4100 AWL/Thomas_ch08p553-641 8/20/04 10:08 AM Page 619

8.8 Improper Integrals 619

8.8 Improper Integrals

Up to now, definite integrals have been required to have two properties. First, that the do-
main of integration [a, b] be finite. Second, that the range of the integrand be finite on this
domain. In practice, we may encounter problems that fail to meet one or both of these con-
ditions. The integral for the area under the curve y = sln xd>x 2 from x = 1 to x = q is
an example for which the domain is infinite (Figure 8.17a). The integral for the area under
the curve of y = 1> 1x between x = 0 and x = 1 is an example for which the range of
the integrand is infinite (Figure 8.17b). In either case, the integrals are said to be improper
and are calculated as limits. We will see that improper integrals play an important role
when investigating the convergence of certain infinite series in Chapter 11.

y
y

y 0.2 y
ln2x y
1
x 兹x

0.1 1
Area
2
x x
0 1 2 3 4 5 6 0 1

x (a) (b)

(a) FIGURE 8.17 Are the areas under these infinite curves finite?
y
Infinite Limits of Integration

Area
2e –b/2  2

Consider the infinite region that lies under the curve y = e -x>2 in the first quadrant
(Figure 8.18a). You might think this region has infinite area, but we will see that the
natural value to assign is finite. Here is how to assign a value to the area. First find the area
x A(b) of the portion of the region that is bounded on the right by x = b (Figure 8.18b).
b b b
(b) Asbd = e-x>2 dx = -2e-x>2 d = -2e-b>2 + 2
L0 0

FIGURE 8.18 (a) The area in the first Then find the limit of A(b) as b : q
quadrant under the curve y = e -x>2 is
(b) an improper integral of the first type. lim Asbd = lim s -2e-b>2 + 2d = 2.
b: q b: q

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch08p553-641 8/20/04 10:08 AM Page 620

620 Chapter 8: Techniques of Integration

The value we assign to the area under the curve from 0 to q is

q b
e-x>2 dx = lim e-x>2 dx = 2.
L0 b: q L0

DEFINITION Type I Improper Integrals

Integrals with infinite limits of integration are improper integrals of Type I.
1. If ƒ(x) is continuous on [a, q d, then
q b
ƒsxd dx = lim ƒsxd dx.
La b: q La

2. If ƒ(x) is continuous on s - q , b], then

b b
ƒsxd dx = lim ƒsxd dx.
L- q a: - q La

3. If ƒ(x) is continuous on s - q , q d, then

q c q
ƒsxd dx = ƒsxd dx + ƒsxd dx,
L- q L- q Lc
where c is any real number.
In each case, if the limit is finite we say that the improper integral converges and
that the limit is the value of the improper integral. If the limit fails to exist, the
improper integral diverges.

It can be shown that the choice of c in Part 3 of the definition

q
is unimportant. We can
evaluate or determine the convergence or divergence of 1- q ƒsxd dx with any convenient
choice.
Any of the integrals in the above definition can be interpreted as an area if ƒ Ú 0 on
the interval of integration. For instance, we interpreted the improper integral in Figure 8.18
as an area. In that case, the area has the finite value 2. If ƒ Ú 0 and the improper integral
diverges, we say the area under the curve is infinite.

y
EXAMPLE 1 Evaluating an Improper Integral on [1, q d
0.2 y
ln2x Is the area under the curve y = sln xd>x2 from x = 1 to x = q finite? If so, what is it?
x

0.1 Solution We find the area under the curve from x = 1 to x = b and examine the limit
as b : q . If the limit is finite, we take it to be the area under the curve (Figure 8.19). The
x area from 1 to b is
0 1 b
b b b
dx = csln xd a- x b d - a- x b a x b dx
ln x 1 1 1 Integration by parts with
FIGURE 8.19 The area under this curve u = ln x, dy = dx>x 2,
L1 x L1
2
1 du = dx>x, y = -1>x.
is an improper integral (Example 1).
b
- cx d
ln b 1
= -
b 1

ln b 1
= - - + 1.
b b

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch08p553-641 8/20/04 10:08 AM Page 621

8.8 Improper Integrals 621

The limit of the area as b : q is

q b
ln x ln x
dx = lim dx
L1 b: L
2
x q 1 x2

= lim c- - + 1d
ln b 1
b: q b b

= - c lim d - 0 + 1
ln b
b: q b

= - c lim d + 1 = 0 + 1 = 1.
1>b
l’Hôpital’s Rule
b: q 1

Thus, the improper integral converges and the area has finite value 1.

EXAMPLE 2 Evaluating an Integral on s - q , q d

Evaluate
q
dx
.
L- q 1 + x2
Solution According to the definition (Part 3), we can write
q 0 q
dx dx dx
= + .
L- q L- q 1 + x L0
2 2
1 + x 1 + x2
HISTORICAL BIOGRAPHY Next we evaluate each improper integral on the right side of the equation above.
Lejeune Dirichlet 0 0
dx dx
(1805–1859) = lim
L- q 1 + x a: - q L
2
a 1 + x2
0
= lim tan-1 x d
a: - q a

lim stan-1 0 - tan-1 ad = 0 - a- b =

p p
=
a: - q 2 2
q b
dx dx
= lim
L0 1 + x2 b: q L 2
0 1 + x
y b
= lim tan-1 x d
b: q 0

1 p p
y
= lim stan-1 b - tan-1 0d = - 0 =
1  x2 Area
␲ b: q 2 2
Thus,
x q
0 dx p p
= + = p.
NOT TO SCALE L- q 1 + x2 2 2
FIGURE 8.20 The area under this curve Since 1>s1 + x2 d 7 0, the improper integral can be interpreted as the (finite) area
is finite (Example 2). beneath the curve and above the x-axis (Figure 8.20).

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch08p553-641 8/20/04 10:08 AM Page 622

622 Chapter 8: Techniques of Integration

q
dx
The Integral
L1 xp
The function y = 1>x is the boundary between the convergent and divergent improper
integrals with integrands of the form y = 1>x p . As the next example shows, the improper
integral converges if p 7 1 and diverges if p … 1.

EXAMPLE 3 Determining Convergence

q
For what values of p does the integral 11 dx>x p converge? When the integral does con-
verge, what is its value?

Solution If p Z 1,
b b
d = a p - 1 - 1b .
dx x -p + 1 1 1 1
p = sb -p + 1 - 1d =
L1 x -p + 1 1 1 - p 1 - p b
Thus,
q b
dx dx
= lim
L1 xp b: q L
p
1 x

1
= lim c a p - 1 - 1b d = • p - 1
1 1 , p 7 1
b: q 1 - p b
q, p 6 1
because

= e
1 0, p 7 1
lim p-1
b: q b q, p 6 1.
Therefore, the integral converges to the value 1>sp - 1d if p 7 1 and it diverges if
p 6 1.
If p = 1, the integral also diverges:
q q
dx dx
= x
L1 xp L1
b
dx
= lim x
b: q L1

= lim ln x D 1
b
b: q

= lim sln b - ln 1d = q .
b: q

Integrands with Vertical Asymptotes

Another type of improper integral arises when the integrand has a vertical asymptote—an
infinite discontinuity—at a limit of integration or at some point between the limits of
integration. If the integrand ƒ is positive over the interval of integration, we can again
interpret the improper integral as the area under the graph of ƒ and above the x-axis
between the limits of integration.

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch08p553-641 8/20/04 10:08 AM Page 623

8.8 Improper Integrals 623

y Consider the region in the first quadrant that lies under the curve y = 1> 1x from
x = 0 to x = 1 (Figure 8.17b). First we find the area of the portion from a to 1
y
1 (Figure 8.21).
兹x
1 1
= 21x d = 2 - 21a
dx
Area
2  2兹a La 1x a

Then we find the limit of this area as a : 0 +:

= lim+ A 2 - 21a B = 2.
1 1
dx
lim+
a:0 La 1x a:0

x
The area under the curve from 0 to 1 is finite and equals
0 a 1
1 1
dx dx
= lim+ = 2.
L0 1x a:0 La 1x
FIGURE 8.21 The area under this curve
is
1
a b dx = 2 ,
1
lim+ DEFINITION Type II Improper Integrals
a: 0 La 2x
Integrals of functions that become infinite at a point within the interval of inte-
an improper integral of the second kind. gration are improper integrals of Type II.
1. If ƒ(x) is continuous on (a, b] and is discontinuous at a then
b b
ƒsxd dx = lim+ ƒsxd dx.
La c:a Lc
2. If ƒ(x) is continuous on [a, b) and is discontinuous at b, then
b c
ƒsxd dx = lim- ƒsxd dx.
La c:b La
3. If ƒ(x) is discontinuous at c, where a 6 c 6 b, and continuous on
[a, cd ´ sc, b], then

b c b
ƒsxd dx = ƒsxd dx + ƒsxd dx.
La La Lc
In each case, if the limit is finite we say the improper integral converges and that
the limit is the value of the improper integral. If the limit does not exist, the inte-
gral diverges.

In Part 3 of the definition, the integral on the left side of the equation converges if both in-
tegrals on the right side converge; otherwise it diverges.

EXAMPLE 4 A Divergent Improper Integral

Investigate the convergence of
1
1
dx.
L0 1 - x

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch08p553-641 8/20/04 10:08 AM Page 624

624 Chapter 8: Techniques of Integration

y
1 Solution The integrand ƒsxd = 1>s1 - xd is continuous on [0, 1) but is discontinuous
y

1x at x = 1 and becomes infinite as x : 1- (Figure 8.22). We evaluate the integral as

dx = lim- C -ln ƒ 1 - x ƒ D 0
b
1 b
lim-
b:1 L0 1 - x b:1

= lim- [-ln s1 - bd + 0] = q .
b:1

1 The limit is infinite, so the integral diverges.

EXAMPLE 5 Vertical Asympote at an Interior Point

x Evaluate
0 b 1
3
dx
.
FIGURE 8.22 The limit does not exist: L0 sx - 1d2>3
1 b
a b dx = lim-
1 1 Solution The integrand has a vertical asymptote at x = 1 and is continuous on [0, 1)
dx = q .
L0 1 - x b:1 L0 1 - x
and (1, 3] (Figure 8.23). Thus, by Part 3 of the definition above,
The area beneath the curve and above the 3 1 3
x-axis for [0, 1) is not a real number dx dx dx
= + .
(Example 4). L0 sx - 1d2>3 L0 sx - 1d2>3 L1 sx - 1d2>3
Next, we evaluate each improper integral on the right-hand side of this equation.
1 b
dx dx
= lim-
L0 sx - 1d2>3 b:1 L0 sx - 1d2>3

= lim- 3sx - 1d1>3 D 0

b
y
b:1

1
= lim- [3sb - 1d1>3 + 3] = 3
y
b:1
(x  1)2/3
3 3
dx dx
= lim+
L1 sx - 1d2>3 c:1 Lc sx - 1d2>3

= lim+ 3sx - 1d1>3 D c

3
c:1
1

= lim+ C 3s3 - 1d1>3 - 3sc - 1d1>3 D = 32

3
2
c:1
x We conclude that
0 b c 3
1 3
dx
= 3 + 32
3
2.
FIGURE 8.23 Example 5 shows the L0 sx - 1d2>3
convergence of
3 EXAMPLE 6 A Convergent Improper Integral
1
dx = 3 + 32
3
2,
L0 sx - 1d2>3 Evaluate
q
so the area under the curve exists (so it is a x + 3
dx.
real number). L2 sx - 1dsx 2 + 1d

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch08p553-641 8/20/04 10:08 AM Page 625

8.8 Improper Integrals 625

Solution
q b
x + 3 x + 3
dx = lim dx
L2 L2 sx - 1dsx + 1d
2 2
sx - 1dsx + 1d b: q

b
a b dx
2 2x + 1
= lim - 2 Partial fractions
b: q L2 x - 1 x + 1

= lim C 2 ln sx - 1d - ln sx 2 + 1d - tan-1 x D 2
b
b: q
sx - 1d2 b
= lim cln 2
- tan-1 x d Combine the logarithms.
b: q x + 1 2

sb - 1d2
= lim cln a b - tan-1 b d - ln a b + tan-1 2
1
b: q b2 + 1 5

p
= 0 - + ln 5 + tan-1 2 L 1.1458
2
Notice that we combined the logarithms in the antiderivative before we calculated the
limit as b : q . Had we not done so, we would have encountered the indeterminate form
lim s2 ln sb - 1d - ln sb 2 + 1dd = q - q .
b: q
The way to evaluate the indeterminate form, of course, is to combine the logarithms, so we
would have arrived at the same answer in the end.
Computer algebra systems can evaluate many convergent improper integrals. To eval-
uate the integral in Example 6 using Maple, enter
7 ƒ:= sx + 3d>ssx - 1d * sx ¿2 + 1dd;
Then use the integration command
7 intsƒ, x = 2..infinityd;
Maple returns the answer
1
- p + ln s5d + arctan s2d.
2
To obtain a numerical result, use the evaluation command evalf and specify the number of
digits, as follows:
7 evalfs%, 6d;
The symbol % instructs the computer to evaluate the last expression on the screen, in this
case s -1>2dp + ln s5d + arctan s2d. Maple returns 1.14579.
Using Mathematica, entering
In [1]:= Integrate [sx + 3d>ssx - 1dsx ¿2 + 1dd, 5x, 2, Infinity6]
returns
-Pi
Out [1]= + ArcTan [2] + Log [5].
2
To obtain a numerical result with six digits, use the command “N[%, 6]”; it also yields
1.14579.

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch08p553-641 9/2/04 10:37 AM Page 626

626 Chapter 8: Techniques of Integration

y EXAMPLE 7 Finding the Volume of an Infinite Solid

2 The cross-sections of the solid horn in Figure 8.24 perpendicular to the x-axis are circular
disks with diameters reaching from the x-axis to the curve y = e x, - q 6 x … ln 2.
y ⫽ ex Find the volume of the horn.
1

Solution The area of a typical cross-section is

b
2
Asxd = psradiusd2 = p a yb = e 2x .
0 1 p
ln 2 x 2 4
FIGURE 8.24 The calculation in We define the volume of the horn to be the limit as b : - q of the volume of the portion
Example 7 shows that this infinite horn from b to ln 2. As in Section 6.1 (the method of slicing), the volume of this portion is
has a finite volume. ln 2 ln 2 ln 2
e dx = e 2x d
p 2x p
V = Asxd dx =
Lb Lb 4 8 b

p p
= se ln 4 - e 2b d = s4 - e 2b d.
8 8
As b : - q , e 2b : 0 and V : sp>8ds4 - 0d = p>2. The volume of the horn is p>2.

EXAMPLE 8 An Incorrect Calculation

Evaluate
3
dx
.
L0 x - 1
Solution Suppose we fail to notice the discontinuity of the integrand at x = 1, inte-
rior to the interval of integration. If we evaluate the integral as an ordinary integral we
get
3 3
= ln ƒ x - 1 ƒ d = ln 2 - ln 1 = ln 2.
dx
L0 x - 1 0

This result is wrong because the integral is improper. The correct evaluation uses limits:
3 1 3
dx dx dx
= +
L0 x - 1 L0 x - 1 L1 x - 1
where

= lim- ln ƒ x - 1 ƒ D 0
1 b
dx dx b
= lim-
L0 x - 1 b:1 L0 x - 1 b:1

= lim- sln ƒ b - 1 ƒ - ln ƒ -1 ƒ d
b:1
= lim- ln s1 - bd = - q . 1 - b : 0 + as b : 1-
b:1
1 3
Since 10 dx>sx - 1d is divergent, the original integral 10 dx>sx - 1d is divergent.

Example 8 illustrates what can go wrong if you mistake an improper integral for an
b
ordinary integral. Whenever you encounter an integral 1a ƒsxd dx you must examine the
function ƒ on [a, b] and then decide if the integral is improper. If ƒ is continuous on [a, b],
it will be proper, an ordinary integral.

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch08p553-641 8/20/04 10:08 AM Page 627

8.8 Improper Integrals 627

y
Tests for Convergence and Divergence
1 When we cannot evaluate an improper integral directly, we try to determine whether it
2
y
e –x
converges or diverges. If the integral diverges, that’s the end of the story. If it converges,
we can use numerical methods to approximate its value. The principal tests for conver-
gence or divergence are the Direct Comparison Test and the Limit Comparison Test.
(1, e –1)
EXAMPLE 9 Investigating Convergence
y
e –x q 2
Does the integral 11 e -x dx converge?
x
0 1 b
Solution By definition,
2 q b
FIGURE 8.25 The graph of e -x lies 2 2
e -x dx = lim e -x dx.
below the graph of e -x for x 7 1 L1 b: q L1
(Example 9).
We cannot evaluate the latter integral directly because it is nonelementary. But we can
b 2
show that its limit as b : q is finite. We know that 11 e -x dx is an increasing function of
b. Therefore either it becomes infinite as b : q or it has a finite limit as b : q . It does
2
not become infinite: For every value of x Ú 1 we have e -x … e -x (Figure 8.25), so that
b b
2
e -x dx … e -x dx = -e -b + e -1 6 e -1 L 0.36788.
L1 L1
Hence
q b
2 2
e -x dx = lim e -x dx
L1 b: q L1

converges to some definite finite value. We do not know exactly what the value is except
that it is something positive and less than 0.37. Here we are relying on the completeness
property of the real numbers, discussed in Appendix 4.
2
The comparison of e -x and e -x in Example 9 is a special case of the following test.

HISTORICAL BIOGRAPHY
Karl Weierstrass THEOREM 1 Direct Comparison Test
(1815–1897) Let ƒ and g be continuous on [a, q d with 0 … ƒsxd … gsxd for all x Ú a. Then
q q
1. ƒsxd dx converges if gsxd dx converges
La La
q q
2. gsxd dx diverges if ƒsxd dx diverges.
La La

The reasoning behind the argument establishing Theorem 1 is similar to that in

Example 9.
If 0 … ƒsxd … gsxd for x Ú a, then
b b
ƒsxd dx … gsxd dx, b 7 a.
La La

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch08p553-641 8/20/04 10:08 AM Page 628

628 Chapter 8: Techniques of Integration

From this it can be argued, as in Example 9, that

q q
fsxd dx converges if gsxd dx converges.
La La
Turning this around says that
q q
gsxd dx diverges if fsxd dx diverges.
La La

EXAMPLE 10 Using the Direct Comparison Test

q
sin2 x
(a) dx converges because
L1 x2
q
sin2 x 1 1
0 … … 2 on [1, q d and dx converges. Example 3
L1
2
x x x2
q
1
(b) dx diverges because
L1 2x - 0.1
2

q
1 1 1
Ú x on [1, q d and x dx diverges. Example 3
2x - 0.1
2
L1

THEOREM 2 Limit Comparison Test

If the positive functions ƒ and g are continuous on [a, q d and if
ƒsxd
lim = L, 0 6 L 6 q,
x: q gsxd
then
q q
ƒsxd dx and gsxd dx
La La
both converge or both diverge.

A proof of Theorem 2 is given in advanced calculus.

Although the improper integrals of two functions from a to q may both converge,
this does not mean that their integrals necessarily have the same value, as the next example
shows.

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch08p553-641 8/20/04 10:08 AM Page 629

8.8 Improper Integrals 629

EXAMPLE 11 Using the Limit Comparison Test

Show that
q
dx
L1 1 + x2
q
converges by comparison with 11 s1>x 2 d dx. Find and compare the two integral values.

Solution The functions ƒsxd = 1>x 2 and gsxd = 1>s1 + x 2 d are positive and continu-
ous on [1, q d. Also,

ƒsxd 1>x 2 1 + x2
lim = lim 2
= lim
x: q gsxd x: q 1>s1 + x d x: q x2

= lim a
1
+ 1b = 0 + 1 = 1,
x: q x2
q q
dx dx
y a positive finite limit (Figure 8.26). Therefore, converges because
L1 1 + x2 L1 x2
1 y
12 converges.
x The integrals converge to different values, however.
q
dx 1
= = 1 Example 3
L1 2 1
2
x -
y
1
1  x2 and
q b
dx dx
= lim
L1 b: L
2 2
x 1 + x q 1 1 + x
0 1 2 3
p p p
= lim [tan-1 b - tan-1 1] = - =
b: q 2 4 4
FIGURE 8.26 The functions in
Example 11.
EXAMPLE 12 Using the Limit Comparison Test
Show that
q
3
dx
L1 ex + 5
converges.
q q
Solution From Example 9, it is easy to see that 11 e -x dx = 11 s1>e x d dx converges.
Moreover, we have

1>e x
= lim a + b = ,
ex + 5 1 5 1
lim = lim x
x: q
x
3>se + 5d x: q 3e x: q 3 3e x 3

a positive finite limit. As far as the convergence of the improper integral is concerned,
3>se x + 5d behaves like 1>e x .

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley

4100 AWL/Thomas_ch08p553-641 8/20/04 10:08 AM Page 630

630 Chapter 8: Techniques of Integration

Types of Improper Integrals Discussed in This Section

INFINITE LIMITS OF INTEGRATION: TYPE I INTEGRAND BECOMES INFINITE: TYPE II
1. Upper limit 4. Upper endpoint

q b
ln x ln x 1 b
dx = lim dx dx dx
L1 L1 x
2 2 = lim-
x b: q
L0 sx - 1d2>3 b:1 L0 sx - 1d2>3

y
y 1
y

y
ln2x (x  1)2/3
x 1

x x
0 1 0 1 3

2. Lower limit 5. Lower endpoint

0 0
dx dx
= lim 3 3
L- q 1 + x a: - q L dx dx
2
a 1 + x2 = lim+
L1 sx - 1d2>3 d:1 Ld sx - 1d2>3

y
y
y
1
1 (x  1)2/3
y
1 2
1x 1

x
0 x
0 1 3

3. Both limits
6. Interior point
q 0 c
dx dx dx
= lim + lim
L- q b: - q L c: q L
3 1 3
1 + x2 b 1 + x
2
0 1 + x
2
dx dx dx
= +
L0 sx - 1d2>3
L0 sx - 1d2>3
L1 sx - 1d2>3

y
y
y
1
1 1
y
(x  1)2/3
1  x2
1

x
0 x
0 1 3

Copyright © 2005 Pearson Education, Inc., publishing as Pearson Addison-Wesley