Excerpt from "Calculus" ©2013 AoPS Inc.

www.artofproblemsolving.com
6.5. IMPROPER INTEGRALS

6.5 Improper integrals

Rb
As we’ve seen, we use the definite integral a f to compute the area of the region under the graph of y = f (x)
along the interval [a, b]. By definition, these integrals can only be used to compute areas of bounded regions.
In some situations, however, we are interested in unbounded regions—these are regions that extend “towards
infinity” in at least one direction. Yet, many unbounded regions still have finite area.
We start with a basic example of this phenomenon:

1
Problem 6.16: What is the area of the region bordered by the curve y = , the line x = 1, and the x-axis?
x2

Solution for Problem 6.16: We sketch a picture of this region at right. Notice that y
this region is unbounded: the region extends towards +1 as x grows large.
Even though this region is unbounded, we can attempt to determine its area. We
certainly can compute the area of the portion of the region to the left of x = b (for
any b > 1) as the definite integral
Z b
1
dx.
1 x2
1
y= x2
As b grows larger, we expect that the area under the curve on [1, b] approaches
the area of the entire region under the curve on [1, +1). Specifically, this area is
Z b
1
lim dx.
b!1 1 x2 1 x

The integral is easy to evaluate:

Z b b
1 1 1
dx = =1 .
1 x2 x 1 b
Thus, when we take the limit, we get that the area of the region is
Z b ✓ ◆
1 1
lim dx = lim 1 =1 0 = 1.
b!1 1 x2 b!1 b
2
Note the “paradox” here: even though the region is unbounded, it has finite area. Problem 6.16 suggests a
logical definition:

Definition: Let f be a continuous function and a 2 R such that (a, 1) ✓ Dom( f ). We define the improper
integral
Z 1 Z b
f (x) dx = lim f (x) dx,
a b!1 a

provided the limit is defined. If the limit is defined and is not ±1, we say that the improper integral converges.
Otherwise, we say that the improper integral diverges.

There’s an obviously similar definition for improper integrals in the other direction:

207

Copyrighted Material
 Excerpt from "Calculus" ©2013 AoPS Inc.
www.artofproblemsolving.com
CHAPTER 6. INFINITY

Definition: Let f be a continuous function and b 2 R such that ( 1, b) ✓ Dom( f ). We define the improper
integral
Z b Z b
f (x) dx = lim f (x) dx,
1 a! 1 a

provided the limit is defined. If the limit is defined and is not ±1, we say that the improper integral converges.
Otherwise, we say that the improper integral diverges.

Let’s generalize Problem 6.16:

Problem 6.17: Let r be a real number. Compute

Z 1
1
dx.
1 xr

Solution for Problem 6.17: By definition, we compute the improper integral by writing a limit. If r , 1, then we
have: Z b b
1 1
lim dx = lim .
b!1 1 xr b!1 (r 1)xr 1 1
This equals
✓ ◆
1 1
lim 1 .
b!1 r 1 br 1

1
If r > 1, then the term br 1
approaches 0 as b approaches 1. Thus, in this case, the improper integral converges
to r 11 .
1
If r < 1, then the term br 1
grows without bound as b approaches 1. Thus, the integral diverges. We might also
write Z 1
1
r
dx = 1 if r < 1.
1 x

Our original integration was not valid for r = 1, so we have to do that case separately:
Z b b
1
lim dx = lim (log x) = lim (log b).
b!1 1 x b!1
1
b!1

As b goes towards infinity, this grows without bound, so the integral diverges.
In summary: 8
Z 1 >
1 > r 11
< if r > 1,
dx = >
>
1 xr :diverges if r  1.
2
The next problem is another common example of an improper integral:
Z 1
Problem 6.18: Compute eax dx, where a is a real number.
0

Solution for Problem 6.18: We compute, for a , 0 (we’ll investigate a = 0 at the end):
Z b
1 1
lim eax dx = lim (eab 1) = lim (eab 1).
b!1 0 b!1 a a b!1

208

Copyrighted Material
 Excerpt from "Calculus" ©2013 AoPS Inc.
www.artofproblemsolving.com
6.5. IMPROPER INTEGRALS

If a is positive, then lim eab = 1, so the integral diverges. If a is negative, then lim eab = 0, so the integral equals
b!1 b!1
1
a. (Note Rthis is a positive number when a is negative, so this answer makes sense.) Finally, if a = 0, then the
1
integral is 0 1 dx, which clearly diverges.
Thus, the integral diverges for nonnegative exponents, and converges for negative exponents. 2
The result of Problem 6.18 is typically written as follows: if r > 0, then
Z 1
1
e rx dx = .
0 r

Problem 6.19: Suppose f and g are continuous functions on [a, 1) and f (x)  g(x) for all x a.
R1 R1
(a) Show that, if a f and a g both converge, then
Z 1 Z 1
f  g.
a a

R1
R1
(b) Show that if both functions are positive, and a
g converges, then a
f converges.
R1 R1
(c) Show that if both functions are positive, and a f diverges, then a g diverges.

Solution for Problem 6.19:

(a) For any b a, we have (g f )(x) 0 for all x 2 [a, b], thus
Z b
(g f )(x) dx 0.
a

Therefore, Z Z
b b
f (x) dx  g(x) dx,
a a
and since limits preserve non-strict inequalities, we conclude that
Z 1 Z b Z b Z 1
f (x) dx = lim f (x) dx  lim g(x) dx = g(x) dx.
a b!1 a b!1 a a

(b) Define a function Z x

F(x) = f (t) dt.
a
R1
Note that F is an increasing function (since f (x) 0 for all x a), and that a
f (x) dx = lim F(x), if this limit
x!1
exists. Also, since 0  f (x)  g(x) for all x a, we have
Z x Z x Z 1
0  F(x) = f (t) dt  g(t) dt  g(t) dt.
a a a
R1
Thus F is increasing and has an upper bound (namely, a g(t) dt, which by assumption converges), so by the
result of Problem 6.5, the limit Z 1
lim F(x) = f (t) dt
x!1 a
exists, so the integral converges.

209

Copyrighted Material
 Excerpt from "Calculus" ©2013 AoPS Inc.
www.artofproblemsolving.com
CHAPTER 6. INFINITY

(c) This is just the contrapositive statement to part (b), so there is nothing additional to prove.
2

WARNING!! We can only use the comparison tests in parts (b) and (c) of Problem 6.19 if
j both functions are positive.
R1 As Ra trivial example, if Rf (x) = 1 and g(x) = 0,
1 1
then for any a 2 R, a g = 0, so a g converges, but a f diverges.

Thus far in this section, we have looked at improper integrals that compute areas of regions that are unbounded
in the x-direction. There is another type of improper integral that occurs when the region that we are examining
is unbounded in the y-direction, as in the following example:
Z 1
1
Problem 6.20: Compute p dx.
0 x

Solution for Problem 6.20: Sketching the graph will immediately show the issue. We have y
1 1
lim+ p = 1. So the area under y = p is potentially infinite (and in fact the function is
x!0 x x
not even defined at 0).
We can do essentially the same thing we did for improper integrals with a limit of
y= p1
integration of ±1. We define x

Z 1 Z 1
1 1
p dx = lim+ p dx.
0 x c!0 c x
Note the “0+ ”—since we only care about the interval (0,1], we only care about what
happens to the right of 0. 1x

This integral is now easy to compute:

Z 1 1
1 p p
p dx = 2 x = 2 2 c.
c x c

As c ! 0+ , this approaches 2. Hence

Z 1
1
p dx = 2.
0 x
Once again, a seemingly infinite area turns out to be finite. 2
We can generalize the definition from Problem 6.20:

Definition: Suppose f is a function, continuous on (a, b], such that lim+ f (x) = ±1. We define the improper
x!a
integral
Z b Z b
f (x) dx = lim+ f (x) dx,
a c!a c

provided this limit is defined. If the limit is defined, we say that this improper integral converges, and if it is
undefined, we say that the improper integral diverges.

Of course, we can do the same thing if the function has a limit of ±1 at the “b” end of [a, b). (We will omit
writing out the formal definition.)

210

Copyrighted Material
 Excerpt from "Calculus" ©2013 AoPS Inc.
www.artofproblemsolving.com
6.5. IMPROPER INTEGRALS

Sidenote: Note that the above definition is consistent with our usual (non-improper) in-
Rb
↵ tegrals. In particular, if a f is defined, then by the Fundamental Theorem of
Calculus, the function Z b
g(x) = f (t) dt
x

is di↵erentiable, hence continuous, and thus

Z b Z b
f (t) dt = g(a) = lim+ g(x) = lim+ f (t) dt.
a x!a x!a x

We know that for regular (not improper) integrals, we can break them apart at any point into two separate
integrals. Specifically, if c 2 (a, b), then
Z b Z c Z b
f = f+ f.
a a c

This is also how we evaluate integrals that are improper at both ends, as in the following example:
Z 1
1
Problem 6.21: Compute dx for all r > 0 (or determine when it diverges).
0 xr

Solution for Problem 6.21: The correct thing to do with an integral that is improper at both ends is to split it
somewhere in the middle. For example, we can write
Z 1 Z 1 Z 1
1 1 1
dx = dx + dx.
0 xr 0 xr 1 xr

(We didn’t have to pick x = 1 as the point at whichR to split them, but it seems convenient since xr is nicely behaved
1
at x = 1.) We already know by Problem 6.17 that 1 x1r dx converges if and only if r > 1. The other integral is

Z 1 Z 1 1
! ✓ ◆
1 1 1 1 1
dx = lim dx = lim = lim 1 .
0 xr a!0+ a xr a!0+ (r 1)xr 1
a r 1 a!0+ ar 1

R1
1
If r > 1, then the fraction gets arbitrarily large, so the limit is infinite. Thus 0 xr
dx diverges for r > 1.
Hence our original doubly-improper integral is never convergent: the integral on (0, 1] diverges for r > 1, and
the integral on [1, 1) diverges for r  1. 2

Z b
Important: If (a, b) ✓ Dom( f ) and f (t) dt is improper at both ends of (a, b), then
a
Z b Z c Z x
f (t) dt = lim+ f (t) dt + lim f (t) dt,
a x!a x x!b c

for any c 2 (a, b).

As noted in the solution to Problem 6.21, it doesn’t matter at which point we break up the doubly-improper
integral.

211

Copyrighted Material
 Excerpt from "Calculus" ©2013 AoPS Inc.
www.artofproblemsolving.com
CHAPTER 6. INFINITY

Concept: We can break an integral apart as

Z b Z c Z b
f = f+ f
a a c

at any c 2 (a, b) that we choose. Thus, choose c to be as convenient as possible.

We will leave it as an exercise to prove this. Also, it is not correct to try to take a shortcut and deal with both
ends of a double-improper integral at once. In particular:

Z 1 Z a
WARNING!! f (x) dx is not the same as lim f (x) dx.
j 1 a!1 a

The correct way to evaluate an integral over all of R is to choose c 2 R, and then compute
Z 1 Z c Z 1 Z c Z b
f (x) dx = f (x) dx + f (x) dx = lim f (x) dx + lim f (x) dx.
1 1 c a! 1 a b!+1 c

We will leave it as an exercise to explore this further.

We also have to be a bit cautious when dealing with functions with domains that are not all of R. Integrals of
such functions might be improper but not immediately appear so. For example:
Z 3
1
Problem 6.22: Compute dx.
2 x2

Solution for Problem 6.22: If you weren’t paying close attention, you might do this:

Bogus Solution: Z 3 3
1 1 1 1 1
dx = = + = .
2 x2 x 2 3 2 6

We can’t do this, because the function is not defined at 0! To be a little more precise, the function x12 does not have
an antiderivative on the interval [ 2, 3], because it is not defined at x = 0, so we cannot apply the Fundamental
Theorem of Calculus.
In order to evaluate the integral, we need to break it up into a sum of two improper integrals at the point at
which the function is undefined: Z 3 Z 0 Z 3
1 1 1
2
dx = 2
dx + 2
dx.
2 x 2 x 0 x
As we saw in Problem 6.21, both of these diverge. Thus, the original integral itself diverges. 2
Z 1
dx
More generally, when computing something like , it might be tempting to say “ 1x is an odd function, so
1 x
the integral from 1 to 0 will cancel out the integral from 0 to 1, and thus the overall integral is 0.” This is also the
result that naive calculation will give:

Bogus Solution: Z 1 1
dx
= log |x| = log(1) log(1) = 0.
1 x 1

212

Copyrighted Material
 Excerpt from "Calculus" ©2013 AoPS Inc.
www.artofproblemsolving.com
REVIEW PROBLEMS

But this is not correct! The only way legally to evaluate this integral is to break it up into its improper parts.
Z 1 Z 0 Z 1
dx dx dx
= + .
1 x 1 x 0 x

Neither part converges, so the integral diverges.

Exercises
6.5.1 Compute the following improper integrals:
Z 1 Z 1 Z 1 Z 2
1 1 x2 1
(a) 2
dx (b) dx (c) xe dx (d) dx
3 (2x 1) 2 x(log x)2 0 0 4 x2
6.5.2
Z 1
1
(a) Compute dx.
0 1 + x2
Z 1
1
(b) Compute dx.
1 1 + x2
Z 1
6.5.3 Compute x2 e x
dx.
0

6.5.4 Show that it doesn’t matter at which point we break up a doubly-improper integral. Specifically, show that,
Rc Rb Rd Rb
for any c, d 2 (a, b), if a f and c f converge, then a f and d f also converge, and
Z c Z b Z d Z b
f+ f = f+ f.
a c a d

Hints: 230, 97
6.5.5?
Z 1 Z 1 Z a
(a) Show that if f (x) dx converges, then f (x) dx. Hints: 165 f (x) dx = lim
1 1 a!1 a
Z a
(b) Show that the converse of part (a) is not true; that is, it is possible that lim f (x) dx converges but that
a!1 a
Z 1
f (x) dx diverges. Hints: 82, 225
1

Review Problems
6.23 Compute the following:
1 3
x2 1 cos2 x 1 10x2 2x
(a) lim (b) lim (c) lim (Source: Rice)
x!1 log x x!0 x2 x!0 e4x2 1
sin at tan at
6.24 Suppose a and b are nonzero real numbers. Find lim and lim . Hints: 32
t!0 sin bt t!0 tan bt

6.25 Compute
Z 1 Z 2 Z 1
1 1
(a) e 2x dx (b) dx (c) dx Hints: 72
1 0 x3 1 x2 + 2x + 2

213

Copyrighted Material