence of points ( y
n
) in [a, b] such that | f (yn) | as yn b.
The improper Cauchy Riemann integral converges absolutely if
a
b
f (x)dx
a
b
f (x) dx
exists (or converges) or equivalently if both and exist or in
a
b
f
+
(x)dx
a
b
f

(x)dx
view of Theorems 45 and 46 if f is Lebesgue integrable on [a, b].
The improper Cauchy Riemann integral converges conditionally if
a
b
f (x)dx
converges and diverges (does not exist) or equivalently f is not
a
b
f (x)dx
a
b
f (x) dx
Lebesgue integrable on [a, b].
Chapter 14 Improper Integral and Lebesgue Integral
19
Ng Tze Beng
If for a point b, there is a sequence of points (xn ) such that | f (xn) | as xn b, i.e.,
f is not bounded in any neighbourhood containing b, then we say f has a singularity
at b. Thus condition (1) is the same as saying that f has a singularity at a and
condition (2) a singularity at b while condition (3) a singularity at both a and b.
14.6 Convergence Tests
Theorem 48. Cauchy Criterion.
Suppose f :[a, b] R is such that f is Riemann integrable on any subinterval in [a,
b). Then the improper Cauchy Riemann integral exists if and only if for
a
b
f (x)dx
every > 0, there exists a number M < b (depending on ) such that for all M < s < t
< b,
.
s
t
f (x)dx < c
The proof is similar to Theorem 2.
Theorem 49.
Let b be a finite point or and f a non-negative function on [a, b). Suppose f is
Riemann integrable on [a, c] for any c, a < c < b with a singularity at b.
Then the improper Cauchy Riemann integral exists if and only if the
a
b
f (x)dx
function defined by is bounded on [a, b). F(x) =
a
x
f (t)dt
Proof. Plainly since F(x) is non-decreasing, exists if and only if F(x) is
xdb

Lim F(x)
bounded by the completeness property of R, i.e., the improper Cauchy Riemann
integral is convergent if and only if F(x) is bounded.
a
b
f (x)dx
Theorem 50. Comparison Test.
Suppose f and g are non-negative functions defined on I = [a, b), integrable on [a,
c] for any c, a < c < b . Suppose f and g each has a singularity at b and that f (x)
g(x) for all x in I.
Then
(1) If the improper Cauchy Riemann integral converges, then so does
a
b
g (x)dx
.
a
b
f (x)dx
(2) If the improper Cauchy Riemann integral diverges, then so does
a
b
f (x)dx
.
a
b
g (x)dx
Proof. This a consequence of Theorem 49. If converges, then
a
b
g (x)dx
a
x
g (t)dt
is bounded on I . Since , is bounded consequently
a
x
f (t)dt [
a
x
g (t)dt
a
x
f (t)dt
is convergent.
a
b
f (x)dx
diverges implies that is unbounded and so is unbounded
a
b
f (x)dx
a
x
f (t)dt
a
x
g (t)dt
and diverges.
a
b
g (x)dx
Theorem 51. Limit Comparison Test
Chapter 14 Improper Integral and Lebesgue Integral
20
Ng Tze Beng
Suppose f and g are positive functions defined on I = [a, b), integrable on [a, c] for
any c, a < c < b . Suppose that . 0 <
xdb

lim
f (x)
g(x)
<
Then either both improper Cauchy Riemann integrals and
a
b
f (x)dx
a
b
g (x)dx
converge or they both diverge.
Proof. Suppose . Then taking = q/2 > 0, there exists > 0 such
xdb

lim
f (x)
g(x)
= q > 0
that
. b ^ < x < b u
f (x)
g(x)
q < c =
q
2
Therefore, taking X = b , for all x such that X < x < b, we have
,
q
2
<
f (x)
g(x)
<
3q
2
i.e., . -------------------------- (1)
q
2
g(x) < f (x) <
3q
2
g(x)
By hypothesis and converge if and only if and
a
b
f (x)dx
a
b
g (x)dx
X
b
f (x)dx
converge. It follows from (1) and Theorem 50 that either both
X
b
g (x)dx
X
b
f (x)dx
and converge or they both diverge. Consequently, either both
X
b
g (x)dx
a
b
f (x)dx
and converge or they both diverge.
a
b
g (x)dx
14.7 Anti-derivative and Improper Integral
We can make use of primitive or more specifically limit of primitive to determine
Lebesgue integral. For example Theorem 45 tells us that the Lebesgue integral is an
improper Cauchy Riemann integral. Then we can make use of the Fundamental
Theorem of Calculus to calculate the improper integral and hence the Lebesgue
integral.
Theorem 52.
(1) Suppose f : [a, ) R is continuous and non-negative on [a, ). Suppose F
is an antiderivative of f . Then f is Lebesgue integrable on [a, ) if and only if
the limit exists (and is finite).
xd
limF(x)
In this case, the Lebesgue integral
.
[a,)
f du =
a

f (x)dx =
xd
limF(x) F(a)
(2) Suppose f : (a, b] R is continuous and non-negative on (a, b] and has a
singularity at a. Suppose F is an antiderivative of f . Then f is Lebesgue integrable
on (a, b] if and only if the limit exists (and is finite).
xda
+
lim F(x)
In this case, the Lebesgue integral
.
(a,b])
f du =
a
b
f (x)dx = F(b)
xda
+
lim F(x)
Proof. Part (1) is a consequence of Theorem 34 and the Fundamental Theorem of
Calculus, as . Part (2) is a consequence of
a

f (x)dx =
xd
lim
a
x
f (t)dt =
xd
lim F(x) F(a)
Theorem 44 and the Fundamental Theorem of Calculus.
Example 53.
Chapter 14 Improper Integral and Lebesgue Integral
21
Ng Tze Beng
For x > 0 and k > 0, an anti-derivative for is for k 1 and ln(x) x
k
=
1
x
k
1
k + 1
x
k+1
for k = 1.
Therefore, for a > 0 ,
.
a
x
1
t
k
dt =





1
k + 1
x
k+1

1
k + 1
a
k+1
, for k > 0 and k ! 1
ln(x) ln(a), for k = 1
Now (-k +1) < 0 if and only if k >1 and so converges when k > 1, since
a

1
t
k
dt
Thus for k >1 and
1
k + 1
x
k+1
d 0 as x d .
a

1
t
k
dt =
1
k 1
a
k+1
=
1
k 1
1
a
k1
diverges when 0 < k < 1. Likewise, diverges because ln(x) as
a

1
t
k
dt
a

1
t
dt
x. Plainly diverges when k 0.
a

1
t
k
dt
Thus we have the following summary
(1) for a > 0 converges if and only if k > 1.
a

1
x
k
dx
We can similarly deduce the following
(2) for a > 0 converges if and only if k < 1.
0
a
1
x
k
dx
It then follows that
(3) for a > x
0
converges if and only if k > 1
a

1
(x x
0
)
k
dx
and
(4) for a > x
0
converges if and only if k < 1.
x
0
a
1
(x x
0
)
k
dx
Example 54.
(1) converges for all p.
1

e
x
x
p
dx
Use a simple comparison test.
. e
x
x
p+2
d0 as x d
Therefore, there exist a number M such that x M . e
x
x
p+2
< 1 ue
x
x
p
<
1
x
2
Since is convergent, by the Comparison Test (Theorem 50) is
1

1
x
2
dx
1

e
x
x
p
dx
convergent.
Hence the Lebesgue integral .
[1,)
e
x
x
p
du =
1

e
x
x
p
dx
(2) , a > 0 diverges for all p.
0
a
e
1
x
x
p
dx
Note that because for n > p+1 and e
1
x
x
p+1
d as x d 0
+
e
1
x
x
p+1
>
1
n!
1
x
n
x
p+1
. Therefore, there exists > 0 such that
xd0
+
lim
1
n!
1
x
n
x
p+1
=
. 0 < x < c ue
1
x
x
p+1
> 1 ue
1
x
x
p
>
1
x
Chapter 14 Improper Integral and Lebesgue Integral
22
Ng Tze Beng
Since is divergent, consequently by the Comparison Test (Theorem 50),
0
c
1
x
dx
is divergent for any p. It follows that is divergent for all p.
0
c
e
1
x
x
p
dx
0
a
e
1
x
x
p
dx
Therefore, by Theorem 44 is not Lebesgue integrable on [0, a]. e
1
x
x
p
(3) converges for any a > 0.
0
a
ln(x)dx

For any 0 < t a, by integration by parts,

t
a
ln(x)dx =
[
x ln(x)
]
t
a

t
a
1dx = a ln(a) a (t ln(t) t).
Now, it can be easily deduced by using L'Hpital's Rule that .
td0
+
limt ln(t) = 0
Therefore,
.
0
a
ln(x)dx =
td0
+
lim
t
a
ln(x)dx = aln(a) a
Hence, by Theorem 45, ln(x) is Lebesgue integrable on [0, a] for any a > 0 and the
Lebesgue integral is aln(a)a.
(4) diverges.
1

1
ln(x)
dx
By using L'Hpital's Rule, we can show that .
xd
lim
x
ln(x)
=
Therefore, there exists K > 0 such that . x m K u
x
ln(x)
> 1 u
1
ln(x)
>
1
x
Since diverges, it follows that diverges by Theorem 50
K

1
x
dx
K

1
ln(x)
dx
(Comparison Test).
Therefore, diverges. Hence is not Lebesgue integrable on [1, ).
1

1
ln(x)
dx
1
ln(x)
For a function, not necessarily non-negative, we can, with sufficient condition
satisfied by the function, formulate a convergence criterion along the line of the
Leibnitz alternating series test.
Theorem 55.
Suppose f :[a, ) R is such that f is Riemann integrable on [a, b] for every b >
a. Suppose there is a strictly increasing sequence (an) in [a, ) with a
0
= a and an
satisfying :
(1) f (x) has constant sign in each interval (an , an+1
),
(2) f (x) changes sign from (an-1
, an) to (an , an+1
),
(3) ,
a
n1
an
f (x)dx m
an
a
n+1
f (x)dx
(4) .
a
n1
an
f (x)dx d0 as n d
Then the improper Riemann integral converges.
a

f (x)dx
Proof. Let . Then by condition (4) cn 0 . By condition (3) ( |cn| ) c
n
=
a
n1
an
f (x)dx
is a decreasing sequence. By condition (2) is an alternating series. Therefore,
E
n=1

c
n
by Leibnitz's Alternating Series Test, (see Chapter 6 Series, Theorem 20), is
E
n=1

c
n
convergent.
We shall show next that is the improper integral.
E
n=1

c
n
Chapter 14 Improper Integral and Lebesgue Integral
23
Ng Tze Beng
Now take any b > a. Since an , there exists an integer n
0
with . Let n
0
be a
n
0
m b
the least integer such that so that . a
n
0
m b a
n
0
1
< b [ a
n
0
Thus, .
a
b
f (x)dx =
E
k=1
n
0
1
c
k
+
a
n
0
1
b
f (x)dx
Since and f (x) is of constant sign on . a
n
0
1
< b [ a
n
0
(a
n
0
1
, a
n
0
)
---------------------- (1)
a
n
0
1
b
f (x)dx [
a
n
0
1
an
0
f (x)dx = |c
n
0
|
Plainly, as b , n
0
.
Now as cn 0, given > 0, there exists an integer N
1
such that
n N
1
|cn | < /2 -------------------- (2)
Since is convergent, there exists an integer N
2
such that
E
n=1

c
n
. ----------------------- (3) k m N
2
u
E
n=k

c
n
< c/2
Let N = max(N
1
, N
2
). Then since n
0
as b , there exists a number K > 0 such
that
b >K n
0
> N.
Therefore, for b >K ,

a
b
f (x)dx
E
k=1

c
k
[
a
b
f (x)dx
E
k=1
k=n
0
1
c
k
+
E
k=n
0

c
k
by (1) [ |c
n
0
| +
E
k=n
0

c
k
< /2 + by (2) since n
0
> N N
1 E
k=n
0

c
k
< /2 + /2 = by (3) since n
0
> N N
2
.
Hence . Thus the improper integral is convergent.
bd
lim
a
b
f (x)dx =
E
k=1

c
k

a

f (x)dx
Example 56. Use of the Lebesgue Dominated Convergence Theorem.
.
nd
lim
0
1
1 e
nx
x
dx = 2
Define f n : (0, 1] R by , Then f n converges pointwise to f
n
(x) =
1 e
nx
x
1
x
on (0, 1].
Now for x 0 so that . Hence 0 < e
nx
[ 1 0 [ 1 e
nx
[ 1
for x in (0, 1]. 0 [ f
n
(x) [
1
x
Note that each f n is continuous on (0, 1] and so is measurable on (0, 1].
Now the improper integral . Therefore, by Theorem
0
1
1
x
dx =
td0
+
lim 2 x
t
1
= 2
52(2) , is Lebesgue integrable on (0, 1] and its Lebesgue integral,
1
x
.
(0,1]
1
x
du =
0
1
1
x
dx = 2
Therefore, by the Lebesgue Dominated Convergence Theorem,
.
(0,1]
f
n
du =
0
1
1 e
nx
x
dx d
(0,1]
1
x
du = 2
Chapter 14 Improper Integral and Lebesgue Integral
24
Ng Tze Beng
Note that by the Comparison Test (Theorem 50), the improper integral
is convergent and equals the Lebesgue integral of
0
1
1 e
nx
x
dx =
0
1
f
n
(x)dx
.
(0,1]
f
n
du
14. 8 Differentiation Under the Integral Sign and Lebesgue Integral
Since differentiation is a process of limit, we shall investigate the processes of limits
before and after a Lebesgue integral.
Here is a result that makes use of the Lebesgue Dominated Convergence Theorem.
Theorem 57. Suppose E is a non-empty (Lebesgue) measurable subset of R and I is
an interval. Suppose f : E I R is a function satisfying
(1) for each t in I, the function
f t : E R defined by f t (x) = f (x, t)
is measurable;
(2) there exists an integrable function g: E R such that for almost all x in E and all
t in I,
| f t (x) | = | f (x, t) | g(x);
(3) for some point t
0
(t
0
may be ,) in the closure of I, , there exists a function I
h : E R such that, for almost all x in E, .
tdt
0
limf
t
(x) =
tdt
0
limf (x, t) = h(x)
Then
(1) h is Lebesgue integrable and
(2) .
tdt
0
lim
E
f
t
(x)du(x) =
E tdt
0
lim f
t
(x)du(x) =
E
h(x)du(x)
Proof. Since t
0
, there exists a sequence ( tn ) in I such that t n t
0
. For each n I
in P define the function, gn :E R by
, x in E. g
n
(x) = f
tn
(x) = f (x, t
n
)
By Condition (2), is Lebesgue integrable for each t in I and so gn is Lebsgue f
t
integrable on E. (Note that f t is measurable implies that | f t | is measurable. Then by
Lemma 27, there is a sequence of increasing simple non-negative function tending
pointwise to | f t |, then by condition (2), the integral of these non-negative simple
functions is bounded above by the Lebesgue integral of g and so by the Lebesgue
Monotone Convergence Theorem (Theorem 28), | f t | is Lebesgue integrable.
Consequently, f t is Lebesgue integrable.) By Condition (3) gn converges pointwise to
h(x) almost everywhere. Therefore, by the Lebesgue Dominated Convergence
Theorem, since |gn (x) | g(x) for almost all x in E and g is Lebesgue integrable,
.
E
g
n
(x)du d
E
h(x)du
Note that since , for any sequence ( tn ) in I such that t n t
0
, we conclude g
n
= f
tn
that
.
tdt
0
lim
E
f
t
(x)du =
nd
lim
E
g
n
(x)du =
E
h(x)du =
E tdt
0
lim f
t
(x)du
Now we specialize Theorem 58 to differentiation under an integration sign.
Theorem 58. Suppose E is a non-empty (Lebesgue) measurable subset of R and f :
E (a, b) R is a function satisfying:
Chapter 14 Improper Integral and Lebesgue Integral
25
Ng Tze Beng
(1) For each t in I, the function
f t : E R defined by f t (x) = f (x, t)
is Lebesgue integrable;
(2) For some t
0
in (a, b), the partial derivative exists for almost all x in E;
f
t
(x, t
0
)
and
(3) There exists a neighbourhood V of t
0
and a Lebesgue integrable function g E
R such that for almost all x in E and all t in V,
, D
t
0
(x, t) =
f (x, t) f (x, t
0
)
t t
0
[ g(x)
where . D
t
0
(x, t) =





f (x, t) f (x, t
0
)
t t
0
, t ! t
0
0, if t = t
0
Then,
(i) the function F: E R defined by is a (Lebesgue) integrable F(x) =
f
t
(x, t
0
)
function
and
(ii) the function H: (a, b) R, defined by is H(t) =
E
f
t
du =
E
f (x, t)du
differentiable at t0 and
. H

(t
0
) =
E
F(x)du =
E
f
t
(x, t
0
)du

Proof. Let for t in V, G(t) =
H(t) H(t
0
)
t t
0
=
E
f
t
f
t
0
t t
0
du =
E
f (x, t) f (x, t
0
)
t t
0
du
. =
E
D
t
0
(x, t)du
Note that for each t in V, the function is measurable since are x hD
t
0
(x, t) f
t
and f
t
0
integrable and so measurable. Therefore, satisfies the condition of Theorem D
t
0
(x, t)
57 and
for almost all x in E.
tdt
0
limD
t
0
(x, t) =
f
t
(x, t
0
)
Hence, by Theorem 57,
for t in V, H

(t
0
) =
tdt
0
lim
H(t) H(t
0
)
t t
0
=
tdt
0
lim
E
f
t
f
t
0
t t
0
du =
tdt
0
lim
E
D
t
0
(x, t)du
. =
E tdt
0
limD
t
0
(x, t)du =
E
f
t
(x, t
0
)du
This shows that H is differentiable at t
0
and the derivative . H

(t
0
) =
E
f
t
(x, t
0
)du
Remark.
(1) Conditions (2) and (3) are satisfied by any function f : [c, d] (a, b) R whose
partial derivative is continuous on [c, d] (a, b) R so that it is uniformly
f
t
(x, t)
continuous on [c, d] V, where V is a closed interval in (a, b) containing t
0
. In this
case, by continuity and compactness of [c, d] V, is bounded on [c, d] V ,
f
t
(x, t)
say by M. By the Mean Value Theorem, for any (x, t) in [c, d] V,
D
t
0
(x, t) =
f (x, t) f (x, t
0
)
t t
0
=
f
t
(x, t
1
)
for some t
1
in V. Therefore, . We can just take g in Theorem 58 to be D
t
0
(x, t) [ M
the constant function M.
Chapter 14 Improper Integral and Lebesgue Integral
26
Ng Tze Beng
More generally, if exists for all t in a neighbourhood V of t
0
and for almost
f
t
(x, t)
all x in E and if there exists an integrable function g(x) such that for all
f
t
(x, t) [ g(x)
t in V and for almost all x in E, then by using the Mean Value Theorem, condition (3)
of Theorem 58 is satisfied.
(2) We can replace the interval (a, b) in Theorem 58 by [a, b ), or (a, b] or [a, b].
Taking the derivative at any end point to be the appropriate one sided limit, the
conclusion of Theorem 58 is still valid.
The next Theorem is a special case.
Theorem 59. Suppose f : [c, d] [a, b] R is a continuous function such the
partial derivative exists for all (x, t) in [c, d] [a, b] and is continuous on [c,
f
t
(x, t)
d] [a, b]. Let F: [a, b] R be defined by F(t) =
c
d
f (x, t)dx.
Then, F is differentiable and . F

(t) =
c
d
f
t
(x, t)dx
Proof. For any t0 in [a, b], there exists a closed interval V containing t
0
such that
is continuous on [c, d] V. Therefore, by the remark preceding the theorem,
f
t
(x, t)
conditions (2) and (3) of Theorem 58 are satisfied. Condition (1) of Theorem 58 is of
course satisfied since for each t, f t : [a, b] R defined by f t (x) = f (x, t), is
continuous and so Riemann integrable, hence Lebesgue integrable and the Lebesgue
integral is equal to the Riemann integral Similarly the
[c,d]
f (x, t)du(x)
c
d
f (x, t)dx.
Lebesgue integral is the same as the Riemann integral
[c,d]
f
t
(x, t)du(x)
c
d
f
t
(x, t)dx
. Therefore, by Theorem 58,
. F

(t) =
c
d
f
t
(x, t)dx
What happens if the integral is given by improper integral? If the improper integral is
of the type with singularity at d, or d = , then even if exists for all (x, t),
f
t
(x, t)
may not be continuous on [c, d] [a, b] or that [c, d] is unbounded when (d =
f
t
(x, t)
). is not generally uniformly bounded. Hence we may need to introduce
f
t
(x, t)
the notion of uniform convergence for the improper Riemann Cauchy integral
.
c
d
f
t
(x, t)dx
Theorem 60. Suppose f : [c, ) [a, b] R is a continuous function such that the
partial derivative exists for all (x, t) in [c, ) [a, b] and is continuous on [c,
f
t
(x, t)
) [a, b].
Chapter 14 Improper Integral and Lebesgue Integral
27
Ng Tze Beng
Suppose that the improper Cauchy Riemann integral converges for each t
c

f (x, t)dx
in [a, b]. Suppose furthermore the improper integral converges
c

f
t
(x, t)dx
uniformly for t in [a, b]. Let F: [a,b] R be defined by . F(t) =
c

f (x, t)dx
Then, F is differentiable and . F

(t) =
c

f
t
(x, t)dx
Proof. converge uniformly in [a, b] means that
c

f
t
(x, t)dx
as d for each t in [a, b],
c
d
f
t
(x, t)dx d G(t)
where , and that given > 0, there exists a real number K > 0 such G(t) =
c

f
t
(x, t)dx
that for all d > K and for all t in [a, b],
----------------------------- (1)
c
d
f
t
(x, t)dx G(t) < c
For each d > c, define Then Fd converges pointwise to F on [a, F
d
(t) =
c
d
f (x, t)dx.
b], where . By Theorem 59, Fd is differentiable and F(t) =
c

f (x, t)dx
. F
d

(t) =
c
d
f
t
(x, t)dx
Then condition (1) says that Fd ' (t) converges uniformly to G(t) on [a, b] as d .
By taking any sequence ( dn ) in (c, ) such that dn and letting , by the H
n
= F
dn
above argument, we have that
Hn converges pointwise to F on [a, b],
Hd ' converges uniformly to G on [a, b].
Then by Theorem 8, Chapter 8, F is differentiable and F' = G.
Remark. As remark in Chapter 8, we only need require that there exists a t
0
in [a, b]
such that Hn(t
0
) converges to F(t
0
). This is equivalent to is convergent.
c

f (x, t
0
)dx
In Theorem 60, instead of requiring to be convergent for each t in [a, b],
c

f (x, t)dx
we need only require convergence at some point t
0
in [a, b]. That it is convergent for
all t is a consequence of the uniform convergence of Fd ' as d tends to infinity.
As an application of the theorems in this section, we shall give an indirect method of
computing the probability integral, more specifically .
0

e
x
2
dx
Example 61. .
0

e
x
2
dx =
v
2
By Theorem 10 (a comparison test), since for x 1 and is e
x
2
[ e
x

e
x
dx
convergent, is convergent. Consequently, the improper Riemann integral
1

e
x
2
dx
is convergent. Since is non-negative, by Theorem 34, is a
0

e
x
2
dx e
x
2

e
x
2
dx
Lebesgue integral on [0, ). Consider the following two functions:
and . f (t) =
0
t
e
x
2
dx
2
g(t) =
0
1
e
t
2
(x
2
+1)
x
2
+ 1
dx
We shall take the limit of these two functions to determine .
0

e
x
2
dx
Observe that by the Chain Rule and the Fundamental Theorem of Calculus,
---------------------- (1) f

(t) = 2e
t
2

0
t
e
x
2
dx
Let . Then h is continuous on [0,1][0,) and so ht(x) = h(x,t) is h(x, t) =
e
t
2
(x
2
+1)
x
2
+ 1
continuous on [0,1] and so is Riemann integrable on [0, 1] and hence Lebesgue
Chapter 14 Improper Integral and Lebesgue Integral
28
Ng Tze Beng
integrable on [0, 1]. The partial derivative exists for all x in [0,
h
t
(x, t) = 2te
t
2
(x
2
+1)
1] and for all t > 0. In particular, for some
h
t
(x, t) = 2te
t
2
(x
2
+1)
[ |2t| [ M
constant number M (depending on t
0
> 0

) for some neighbourhood of any t
0
.
Therefore, by Theorem 58,
. g

(t) =
0
1
h
t
(x, t)dx =
0
1
2te
t
2
(x
2
+1)
dx = 2te
t
2

0
1
e
t
2
x
2
dx
For t > 0, using a change of variable y = xt ,
-------------------- (2) g

(t) = 2e
t
2

0
t
e
y
2
dy = 2e
t
2

0
t
e
x
2
dx
Thus from (1) and (2), we see that f '(t) + g'(t) = 0 for all t > 0. Therefore, f (t) + g(t)
= c for some constant c, for all t > 0. Note that f is obviously continuous at 0.
Since and is Lebesgue integrable on [0,1], by h(x, t) =
e
t
2
(x
2
+1)
x
2
+ 1
[
1
x
2
+ 1
1
x
2
+ 1
Theorem 57, g(t) is continuous at t = 0. Hence f (t) + g(t) is continuous at t = 0.
Therefore, f (t) + g(t) = c
for some constant c, for all t 0. Thus
. c = f (0) + g(0) = 0 +
0
1
1
x
2
+ 1
dx = tan
1
(1) =
v
4
So we have the equation,
------------------------------- (3). f (t) + g(t) =
v
4
By Theorem 57,
since for all x in [0, 1].
td
lim g(t) =
0
1
td
lim
e
t
2
(x
2
+1)
x
2
+ 1
dx = 0
td
lim
e
t
2
(x
2
+1)
x
2
+ 1
= 0
Thus, passing to the limit (3) becomes

td
lim f (t) +
td
lim g(t) =
v
4
and so . This means and so .
td
lim f (t) =
v
4

0

e
x
2
dx
2
=
v
4

0

e
x
2
dx =
v
2
Exercises 62.
1. Determine for each of the following integrals whether it is convergent,
absolute convergent, or conditionally convergent. Evaluate the convergent ones
if possible.
(a) , n 1, (b) , n > 1, (c) , n 1,
0
1
1
x
n
dx
1

1
x
n
dx
1

1
x
n
dx
(d) , (e) , n > 1 (f) ,
0
v/2
tan(x)dx
1
sin(x)
x
n
dx
0
1
(
ln(x)
)
2
dx
(g) , (h) , (i) , n > 0 ,
0

e
x
cos(2x)dx
0
v/2
1
1 cos(x)
dx
0

x
n
e
x
dx
(j) (k) , (l)
0

sin
2
(2x) cos
2
(2x)dx
0

e
x
x
dx
0

1
1 + e
x
dx
(m) , (n) , (o) , n 0.
0

x
2
e
x
2
dx
0

1
x (1 + x)
dx
0

x
2n+1
e
x
2
dx
Chapter 14 Improper Integral and Lebesgue Integral
29
Ng Tze Beng
2. Suppose is an absolutely convergent improper integral and g is a
a
b
f (x)dx
bounded integrable function on [a, b]. Prove that converges
a
b
f (x)g(x)dx
absolutely.
3. Prove that and are convergent.
0

sin(x
2
)dx
0

cos(x
2
)dx
4. Show that .
0
sin
2
(x)
x
2
dx =
0
sin(x)
x
dx
5. Determine .
0
v/2
ln(sin(x))dx
6. Test the following improper integrals for convergence.
(a) , 1 , (b) , (c) ,
0

sin(x
c
)dx
0

x
2
(
1 + x
2
)
2
dx
0

x
2
(
1 + x
2
)
2
dx
(d) , (e) , (f) ,
1

1
x
1+1/x
dx
1
4 x
ln
(
x
)

0

1
x
cos
1
x
dx
(g) , (h) (i) .
0
v
x
sin(x)
dx
0

x sin
(
e
x
)
dx
0

e
cos(x)
x
dx
7. Suppose f is a positive continuous function defined on [a, ). Prove that
is convergent if .
a
b
f (x)dx
x d
lim
[ f (x + 1) ]
f (x)
< 1
8. Prove that converges absolutely if > 2. [Hint: consider the
0

e
t
t

sin(t)dt
integral from 0 to 1 and from 1 to separately. ]
9. Test the integral for convergence.
0
v/2
1
1 sin(x)
dx
10. Prove that for t 0, . Hence deduce that
0

e
t x
sin(x)
x
dx =
v
2
tan
1
(t)
.
0
sin(x)
x
dx =
v
2
11. Show that for t > 0.
0

e
tx
2
dx =
1
2
v
t
12. Show that the function f (x) = ln(x) / x is Lebesgue integrable over [1, ) and
that the Lebesgue integral .
[1, )
f du = 1
13. For any t in R, prove that .
0

e
x
2
cos(2tx)dx =
v
2
e
t
2



Chapter 14 Improper Integral and Lebesgue Integral
30
Ng Tze Beng