FUNCTIONS DEFINED BY

IMPROPER INTEGRALS
William F. Trench
Professor Emeritus
Department of Mathematics
Trinity University
San Antonio, Texas, USA
[email protected]
Copyright May 2012, WilliamF. Trench
This is a supplement to the authors
Introduction to Real Analysis
Reproduction and online posting are permitted for any valid noncommercial edu-
cational, mathematical, or scientic purpose. However, charges for prot beyond
reasonable printing costs are strictly prohibited. A complete instructors solution
manual is available on request by email to the author, subject to verication of the
requestors faculty status.
1 Foreword
This is a revised version of Section 7.5 of my Advanced Calculus (Harper & Row,
1978). It is a supplement to my textbook Introduction to Real Analysis, which is refer-
enced several times here. You should review Section 3.4 (Improper Integrals) of that
book before reading this document.
2 Introduction
In Section 7.2 (pp. 462484) we considered functions of the form
J(,) =
_
b
a
(.. ,) J.. c _ , _ J.
We saw that if is continuous on a. b| c. J|, then J is continuous on c. J| (Exer-
cise 7.2.3, p. 481) and that we can reverse the order of integration in
_
d
c
J(,) J, =
_
d
c
_
_
b
a
(.. ,) J.
_
J,
to evaluate it as
_
d
c
J(,) J, =
_
b
a
_
_
d
c
(.. ,) J,
_
J.
(Corollary 7.2.3, p. 466).
Here is another important property of J.
Theorem 1 If and
y
are continuous on a. b| c. J|. then
J(,) =
_
b
a
(.. ,) J.. c _ , _ J. (1)
is continuously differentiable on c. J| and J
0
(,) can be obtained by differentiating (1)
under the integral sign with respect to ,: that is,
J
0
(,) =
_
b
a

y
(.. ,) J.. c _ , _ J. (2)
Here J
0
(a) and
y
(.. a) are derivatives from the right and J
0
(b) and
y
(.. b) are
derivatives from the left.
Proof If , and , z, are in c. J| and z, = 0, then
J(, z,) J(,)
z,
=
_
b
a
(.. , z,) (.. ,)
z,
J.. (3)
From the mean value theorem (Theorem 2.3.11, p. 83), if . a. b| and ,, , z,
c. J|, there is a ,(.) between , and , z, such that
(.. ,z,)(.. ,) =
y
(.. ,)z, =
y
(.. ,(.))z,(
y
(.. ,(.)
y
(.. ,))z,.
2
From this and (3),

J(, z,) J(,)

z,

_
b
a

y
(.. ,) J.

_
_
b
a
[
y
(.. ,(.))
y
(.. ,)[ J.. (4)
Now suppose c > 0. Since
y
is uniformly continuous on the compact set a. b|c. J|
(Corollary 5.2.14, p. 314) and ,(.) is between , and , z,, there is a > 0 such
that if [z[ < then
[
y
(.. ,)
y
(.. ,(.))[ < c. (.. ,) a. b| c. J|.
This and (4) imply that

J(, z, J(,))
z,

_
b
a

y
(.. ,) J.

< c(b a)
if , and , z, are in c. J| and 0 < [z,[ < . This implies (2). Since the integral
in (2) is continuous on c. J| (Exercise 7.2.3, p. 481, with replaced by
y
), J
0
is
continuous on c. J|.
Example 1 Since
(.. ,) = cos ., and
y
(.. ,) = . sin .,
are continuous for all (.. ,), Theorem 1 implies that if
J(,) =
_

0
cos ., J.. o< , < o. (5)
then
J
0
(,) =
_

0
. sin ., J.. o< , < o. (6)
(In applying Theorem 1 for a specic value of ,, we take 1 = 0. | j. j|, where
j > [,[.) This provides a convenient way to evaluate the integral in (6): integrating the
right side of (5) with respect to . yields
J(,) =
sin .,
,

xD0
=
sin ,
,
. , = 0.
Differentiating this and using (6) yields
_

0
. sin ., J. =
sin ,
,
2

cos ,
,
. , = 0.
To verify this, use integration by parts.
We will study the continuity, differentiability, and integrability of
J(,) =
_
b
a
(.. ,) J.. , S.
3
where S is an interval or a union of intervals, and J is a convergent improper integral
for each , S. If the domain of is a. b) S where o < a < b _ o, we say
that J is pointwise convergent on S or simply convergent on S, and write
_
b
a
(.. ,) J. = lim
r!b
_
r
a
(.. ,) J. (7)
if, for each , S and every c > 0, there is an r = r
0
(,) (which also depends on c)
such that

J(,)
_
r
a
(.. ,) J.

_
b
r
(.. ,) J.

< c. r
0
(,) _ , < b. (8)
If the domain of is (a. b| S where o_ a < b < o, we replace (7) by
_
b
a
(.. ,) J. = lim
r!aC
_
b
r
(.. ,) J.
and (8) by

J(,)
_
b
r
(.. ,) J.

_
r
a
(.. ,) J.

< c. a < r _ r
0
(,).
In general, pointwise convergence of J for all , S does not imply that J is
continuous or integrable on c. J|, and the additional assumptions that
y
is continuous
and
_
b
a

y
(.. ,) J. converges do not imply (2).
Example 2 The function
(.. ,) = ,e
jyjx
is continuous on 0. o) (o. o) and
J(,) =
_
1
0
(.. ,) J. =
_
1
0
,e
jyjx
J.
converges for all ,, with
J(,) =
_

_
1 , < 0.
0 , = 0.
1 , > 0:
therefore, J is discontinuous at , = 0.
Example 3 The function
(.. ,) = ,
3
e
y
2
x
is continuous on 0. o) (o. o). Let
J(,) =
_
1
0
(.. ,) J. =
_
1
0
,
3
e
y
2
x
J. = ,. o< , < o.
4
Then
J
0
(,) = 1. o< , < o.
However,
_
1
0
d
d,
(,
3
e
y
2
x
) J. =
_
1
0
(3,
2
2,
4
.)e
y
2
x
J. =
_
1. , = 0.
0. , = 0.
so
J
0
(,) =
_
1
0
d(.. ,)
d,
J. if , = 0.
3 Preparation
We begin with two useful convergence criteria for improper integrals that do not involve
a parameter. Consistent with the denition on p. 152, we say that is locally integrable
on an interval 1 if it is integrable on every nite closed subinterval of 1.
Theorem 2 ( Cauchy Criterion for Convergence of an Improper Integral I) Suppose
g is locally integrable on a. b) and denote
G(r) =
_
r
a
g(.) J.. a _ r < b.
Then the improper integral
_
b
a
g(.) J. converges if and only if. for each c > 0. there
is an r
0
a. b) such that
[G(r) G(r
1
)[ < c. r
0
_ r. r
1
< b. (9)
Proof For necessity, suppose
_
b
a
g(.) J. = 1. By denition, this means that for
each c > 0 there is an r
0
a. b) such that
[G(r) 1[ <
c
2
and [G(r
1
) 1[ <
c
2
. r
0
_ r. r
1
< b.
Therefore
[G(r) G(r
1
)[ = [(G(r) 1) (G(r
1
) 1)[
_ [G(r) 1[ [G(r
1
) 1[ < c. r
0
_ r. r
1
< b.
For sufciency, (9) implies that
[G(r)[ = [G(r
1
) (G(r) G(r
1
))[ < [G(r
1
)[ [G(r) G(r
1
)[ _ [G(r
1
)[ c.
r
0
_ r _ r
1
< b. Since G is also bounded on the compact set a. r
0
| (Theorem 5.2.11,
p. 313), G is bounded on a. b). Therefore the monotonic functions
G(r) = sup

G(r
1
)

r _ r
1
< b
_
and G(r) = inf

G(r
1
)

r _ r
1
< b
_
5
are well dened on a. b), and
lim
r!b
G(r) = 1 and lim
r!b
G(r) = 1
both exist and are nite (Theorem 2.1.11, p. 47). From (9),
[G(r) G(r
1
)[ = [(G(r) G(r
0
)) (G(r
1
) G(r
0
))[
_ [G(r) G(r
0
)[ [G(r
1
) G(r
0
)[ < 2c.
so
G(r) G(r) _ 2c. r
0
_ r. r
1
< b.
Since c is an arbitrary positive number, this implies that
lim
r!b
(G(r) G(r)) = 0.
so 1 = 1. Let 1 = 1 = 1. Since
G(r) _ G(r) _ G(r).
it follows that lim
r!b
G(r) = 1.
We leave the proof of the following theorem to you (Exercise 2).
Theorem 3 (Cauchy Criterion for Convergence of an Improper Integral II) Suppose
g is locally integrable on (a. b| and denote
G(r) =
_
b
r
g(.) J.. a _ r < b.
Then the improper integral
_
b
a
g(.) J. converges if and only if. for each c > 0. there
is an r
0
(a. b| such that
[G(r) G(r
1
)[ < c. a < r. r
1
_ r
0
.
To see why we associate Theorems 2 and 3 with Cauchy, compare them with The-
orem 4.3.5 (p. 204)
4 Uniform convergence of improper integrals
Henceforth we deal with functions = (.. ,) with domains 1 S, where S is an
interval or a union of intervals and 1 is of one of the following forms:
v a. b) with o< a < b _ o;
v (a. b| with o_ a < b < o;
v (a. b) with o_ a _ b _ o.
6
In all cases it is to be understood that is locally integrable with respect to . on 1.
When we say that the improper integral
_
b
a
(.. ,) J. has a stated property on S we
mean that it has the property for every , S.
Denition 1 If the improper integral
_
b
a
(.. ,) J. = lim
r!b
_
r
a
(.. ,) J. (10)
converges on S. it is said to converge uniformly (or be uniformly convergent) on S if.
for each c > 0. there is an r
0
a. b) such that

_
b
a
(.. ,) J.
_
r
a
(.. ,) J.

< c. , S. r
0
_ r < b.
or. equivalently.

_
b
r
(.. ,) J.

< c. , S. r
0
_ r < b. (11)
The crucial difference between pointwise and uniform convergence is that r
0
(,) in
(8) may depend upon the particular value of ,, while the r
0
in (11) does not: one choice
must work for all , S. Thus, uniform convergence implies pointwise convergence,
but pointwise convergence does not imply uniform convergence.
Theorem 4 (Cauchy Criterion for Uniform Convergence I) The improper integral
in (10) converges uniformly on S if and only if. for each c > 0. there is an r
0
a. b)
such that

_
r
1
r
(.. ,) J.

< c. , S. r
0
_ r. r
1
< b. (12)
Proof Suppose
_
b
a
(.. ,) J. converges uniformly on S and c > 0. From Deni-
tion 1, there is an r
0
a. b) such that

_
b
r
(.. ,) J.

<
c
2
and

_
b
r
1
(.. ,) J.

<
c
2
. , S. r
0
_ r. r
1
< b. (13)
Since
_
r
1
r
(.. ,) J. =
_
b
r
(.. ,) J.
_
b
r
1
(.. ,) J..
(13) and the triangle inequality imply (12).
For the converse, denote
J(,) =
_
r
a
(.. ,) J..
7
Since (12) implies that
[J(r. ,) J(r
1
. ,)[ < c. , S. r
0
_ r. r
1
< b. (14)
Theorem 2 with G(r) = J(r. ,) (, xed but arbitrary in S) implies that
_
b
a
(.. ,) J.
converges pointwise for , S. Therefore, if c > 0 then, for each , S, there is an
r
0
(,) a. b) such that

_
b
r
(.. ,) J.

< c. , S. r
0
(,) _ r < b. (15)
For each , S, choose r
1
(,) _ maxr
0
(,). r
0
|. (Recall (14)). Then
_
b
r
(.. ,) J. =
_
r
1
.y/
r
(.. ,) J.
_
b
r
1
.y/
(.. ,) J..
so (12), (15), and the triangle inequality imply that

_
b
r
(.. ,) J.

< 2c. , S. r
0
_ r < b.
In practice, we dont explicitly exhibit r
0
for each given c. It sufces to obtain
estimates that clearly imply its existence.
Example 4 For the improper integral of Example 2,

_
1
r
(.. ,) J.

=
_
1
r
[,[e
jyjx
= e
rjyj
. , = 0.
If [,[ _ j, then

_
1
r
(.. ,) J.

_ e
r
.
so
_
1
0
(.. ,) J. converges uniformly on (o. j| L j. o) if j > 0; however, it
does not converge uniformly on any neighborhood of , = 0, since, for any r > 0,
e
rjyj
>
1
2
if [,[ is sufciently small.
Denition 2 If the improper integral
_
b
a
(.. ,) J. = lim
r!aC
_
b
r
(.. ,) J.
converges on S. it is said to converge uniformly (or be uniformly convergent) on S if.
for each c > 0. there is an r
0
(a. b| such that

_
b
a
(.. ,) J.
_
b
r
(.. ,) J.

< c. , S. a < r _ r
0
.
or. equivalently.

_
r
a
(.. ,) J.

< c. , S. a < r _ r
0
.
8
We leave proof of the following theorem to you (Exercise 3).
Theorem 5 (Cauchy Criterion for Uniform Convergence II) The improper integral
_
b
a
(.. ,) J. = lim
r!aC
_
b
r
(.. ,) J.
converges uniformly on S if and only if. for each c > 0. there is an r
0
(a. b| such
that

_
r
r
1
(.. ,) J.

< c. , S. a < r. r
1
_ r
0
.
We need one more denition, as follows.
Denition 3 Let = (.. ,) be dened on (a. b) S. where o _ a < b _ o.
Suppose is locally integrable on (a. b) for all , S and let c be an arbitrary point
in (a. b). Then
_
b
a
(.. ,) J. is said to converge uniformly on S if
_
c
a
(.. ,) J. and
_
b
c
(.. ,) J. both converge uniformly on S.
We leave it to you (Exercise 4) to show that this denition is independent of c;
that is, if
_
c
a
(.. ,) J. and
_
b
c
(.. ,) J. both converge uniformly on S for some
c (a. b), then they both converge uniformly on S for every c (a. b).
We also leave it you (Exercise 5) to show that if is bounded on a. b| c. J|
and
_
b
a
(.. ,) J. exists as a proper integral for each , c. J|, then it converges
uniformly on c. J| according to all three Denitions 13.
Example 5 Consider the improper integral
J(,) =
_
1
0
.
1=2
e
xy
J..
which diverges if , _ 0 (verify). Denition 3 applies if , > 0, so we consider the
improper integrals
J
1
(,) =
_
1
0
.
1=2
e
xy
J. and J
2
(,) =
_
1
1
.
1=2
e
xy
J.
separately. Moreover, we could just as well dene
J
1
(,) =
_
c
0
.
1=2
e
xy
J. and J
2
(,) =
_
1
c
.
1=2
e
xy
J.. (16)
where c is any positive number.
Denition 2 applies to J
1
. If 0 < r
1
< r and , _ 0, then

_
r
1
r
.
1=2
e
xy
J.

<
_
r
r
1
.
1=2
J. < 2r
1=2
.
so J
1
(,) converges for uniformly on 0. o).
9
Denition 1 applies to J
2
. Since

_
r
1
r
.
1=2
e
xy
J.

< r
1=2
_
1
r
e
xy
J. =
e
ry
,r
1=2
.
J
2
(,) converges uniformly on j. o) if j > 0. It does not converge uniformly on
(0. j), since the change of variable u = ., yields
_
r
1
r
.
1=2
e
xy
J. = ,
1=2
_
r
1
y
ry
u
1=2
e
u
Ju.
which, for any xed r > 0, can be made arbitrarily large by taking , sufciently small
and r = 1,,. Therefore we conclude that J(,) converges uniformly on j. o) if
j > 0.
Note that that the constant c in (16) plays no role in this argument.
Example 6 Suppose we take
_
1
0
sin u
u
Ju =

2
(17)
as given (Exercise 31(b)). Substituting u = ., with , > 0 yields
_
1
0
sin .,
.
J. =

2
. , > 0. (18)
What about uniform convergence? Since (sin .,),. is continuous at . = 0, Deni-
tion 1 and Theorem 4 apply here. If 0 < r < r
1
and , > 0, then
_
r
1
r
sin .,
.
J. =
1
,
_
cos .,
.

r
1
r

_
r
1
r
cos .,
.
2
J.
_
. so

_
r
1
r
sin .,
.
J.

<
3
r,
.
Therefore (18) converges uniformly on j. o) if j > 0. On the other hand, from (17),
there is a > 0 such that
_
1
u
0
sin u
u
Ju >

4
. 0 _ u
0
< .
This and (18) imply that
_
1
r
sin .,
.
J. =
_
1
yr
sin u
u
Ju >

4
for any r > 0 if 0 < , < ,r. Hence, (18) does not converge uniformly on any interval
(0. j| with j > 0.
10
5 Absolutely Uniformly Convergent Improper Integrals
Denition 4 (Absolute Uniform Convergence I) The improper integral
_
b
a
(.. ,) J. = lim
r!b
_
r
a
(.. ,) J.
is said to converge absolutely uniformly on S if the improper integral
_
b
a
[(.. ,)[ J. = lim
r!b
_
r
a
[(.. ,)[ J.
converges uniformly on S; that is, if, for each c > 0, there is an r
0
a. b) such that

_
b
a
[(.. ,)[ J.
_
r
a
[(.. ,)[ J.

< c. , S. r
0
< r < b.
To see that this denition makes sense, recall that if is locally integrable on a. b)
for all , in S, then so is [ [ (Theorem 3.4.9, p. 161). Theorem 4 with replaced by
[ [ implies that
_
b
a
(.. ,) J. converges absolutely uniformly on S if and only if, for
each c > 0, there is an r
0
a. b) such that
_
r
1
r
[(.. ,)[ J. < c. , S. r
0
_ r < r
1
< b.
Since

_
r
1
r
(.. ,) J.

_
_
r
1
r
[(.. ,)[ J..
Theorem 4 implies that if
_
b
a
(.. ,) J. converges absolutely uniformly on S then it
converges uniformly on S.
Theorem 6 ( Weierstrasss Test for Absolute Uniform Convergence I) Suppose
M = M(.) is nonnegative on a. b).
_
b
a
M(.) J. < o. and
[(.. ,)[ _ M(.). , S. a _ . < b. (19)
Then
_
b
a
(.. ,) J. converges absolutely uniformly on S.
Proof Denote
_
b
a
M(.) J. = 1 < o. By denition, for each c > 0 there is an
r
0
a. b) such that
1 c <
_
r
a
M(.) J. _ 1. r
0
< r < b.
Therefore, if r
0
< r _ r
1
. then
0 _
_
r
1
r
M(.) J. =
__
r
1
a
M(.) J. 1
_

__
r
a
M(.) J. 1
_
< c
11
This and (19) imply that
_
r
1
r
[(.. ,)[ J. _
_
r
1
r
M(.) J. < c. , S. a _ r
0
< r < r
1
< b.
Now Theorem 4 implies the stated conclusion.
Example 7 Suppose g = g(.. ,) is locally integrable on 0. o) for all , S and, for
some a
0
_ 0, there are constants 1 and
0
such that
[g(.. ,)[ _ 1e
p
0
x
. , S. . _ a
0
.
If >
0
and r _ a
0
, then
_
1
r
e
px
[g(.. ,)[ J. =
_
1
r
e
.pp
0
/x
e
p
0
x
[g(.. ,)[ J.
_ 1
_
1
r
e
.pp
0
/x
J. =
1e
.pp
0
/r

0
.
so
_
1
0
e
px
g(.. ,) J. converges absolutely on S. For example, since
[.

sin .,[ < e

p
0
x
and [.

cos .,[ < e

p
0
x
for . sufciently large if
0
> 0, Theorem 4 implies that
_
1
0
e
px
.

sin ., J.
and
_
1
0
e
px
.

cos ., J. converge absolutely uniformly on (o. o) if > 0 and

_ 0. As a matter of fact,
_
1
0
e
px
.

sin ., J. converges absolutely on (o. o)

if > 0 and > 1. (Why?)
Denition 5 (Absolute Uniform Convergence II) The improper integral
_
b
a
(.. ,) J. = lim
r!aC
_
b
r
(.. ,) J.
is said to converge absolutely uniformly on S if the improper integral
_
b
a
[(.. ,)[ J. = lim
r!aC
_
b
r
[(.. ,)[ J.
converges uniformly on S; that is, if, for each c > 0, there is an r
0
(a. b| such that

_
b
a
[(.. ,)[ J.
_
b
r
[(.. ,)[ J.

< c. , S. a < r < r

0
_ b.
We leave it to you (Exercise 7) to prove the following theorem.
Theorem 7 (Weierstrasss Test for Absolute Uniform Convergence II) Suppose
M = M(.) is nonnegative on (a. b|.
_
b
a
M(.) J. < o. and
[(.. ,)[ _ M(.). , S. . (a. b|.
Then
_
b
a
(.. ,) J. converges absolutely uniformly on S.
12
Example 8 If g = g(.. ,) is locally integrable on (0. 1| for all , S and
[g(.. ,)[ _ .

. 0 < . _ .
0
.
for each , S, then
_
1
0
.

g(.. ,) J.
converges absolutely uniformly on S if > 1. To see this, note that if 0 < r <
r
1
_ .
0
, then
_
r
r
1
.

[g(.. ,)[ J. _
_
r
r
1
.

J. =
.
C1
1

r
r
1
<
r
C1
1
.
Applying this with = 0 shows that
J(,) =
_
1
0
.

cos ., J.
converges absolutely uniformly on (o. o) if > 1 and
G(,) =
_
1
0
.

sin ., J.
converges absolutely uniformly on (o. o) if > 2.
By recalling Theorem 4.4.15 (p. 246), you can see why we associate Theorems 6
and 7 with Weierstrass.
6 Dirichlets Tests
Weierstrasss test is useful and important, but it has a basic shortcoming: it applies
only to absolutely uniformly convergent improper integrals. The next theorem applies
in some cases where
_
b
a
(.. ,) J. converges uniformly on S, but
_
b
a
[(.. ,)[ J.
does not.
Theorem 8 (Dirichlets Test for Uniform Convergence I) If g. g
x
. and h are con-
tinuous on a. b) S. then
_
b
a
g(.. ,)h(.. ,) J.
converges uniformly on S if the following conditions are satised:
(a) lim
x!b
_
sup
y2S
[g(.. ,)[
_
= 0:
(b) There is a constant M such that
sup
y2S

_
x
a
h(u. ,) Ju

< M. a _ . < b:
13
(c)
_
b
a
[g
x
(.. ,)[ J. converges uniformly on S.
Proof If
H(.. ,) =
_
x
a
h(u. ,) Ju. (20)
then integration by parts yields
_
r
1
r
g(.. ,)h(.. ,) J. =
_
r
1
r
g(.. ,)H
x
(.. ,) J.
= g(r
1
. ,)H(r
1
. ,) g(r. ,)H(r. ,) (21)

_
r
1
r
g
x
(.. ,)H(.. ,) J..
Since assumption (b) and (20) imply that [H(.. ,)[ _ M. (.. ,) (a. b|S, Eqn. (21)
implies that

_
r
1
r
g(.. ,)h(.. ,) J.

< M
_
2 sup
xr
[g(.. ,)[
_
r
1
r
[g
x
(.. ,)[ J.
_
(22)
on r. r
1
| S.
Now suppose c > 0. From assumption (a), there is an r
0
a. b) such that
[g(.. ,)[ < c on S if r
0
_ . < b. From assumption (c) and Theorem 6, there is
an s
0
a. b) such that
_
r
1
r
[g
x
(.. ,)[ J. < c. , S. s
0
< r < r
1
< b.
Therefore (22) implies that

_
r
1
r
g(.. ,)h(.. ,)

< 3Mc. , S. max(r

0
. s
0
) < r < r
1
< b.
Now Theorem 4 implies the stated conclusion.
The statement of this theorem is complicated, but applying it isnt; just look for a
factorization = gh, where h has a bounded antderivative on a. b) and g is small
near b. Then integrate by parts and hope that something nice happens. A similar
comment applies to Theorem 9, which follows.
Example 9 Let
1(,) =
_
1
0
cos .,
. ,
J.. , > 0.
The obvious inequality

cos .,
. ,

_
1
. ,
is useless here, since
_
1
0
J.
. ,
= o.
14
However, integration by parts yields
_
r
1
r
cos .,
. ,
J. =
sin .,
,(. ,)

r
1
r

_
r
1
r
sin .,
,(. ,)
2
J.
=
sin r
1
,
,(r
1
,)

sin r,
,(r ,)

_
r
1
r
sin .,
,(. ,)
2
J..
Therefore, if 0 < r < r
1
, then

_
r
1
r
cos .,
. ,
J.

<
1
,
_
2
r ,

_
1
r
1
(. ,)
2
_
_
3
,(r ,)
2
_
3
j(r j)
if , _ j > 0. Now Theorem 4 implies that 1(,) converges uniformly on j. o) if
j > 0.
We leave the proof of the following theorem to you (Exercise 10).
Theorem 9 (Dirichlets Test for Uniform Convergence II) If g. g
x
. and h are con-
tinuous on (a. b| S. then
_
b
a
g(.. ,)h(.. ,) J.
converges uniformly on S if the following conditions are satised:
(a) lim
x!aC
_
sup
y2S
[g(.. ,)[
_
= 0:
(b) There is a constant M such that
sup
y2S

_
b
x
h(u. ,) Ju

_ M. a < . _ b:
(c)
_
b
a
[g
x
(.. ,)[ J. converges uniformly on S.
By recalling Theorems 3.4.10 (p. 163), 4.3.20 (p. 217), and 4.4.16 (p. 248), you
can see why we associate Theorems 8 and 9 with Dirichlet.
7 Consequences of uniform convergence
Theorem 10 If = (.. ,) is continuous on either a. b) c. J| or (a. b| c. J|
and
J(,) =
_
b
a
(.. ,) J. (23)
converges uniformly on c. J|. then J is continuous on c. J|. Moreover.
_
d
c
_
_
b
a
(.. ,) J.
_
J, =
_
b
a
_
_
d
c
(.. ,) J,
_
J.. (24)
15
Proof We will assume that is continuous on (a. b| c. J|. You can consider the
other case (Exercise 14).
We will rst show that J in (23) is continuous on c. J|. Since J converges uni-
formly on c. J|, Denition 1 (specically, (11)) implies that if c > 0, there is an
r a. b) such that

_
b
r
(.. ,) J.

< c. c _ , _ J.
Therefore, if c _ ,. ,
0
_ J|, then
[J(,) J(,
0
)[ =

_
b
a
(.. ,) J.
_
b
a
(.. ,
0
) J.

_
r
a
(.. ,) (.. ,
0
)| J.

_
b
r
(.. ,) J.

_
b
r
(.. ,
0
) J.

.
so
[J(,) J(,
0
)[ _
_
r
a
[(.. ,) (.. ,
0
)[ J. 2c. (25)
Since is uniformly continuous on the compact set a. r|c. J| (Corollary 5.2.14,
p. 314), there is a > 0 such that
[(.. ,) (.. ,
0
)[ < c
if (.. ,) and (.. ,
0
) are in a. r| c. J| and [, ,
0
[ < . This and (25) imply that
[J(,) J(,
0
)[ < (r a)c 2c < (b a 2)c
if , and ,
0
are in c. J| and [, ,
0
[ < . Therefore J is continuous on c. J|, so the
integral on left side of (24) exists. Denote
1 =
_
d
c
_
_
b
a
(.. ,) J.
_
J,. (26)
We will show that the improper integral on the right side of (24) converges to 1. To
this end, denote
1(r) =
_
r
a
_
_
d
c
(.. ,) J,
_
J..
Since we can reverse the order of integration of the continuous function over the
rectangle a. r| c. J| (Corollary 7.2.2, p. 466),
1(r) =
_
d
c
__
r
a
(.. ,) J.
_
J,.
16
From this and (26),
1 1(r) =
_
d
c
_
_
b
r
(.. ,) J.
_
J,.
Now suppose c > 0. Since
_
b
a
(.. ,) J. converges uniformly on c. J|, there is an
r
0
(a. b| such that

_
b
r
(.. ,) J.

< c. r
0
< r < b.
so [1 1(r)[ < (J c)c if r
0
< r < b. Hence,
lim
r!b
_
r
a
_
_
d
c
(.. ,) J,
_
J. =
_
d
c
_
_
b
a
(.. ,) J.
_
J,.
which completes the proof of (24).
Example 10 It is straightforward to verify that
_
1
0
e
xy
J. =
1
,
. , > 0.
and the convergence is uniformon j. o) if j > 0. Therefore Theorem 10 implies that
if 0 < ,
1
< ,
2
, then
_
y
2
y
1
J,
,
=
_
y
2
y
1
__
1
0
e
xy
J.
_
J, =
_
1
0
__
y
2
y
1
e
xy
J,
_
J,
=
_
1
0
e
xy
1
e
xy
2
.
J..
Since
_
y
2
y
1
J,
,
= log
,
2
,
1
. ,
2
_ ,
1
> 0.
it follows that
_
1
0
e
xy
1
e
xy
2
.
J. = log
,
2
,
1
. ,
2
_ ,
1
> 0.
Example 11 From Example 6,
_
1
0
sin .,
.
J. =

2
. , > 0.
and the convergence is uniform on j. o) if j > 0. Therefore, Theorem 10 implies
that if 0 < ,
1
< ,
2
, then

2
(,
2
,
1
) =
_
y
2
y
1
__
1
0
sin .,
.
J.
_
J, =
_
1
0
__
y
2
y
1
sin .,
.
J,
_
J.
=
_
1
0
cos .,
1
cos .,
2
.
2
J.. (27)
17
The last integral converges uniformly on (o. o) (Exercise 10(h)), and is therefore
continuous with respect to ,
1
on (o. o), by Theorem 10; in particular, we can let
,
1
0in (27) and replace ,
2
by , to obtain
_
1
0
1 cos .,
.
2
J. =
,
2
. , _ 0.
The next theorem is analogous to Theorem 4.4.20 (p. 252).
Theorem 11 Let and
y
be continuous on either a. b) c. J| or (a. b| c. J|.
Suppose that the improper integral
J(,) =
_
b
a
(.. ,) J.
converges for some ,
0
c. J| and
G(,) =
_
b
a

y
(.. ,) J.
converges uniformly on c. J|. Then J converges uniformly on c. J| and is given ex-
plicitly by
J(,) = J(,
0
)
_
y
y
0
G(t ) Jt. c _ , _ J.
Moreover, J is continuously differentiable on c. J|; specically,
J
0
(,) = G(,). c _ , _ J. (28)
where J
0
(c) and
y
(.. c) are derivatives from the right, and J
0
(J) and
y
(.. J) are
derivatives from the left.
Proof We will assume that and
y
are continuous on a. b) c. J|. You can
consider the other case (Exercise 15).
Let
J
r
(,) =
_
r
a
(.. ,) J.. a _ r < b. c _ , _ J.
Since and
y
are continuous on a. r| c. J|, Theorem 1 implies that
J
0
r
(,) =
_
r
a

y
(.. ,) J.. c _ , _ J.
Then
J
r
(,) = J
r
(,
0
)
_
y
y
0
__
r
a

y
(.. t ) J.
_
Jt
= J(,
0
)
_
y
y
0
G(t ) Jt
(J
r
(,
0
) J(,
0
))
_
y
y
0
_
_
b
r

y
(.. t ) J.
_
Jt. c _ , _ J.
18
Therefore,

J
r
(,) J(,
0
)
_
y
y
0
G(t ) Jt

_ [J
r
(,
0
) J(,
0
)[

_
y
y
0
_
b
r

y
(.. t ) J.

Jt. (29)
Now suppose c > 0. Since we have assumed that lim
r!b
J
r
(,
0
) = J(,
0
) exists,
there is an r
0
in (a. b) such that
[J
r
(,
0
) J(,
0
)[ < c. r
0
< r < b.
Since we have assumed that G(,) converges for , c. J|, there is an r
1
a. b) such
that

_
b
r

y
(.. t ) J.

< c. t c. J|. r
1
_ r < b.
Therefore, (29) yields

J
r
(,) J(,
0
)
_
y
y
0
G(t ) Jt

< c(1 [, ,
0
[) _ c(1 J c)
if max(r
0
. r
1
) _ r < b and t c. J|. Therefore J(,) converges uniformly on c. J|
and
J(,) = J(,
0
)
_
y
y
0
G(t ) Jt. c _ , _ J.
Since G is continuous on c. J| by Theorem 10, (28) follows from differentiating this
(Theorem 3.3.11, p. 141).
Example 12 Let
1(,) =
_
1
0
e
yx
2
J.. , > 0.
Since
_
r
0
e
yx
2
J. =
1
_
,
_
r
p
y
0
e
t
2
Jt.
it follows that
1(,) =
1
_
,
_
1
0
e
t
2
Jt.
and the convergence is uniform on j. o) if j > 0 (Exercise 8(i)). To evaluate the last
integral, denote J(j) =
_

0
e
t
2
Jt ; then
J
2
(j) =
__

0
e
u
2
Ju
___

0
e
v
2
J
_
=
_

0
_

0
e
.u
2
Cv
2
/
JuJ.
Transforming to polar coordinates r = r cos 0, = r sin 0 yields
J
2
(j) =
_
=2
0
_

0
re
r
2
Jr J0 =
(1 e

2
)
4
. so J(j) =
_
(1 e

2
)
2
.
19
Therefore
_
1
0
e
t
2
Jt = lim
!1
J(j) =
_

2
and
_
1
0
e
yx
2
J. =
1
2
_

,
. , > 0.
Differentiating this n times with respect to , yields
_
1
0
.
2n
e
yx
2
J. =
1 3 (2n 1)
_

2
n
,
nC1=2
, > 0. n = 1. 2. 3. . . . .
where Theorem 11 justies the differentiation for every n, since all these integrals
converge uniformly on j. o) if j > 0 (Exercise 8(i)).
Some advice for applying this theorem: Be sure to check rst that J(,
0
) =
_
b
a
(.. ,
0
) J. converges for at least one value of ,. If so, differentiate
_
b
a
(.. ,) J.
formally to obtain
_
b
a

y
(.. ,) J.. Then J
0
(,) =
_
b
a

y
(.. ,) J. if , is in some
interval on which this improper integral converges uniformly.
8 Applications to Laplace transforms
The Laplace transform of a function locally integrable on 0. o) is
J(s) =
_
1
0
e
sx
(.) J.
for all s such that integral converges. Laplace transforms are widely applied in mathe-
matics, particularly in solving differential equations.
We leave it to you to prove the following theorem (Exercise 26).
Theorem 12 Suppose is locally integrable on 0. o) and [(.)[ _ Me
s
0
x
for suf-
ciently large .. Then the Laplace transform of J converges uniformly on s
1
. o) if
s
1
> s
0
.
Theorem 13 If is continuous on 0. o) and H(.) =
_
1
0
e
s
0
u
(u) Ju is bounded
on 0. o). then the Laplace transform of converges uniformly on s
1
. o) if s
1
> s
0
.
Proof If 0 _ r _ r
1
,
_
r
1
r
e
sx
(.) J. =
_
r
1
r
e
.ss
0
/x
e
s
0
x
(.) Jt =
_
r
1
r
e
.ss
0
/t
H
0
(.) Jt.
Integration by parts yields
_
r
1
r
e
sx
(.) Jt = e
.ss
0
/x
H(.)

r
1
r
(s s
0
)
_
r
1
r
e
.ss
0
/x
H(.) J..
20
Therefore, if [H(.)[ _ M, then

_
r
1
r
e
sx
(.) J.

_ M

e
.ss
0
/r
1
e
.ss
0
/r
(s s
0
)
_
r
1
r
e
.ss
0
/x
J.

_ 3Me
.ss
0
/r
_ 3Me
.s
1
s
0
/r
. s _ s
1
.
Now Theorem 4 implies that J(s) converges uniformly on s
1
. o).
The following theorem draws a considerably stonger conclusion from the same
assumptions.
Theorem 14 If is continuous on 0. o) and
H(.) =
_
x
0
e
s
0
u
(u) Ju
is bounded on 0. o). then the Laplace transform of is innitely differentiable on
(s
0
. o). with
J
.n/
(s) = (1)
n
_
1
0
e
sx
.
n
(.) J.: (30)
that is, the n-th derivative of the Laplace transform of (.) is the Laplace transform
of (1)
n
.
n
(.).
Proof First we will show that the integrals
1
n
(s) =
_
1
0
e
sx
.
n
(.) J.. n = 0. 1. 2. . . .
all converge uniformly on s
1
. o) if s
1
> s
0
. If 0 < r < r
1
, then
_
r
1
r
e
sx
.
n
(.) J. =
_
r
1
r
e
.ss
0
/x
e
s
0
x
.
n
(.) J. =
_
r
1
r
e
.ss
0
/x
.
n
H
0
(.) J..
Integrating by parts yields
_
r
1
r
e
sx
.
n
(.) J. = r
n
1
e
.ss
0
/r
1
H(r) r
n
e
.ss
0
/r
H(r)

_
r
1
r
H(.)
_
e
.ss
0
/x
.
n
_
0
J..
where
0
indicates differentiation with respect to .. Therefore, if [H(.)[ _ M _ oon
0. o), then

_
r
1
r
e
sx
.
n
(.) J.

_ M
_
e
.ss
0
/r
r
n
e
.ss
0
/r
r
n

_
1
r
[(e
.ss
0
/x
).
n
)
0
[ J.
_
.
Therefore, since e
.ss
0
/r
r
n
decreases monotonically on (n. o) if s > s
0
(check!),

_
r
1
r
e
sx
.
n
(.) J.

< 3Me
.ss
0
/r
r
n
. n < r < r
1
.
so Theorem 4 implies that 1
n
(s) converges uniformly s
1
. o) if s
1
> s
0
. Now The-
orem 11 implies that J
nC1
= J
0
n
, and an easy induction proof yields (30) (Exer-
cise 25).
21
Example 13 Here we apply Theorem 12 with (.) = cos a. (a = 0) and s
0
= 0.
Since
_
x
0
cos auJu =
sin a.
a
is bounded on (0. o), Theorem 12 implies that
J(s) =
_
1
0
e
sx
cos a. J.
converges and
J
.n/
(s) = (1)
n
_
1
0
e
sx
.
n
cos a. J.. s > 0. (31)
(Note that this is also true if a = 0.) Elementary integration yields
J(s) =
s
s
2
a
2
.
Hence, from (31),
_
1
0
e
sx
.
n
cos a. = (1)
n
J
n
Js
n
s
s
2
a
2
. n = 0. 1. . . . .
22
9 Exercises
1. Suppose g and h are differentiable on a. b|, with
a _ g(,) _ b and a _ h(,) _ b. c _ , _ J.
Let and
y
be continuous on a. b| c. J|. Derive Liebnizs rule:
J
J,
_
h.y/
g.y/
(.. ,) J. = (h(,). ,)h
0
(,) (g(,). ,)g
0
(,)

_
h.y/
g.y/

y
(.. ,) J..
(Hint: Dene H(,. u. ) =
_
v
u
(.. ,) J. and use the chain rule.)
2. Adapt the proof of Theorem 2 to prove Theorem 3.
3. Adapt the proof of Theorem 4 to prove Theorem 5.
4. Showthat Denition 3 is independent of c; that is, if
_
c
a
(.. ,) J. and
_
b
c
(.. ,) J.
both converge uniformly on S for some c (a. b), then they both converge uni-
formly on S and every c (a. b).
5. (a) Show that if is bounded on a. b| c. J| and
_
b
a
(.. ,) J. exists as a
proper integral for each , c. J|, then it converges uniformly on c. J|
according to all of Denition 13.
(b) Give an example to show that the boundedness of is essential in (a).
6. Working directly from Denition 1, discuss uniform convergence of the follow-
ing integrals:
(a)
_
1
0
J.
1 ,
2
.
2
J. (b)
_
1
0
e
xy
.
2
J.
(c)
_
1
0
.
2n
e
yx
2
J. (d)
_
1
0
sin .,
2
J.
(e)
_
1
0
(3,
2
2.,)e
y
2
x
J. (f)
_
1
0
(2., ,
2
.
2
)e
xy
J.
7. Adapt the proof of Theorem 6 to prove Theorem 7.
8. Use Weierstrasss test to show that the integral converges uniformly on S :
(a)
_
1
0
e
xy
sin . J., S = j. o), j > 0
(b)
_
1
0
sin .
.
y
J., S = c. J|, 1 < c < J < 2
23
(c)
_
1
1
e
px
sin .,
.
J., > 0, S = (o. o)
(d)
_
1
0
e
xy
(1 .)
y
J., S = (o. b), b < 1
(e)
_
1
1
cos .,
1 .
2
,
2
J., S = (o. j| Lj. o), j > 0.
(f)
_
1
1
e
x=y
J., S = j. o), j > 0
(g)
_
1
1
e
xy
e
x
2
J., S = j. j|, j > 0
(h)
_
1
0
cos ., cos a.
.
2
J., S = (o. o)
(i)
_
1
0
.
2n
e
yx
2
J., S = j. o), j > 0, n = 0, 1, 2,. . .
9. (a) Show that
1(,) =
_
1
0
.
y1
e
x
J.
converges if , > 0, and uniformly on c. J| if 0 < c < J < o.
(b) Use integration by parts to show that
1 (,) =
1(, 1)
,
. , _ 0.
and then show by induction that
1 (,) =
1 (, n)
,(, 1) (, n 1)
. , > 0. n = 1. 2. 3. . . . .
How can this be used to dene 1(,) in a natural way for all , = 0, 1,
2, . . . ? (This function is called the gamma function.)
(c) Show that 1 (n 1) = n if n is a positive integer.
(d) Show that
_
1
0
e
st
t

Jt = s
1
1( 1). > 1. s > 0.
10. Show that Theorem 8 remains valid with assumption (c) replaced by the as-
sumption that [g
x
(.. ,)[ is monotonic with respect to . for all , S.
11. Adapt the proof of Theorem 8 to prove Theorem 9.
12. Use Dirichlets test to show that the following integrals converge uniformly on
S = j. o) if j > 0:
24
(a)
_
1
1
sin .,
.
y
J. (b)
_
1
2
sin .,
log .
J.
(c)
_
1
0
cos .,
. ,
2
J. (d)
_
1
1
sin .,
1 .,
J.
13. Suppose g. g
x
and h are continuous on a. b) S. and denote H(.. ,) =
_
x
a
h(u. ,) Ju. a _ . < b. Suppose also that
lim
x!b
_
sup
y2S
[g(.. ,)H(.. ,)[
_
= 0 and
_
b
a
g
x
(.. ,)H(.. ,) J.
converges uniformly on S. Showthat
_
b
a
g(.. ,)h(.. ,) J. converges uniformly
on S.
14. Prove Theorem 10 for the case where = (.. ,) is continuous on (a. b|
c. J|.
15. Prove Theorem 11 for the case where = (.. ,) is continuous on (a. b|
c. J|.
16. Show that
C(,) =
_
1
1
(.) cos ., J. and S(,) =
_
1
1
(.) sin ., J.
are continuous on (o. o) if
_
1
1
[(.)[ J. < o.
17. Suppose is continuously differentiable on a. o), lim
x!1
(.) = 0, and
_
1
a
[
0
(.)[ J. < o.
Show that the functions
C(,) =
_
1
a
(.) cos ., J. and S(,) =
_
1
a
(.) sin ., J.
are continuous for all , = 0. Give an example showing that they need not be
continuous at , = 0.
18. Evaluate J(,) and use Theorem 11 to evaluate 1:
(a) J(,) =
_
1
0
J.
1 ,
2
.
2
, , = 0; 1 =
_
1
0
tan
1
a. tan
1
b.
.
J.,
a, b > 0
25
(b) J(,) =
_
1
0
.
y
J., , > 1; 1 =
_
1
0
.
a
.
b
log .
J., a, b > 1
(c) J(,) =
_
1
0
e
xy
cos . J., , > 0
1 =
_
1
0
e
ax
e
bx
.
cos . J., a, b > 0
(d) J(,) =
_
1
0
e
xy
sin . J., , > 0
1 =
_
1
0
e
ax
e
bx
.
sin . J., a, b > 0
(e) J(,) =
_
1
0
e
x
sin ., J.; 1 =
_
1
0
e
x
1 cos a.
.
J.
(f) J(,) =
_
1
0
e
x
cos ., J.; 1 =
_
1
0
e
x
sin a.
.
J.
19. Use Theorem 11 to evaluate:
(a)
_
1
0
(log .)
n
.
y
J., , > 1, n = 0, 1, 2,. . . .
(b)
_
1
0
J.
(.
2
,)
nC1
J., , > 0, n = 0, 1, 2, . . . .
(c)
_
1
0
.
2nC1
e
yx
2
J., , > 0, n = 0, 1, 2,. . . .
(d)
_
1
0
.,
x
J., 0 < , < 1.
20. (a) Use Theorem 11 and integration by parts to show that
J(,) =
_
1
0
e
x
2
cos 2., J.
satises
J
0
2,J = 0.
(b) Use part (a) to show that
J(,) =
_

2
e
y
2
.
21. Show that
_
1
0
e
x
2
sin 2., J. = e
y
2
_
y
0
e
u
2
Ju.
(Hint: See Exercise 20.)
22. State a condition implying that
C(,) =
_
1
a
(.) cos ., J. and S(,) =
_
1
a
(.) sin ., J.
26
are n times differentiable on for all , = 0. (Your condition should imply the
hypotheses of Exercise 16.)
23. Suppose is continuously differentiable on a. o),
_
1
a
[(.
k
(.))
0
[ J. < o. 0 _ k _ n.
and lim
x!1
.
n
(.) = 0. Show that if
C(,) =
_
1
a
(.) cos ., J. and S(,) =
_
1
a
(.) sin ., J..
then
C
.k/
(,) =
_
1
a
.
k
(.) cos ., J. and S
.k/
(,) =
_
1
a
.
k
(.) sin ., J..
0 _ k _ n.
24. Differentiating
J(,) =
_
1
1
cos
,
.
J.
under the integral sign yields

_
1
1
1
.
sin
,
.
J..
which converges uniformly on any nite interval. (Why?) Does this imply that
J is differentiable for all ,?
25. Show that Theorem 11 and induction imply Eq. (30).
26. Prove Theorem 12.
27. Show that if J(s) =
_
1
0
e
sx
(.) J. converges for s = s
0
, then it converges
uniformly on s
0
. o). (Whats the difference between this and Theorem 13?)
28. Prove: If is continuous on 0. o) and
_
1
0
e
s
0
x
(.) J. converges, then
lim
s!s
0
C
_
1
0
e
sx
(.) J. =
_
1
0
e
s
0
x
(.) J..
(Hint: See the proof of Theorem 4.5.12, p. 273.)
29. Under the assumptions of Exercise 28, show that
lim
s!s
0
C
_
1
r
e
sx
(.) J. =
_
1
r
e
s
0
x
(.) J.. r > 0.
27
30. Suppose is continuous on 0. o) and
J(s) =
_
1
0
e
sx
(.) J.
converges for s = s
0
. Show that lim
s!1
J(s) = 0. (Hint: Integrate by parts.)
31. (a) Starting from the result of Exercise 18(d), let b o and invoke Exer-
cise 30 to evaluate
_
1
0
e
ax
sin .
.
J.. a > 0.
(b) Use (a) and Exercise 28 to show that
_
1
0
sin .
.
J. =

2
.
32. (a) Suppose is continuously differentiable on 0. o) and
[(.)[ _ Me
s
0
x
. 0 _ . _ o.
Show that
G(s) =
_
1
0
e
sx

0
(.) J.
converges uniformly on s
1
. o) if s
1
> s
0
. (Hint: Integrate by parts.)
(b) Show from part (a) that
G(s) =
_
1
0
e
sx
.e
x
2
sin e
x
2
J.
converges uniformly on j. o) if j > 0. (Notice that this does not follow
from Theorem 6 or 8.)
33. Suppose is continuous on 0. o),
lim
x!0C
(.)
.
exists, and
J(s) =
_
1
0
e
sx
(.) J.
converges for s = s
0
. Show that
_
1
s
0
J(u) Ju =
_
1
0
e
s
0
x
(.)
.
J..
28
10 Answers to selected exercises
5. (b) If (.. ,) = 1,, for , = 0 and (.. 0) = 1, then
_
b
a
(.. ,) J. does not
converge uniformly on 0. J| for any J > 0.
6. (a), (d), and (e) converge uniformly on (o. j| Lj. o) if j > 0; (b), (c), and (f)
converge uniformly on j. o) if j > 0.
17. Let C(,) =
_
1
1
cos .,
.
J. and S(,) =
_
1
1
sin .,
.
J.. Then C(0) = o and
S(0) = 0, while S(,) = ,2 if , = 0.
18. (a) J(,) =

2[,[
; 1 =

2
log
a
b
(b) J(,) =
1
, 1
; 1 = log
a 1
b 1
(c) J(,) =
,
,
2
1
; 1 =
1
2
b
2
1
a
2
1
(d) J(,) =
1
,
2
1
; 1 = tan
1
b tan
1
a
(e) J(,) =
,
,
2
1
; 1 =
1
2
log(1 a
2
)
(f) J(,) =
1
,
2
1
; 1 = tan
1
a
19. (a) (1)
n
n(, 1)
n1
(b) 2
2n1
_
2n
n
_
,
n1=2
(c)
n
2,
nC1
(log ,)
2
(d)
1
(log .)
2
22.
_
1
1
[.
n
(.)[ J. < o
24. No; the integral dening J diverges for all ,.
31. (a)

2
tan
1
a
29