amsppt

Improper Integrals

Assume that f : [a, ∞) −→ R is a function such that for any b ∈ (a, ∞), f is integrable on
[a, b]. If
Z b
lim f (x)dx.
b→∞ a

exists and is finite, then we say that the improper integral of f on (a, ∞) converges and
we write Z ∞ Z b
f (x)dx = lim f (x)dx.
a b→∞ a

Example. Let p > 0. If p 6= 1 then

Z b
I(b) = 1xp dx = x−p+1 −p + 1|b1 = 11 − p[b1−p − 1]
1

If p > 1 then I(b) −→ 1p − 1 as b −→ ∞ and therefore

Z ∞
1xp dx = 1p − 1, p > 1.
1

R∞
If 0 < p < 1, then I(b) −→ ∞ and the improper integral 1
1xp dx diverges. If p = 1 then
I(b) = lnx|b1 = lnb −→ ∞ as b −→ ∞.
Therefore we have the

Theorem 1 Z ∞
1xp dx = { 1p − 1if
1

p¿1
Z b
I(b) = e−ax dx = −1ae−ax |b0 = 1a(1 − e−ab ) −→ 1a as b −→ ∞.
0
R∞
Therefore, 0 e−ax dx converges to 1a if a > 0. It diverges if a < 0.
Similarly, we define
Z b Z b
f (x)dx = lim f (x)dx
−∞ a→−∞ a

if f is integrable on [a, b] for any a ≤ b and the above limits exist and is finite.
Also, we define
Z ∞ Z b Z c
f (x)dx = lim f (x)dx + lim f (x)dx
−∞ b→∞ c a→−∞ a

if f is integrable on any interval [a, b] and both limits above exist and are finite.

1
If a function f is unbounded and on (a,b] it is integrable on [a + , b] for any  > 0 and
Rb
lim→0 a+ f (x)dx exists and is finite, then we say that the improper integral of f over
(a, b] converges and we write
Z b Z b
f (x)dx = lim f (x)dx.
a →0 a+

Example. Let p > 0. We have

Z 1
I() = 1xp dx = { 11 − p[a − 1−p ]if p 6= 1


If 0 <  < 1 then I() −→ 11 − p as  −→ 0 and if 1 ≤ p then I() −→ ∞ as  −→ 0.

Therefore, we have

Theorem 2.
Z 1
1xp dx = { 11 − pif
0

0¡p¡1
If a function f is integrable on [a, b − ] for any  > 0 and unbounded on [a, b), then we
define Z a Z b−
f (x)dx = lim f (x)dx
b −→0 a

if the above limit exists and is finite.

Finally, we define
Z b Z c Z b−
f (x)dx = lim f (x)dx + lim f (x)dx
a −→0 a+ −→0 c

if f is integrable on any interval [a + , b − ] but unbounded on (a, c] and [c, b), c(a, b),
and if both limits above exist and are finite.

Example. The integral

Z 1
1(1 − x2 )1 4
−1

is improper since
f (x) = 1(1 − x2 )1 4

is unbounded on both intervals [0, 1) and (−1, 0]. We will show later that it converges.
Convergence Tests for Improper Integrals

2
Here we will discuss only infinite integrals of the form
Z ∞
f (x)dx
a

where the function

Ra f is integrable over any interval [a, b], b ≥ a. The tests for the integrals
of the form −∞ f (x)dx are reduced to the above case by the substitution t = −x. Also,
Rb
the substitution t = 1x − a or t = 1x − b reduces the improper integrals a f (x)dx to the
above case.

Theorem 1 (Cauchy criterion). Suppose that R ∞the function f is integrable over any
interval [0, b), b ≤ a. Then the improper integral a f (x)dx converges if and only if for
every  > 0 there exists N = N () such that
Z d
d ≥ c ≥ N =⇒ | f (x)dx| < .
c

* R∞
Remark. In other words, the Cauchy criterion says that a
f (x)dx converges if and
only if
Z d
f (x)dx −→ 0 as c −→ ∞.
c
R∞
Example. The infinite integral 1
dxx2 dx converges since
Z d
| dxx2 | = 1d − 1c ≤ 2c −→ 0 as c −→ ∞.
c

R∞
Example. The integral 1
dxx does not converge since for c ≥ 1 we have
Z c2
| dxx| = ln c2 − ln c = ln c −→ ∞ as c −→ ∞.
c
Rb
Proof. Supposef irstthattheintegralconverges, i.e.I(b) = a f (x)dx −→ I as b −→ ∞.
Then by the definition of the limit for any  > 0 there exists N = N () ≥ a such that

b ≥ N ⇒ |I(b) − I| < 2

Therefore, if d ≥ c ≥ N then
Z d
| f (x)dx| = |I(d) − I(c)| ≤ |I(d) − I| + |I(c) − I| < 2 + 2 = 
c

Conversely, we assume that (*) holds. The sequence of numbers I(k) is a Cauchy sequence.
Such sequences always converge. This is a fact following from the construction of the real

3
numbers. Therefore, there exists a real number I such that I(k) −→ I as k → ∞. Thus
for any b ≥ a we have

I(b) = [I(b) − I(k)] + I(k) −→ I if k ≥ b −→ ∞

since I(k) −→ I and I(k) − I(b) −→ 0 by (*).

Next we have

Theorem 2.R Assume that f (x) ≥ 0 for x ≥ a and that it is integrable on [a, b] for any
∞
b ≥ a. Then a f (x)dx converges if and only if there exists M such that
Z b
f (x)dx ≤ M f orany b ≥ a.
a

Proof. Since f (x) ≥ 0 the function

Z b
I(b) = f (x)dx, a ≥ b
a

is increasing. Therefore, I(b) converges as b −→ ∞ if and only if I(b) is bounded.

R∞
Example. The integral 1
1x2 + x + 1dx converges since
Z b Z b
2
1x + x + 1dx ≤ 1x2 dx = 1 − 1b ≤ 1 f orany b ≥ 1.
1 1

Theorem 3. (Comparison Test). Suppose that f (x), g(x)and h(x) are integrable on [a, b]
for any b ≥ a and that 0 ≤ g(x) ≤ 1|f (x)| ≤ h(x) for any x ≥ A, where A is some number
with A ≥ a. We have the following:
R∞ R∞
a) If a
h(x)dx converges, then a
f (x)dx converges.
R∞ R∞
b) If a
g(x)dx = ∞ and f (x) ≥ 0 then a
f (x)dx = ∞.
R∞
Example. The integral 1 sin xxdx converges since for x ≥ 1 we have |sin xx2 | ≤ 1x2
R∞
and 1 1x2 dx < ∞.
R∞√
Example. The integral 1
x2x + 1dx diverges since for x ≥ 1
√ √ √
x2x + 1 ≥ x2x + x = 131 x
R∞ √
and 1
131 x diverges.

4
 R∞
Proof. We will use the Cauchy criterion for g(x). Let  > 0. Since a
g(x)dx converge,
there exists N = N () ≥ a such that
Z d
d ≥ c ≥ N =⇒ g(x)dx < .
c

Then for d ≥ c ≥ N we have

Z d Z d Z d
| f (x)dx| ≤ |f (x)|dx ≤ g(x)dx < .
c c c
R∞
Therefore a
f (x)dx converges.

Example. (Gamma Function) Let s > 0 and

f (x) = xs−1 e−x , x ∈ (0, ∞).

For s ≥ 1 the function f (x) is bounded on any interval [0, a], a > 0. Since

xs−1 e−x e−12 = xs−1 e−12x −→ 0 as x → ∞

we have that there is A > 0 such that

xs−1 e−x e−12x ≤ 1, f or x ≥ A.

Therefore
0 ≤ xs−1 e−x ≤ e−12x , f or x ≥ A.
R∞
By the comparison test, the integral 0 xs−1 e−x dx converges to a finite number which we
denote by Γ(s), i.e. Z ∞
Γ(s) = xs−1 e−x dx, s ≥ 1.
0

If 0 < s < 1 then the function f (x) = xs−1 e−x is unbounded near 0. Since

0 ≤ xs−1 e−x ≤ 1x1−s , 0 < x ≤ 1

R1 R1
and 0 dxx1−s converges for 0 < s < 1 we have, by the comparison test, that 0 xs−1 e−x dx
converges.
R∞
If 0 < s < 1 and x ≥ 1 then 0 ≤ xs−1 e−x ≤ e−x and since 1 e−x dx converges, by
R∞
the comparison test we conclude that 1 xs−1 e−x dx converges. Therefore, we have that
R ∞ s−1 −x
0
x e dx converges for any s > 0 to a finite number which is denoted by Γ(s) and it
is called the Gamma function, i.e.
Z ∞
Γ(s) = xs−1 e−x dx, s > 0.
0

5
If s > 1, then by integrating by parts we obtain
Z ∞ Z ∞
−x s−1 −x ∞
Γ(s) = s−1
x d(−e ) = −x e |0 + (s − 1) xs−2 e−x dx
0 0

Thus, if s > n by integrating by parts repeatedly we obtain

Γ(s) = (s − 1)(s − 2) · · · (s − n)Γ(s − n)

In particular we obtain Γ(n + 1) = n(n − 1) · 1 · Γ(1) = n! since Γ(1) = 1.

2 R∞ 2
Example. Since e−x ≤ e−x for x > 1 the integral 0 e−x dx converges. It can be shown
that Z ∞
2 √
e−x dx = 12 π.
0

(See p. 371, no. 54.)

Theorem 4. (Limit Comparison Test) Suppose f (x) ≥ 0 and g(x) > 0 and that they
are integrable on any interval [a, b], b ≥ a. If

lim f (x)g(x) = l 6= 0
x→∞

then either both integrals

Z ∞ Z ∞
f (x)dx and g(x)dx
a a

converge or they both diverge.

R∞
Example. The integral 1
2x − 1x5 + x3 + xdx converges since

2x − 1x5 + x3 + 11x4 = 2x5 − x4 x5 + x3 + 1 @ >> x → ∞ > 2 6= 0

R∞
and since 1 1x4 dx < ∞.

Example. The integral Z ∞ p

1 x2 + 5x + 8dx
1

diverges since

p p p
1 x2 + 5x + 81x = x x2 + 5x + 8 = 1 1 + 5x + 8x2 @ >> x → ∞ >1 6= 0
R∞
and 1
1xdx = ∞.

6
Proof of Theorem 4. Since lim f (x)g(x) = l > 0, by the definition of the limit for
x→∞
 = l2 there exists A = A() such that

x ≥ A =⇒ |f (x)g(x) − l| < l2.

The last relation is written as

−l2 < f (x)g(x) − l < l2, x ≥ A

or

l2 < f (x)g(x) < 3l2, x ≥ A

or

l2g(x) < f (x) < 3l2g(x), x ≥ A

By Thoeorem 4 the last relation implies that

Z ∞ Z ∞
f (x)dx g(x)dx
a a

are either both converging or both diverging.

Example. None of the convergence tests we discussed above can be used to show that
Z ∞
sin xxdx
1

converges. This integral converges because of two reasons:

Rb
i) For any b ≥ 1, | 1
sin xdx| ≤ 2.

ii) The function g(x) = 1x is decreasing on any interval [1, b], b ≥ 1.

The convergence of the above integral follows from:

Theorem 5 (Dirichlet’s Test) If

i) f is continuous on [a, ∞) and there is M such that
Z b
| f (x)dx| ≤ M, f oranyb ≥ a
a

and ii) g is a decreasing-continuously differentiable function on [a, ∞) with g(x) → 0 as x →

∞, then

7
 Z ∞
f (x)g(x)dx
a
converges.

Remark. The condition that g is continuously differentiable can be dropped.To prove The-
orem 5, we need the version of the Second Mean Value Theorem for Integrals. Let
f (x) be continuous on [a, b] and g(x) be decreasing and continuously differentiable on [a, b].
Then there exists ξ ∈ [a, b] such that
Z b Z ξ Z b
f (x)g(x)dx = g(a) f (x)dx + g(b) f (x)dx.
a a ξ

Proof. Let Z x
F (x) = f (t)dt.
a

Then by the Fundamental Theorem of Calculus, we have F 0 (x) = f (x). Integrating by

parts, we obtain
Z b Z b Z b
f (x)g(x)dx = g(x)d(F (x)) = g(x)F (x)|ba − F (x)g 0 (x)dx.
a a a

Since −g 0 (x) ≥ 0 on [a, b] by the Mean Value Theorem for integrals we obtain that there
is ξ ∈ [a, b] such that
Z b Z b
0
F (x)(−g (x)]dx = F (ξ) (−g 0 (x))dx.
a a

Therefore Z b Z b
f (x)g(x)dx = g(b)F (b) − g(a)F (a) − F (ξ) g 0 (x)dx
a a

Proof of Theorem 5. Let d ≥ c ≥ a. By the last result there exists ξ ∈ [c, d] such that
Z d Z ξ Z d
f (x)g(x)dx = g(c) f (x)dx + g(d) f (x)dx
c c ξ

since

Z ξ Z ξ Z c Z ξ Z ξ
| f (x)dx| = | f (x)dx − f (x)dx| ≤ | f (x)dx| + | f (x)|dx ≤ 2M.
c a a a a

and similarly
Z d
| f (x)dx| ≤ 2M
ξ

8
we have that
Z d
| f (x)g(x)dx| ≤ g(c)2M + g(d)2M = 4M g(c)@ >> c → ∞ > 0
c
R∞
Therefore, by the Cauchy criterion a
f (x)g(x)dx converges.

Example. If p > 0 then the integral

Z ∞
sin xxp dx
1

converges since, if we let f (x) = sin x and g(x) = 1xp , then f and g satisfy the assumptions
of Theorem 5. In fact, for p > 1 it follows by the Comparison Test since
Z ∞
p p
|sin xx | ≤ 1x and 1xp dx < ∞.
1
R∞
Also, the integral 0 sin xxdx converges since for x ≥ 1 we can use the Dirichlet’s test
and since sin xx is continuous on [0, 1]. It can be shown (see p. 372, no. 55) that
Z ∞
sin xxdx = π2.
0

Remark We remark here that

Z ∞
| sin x|xdx = ∞.
0

In fact, for any k = 2, 3, 4, · · ·

Z ∞ Z Z 3π Z kπ
| sin x|xdx ≥ 2π| sin x|xdx + | sin x|xdx + · · · + |sinx|xdx
0 π 2π (k−1)π

Since Z j+1
1j > 1xdx
j

We have
Z ∞ Z 3 Z 4 Z k+1 Z k+1
| sin x|x ≥ 2π[ 1xdx + 1xdx + · · · + 1xdx] ≥ 2π 1xdx.
0 2 3 k 2

Thus Z ∞ Z ∞
| sin x|x ≥ 2π 1xdx = ∞.
0 2

9
 R∞
By the comparison
R test we have seen that if a
|f (x)|dx converges, then so does the
integral f (x)dx.
R∞
The last remark shows that the converse is not true. If a |f (x)|dx converges, then we
say that the improper integral of f is absolutely convergent.

Exercises.
R∞
1. Show that 0
sin(x2 )dx is convergent. (Hint: Let the substitution t = x2 .)
R∞
2. Show that the integral −1
(1 − x2 )−14 dx is convergent.
R∞
3. Compute the integral 1
x3 + x2 + 1x6 + x3 dx.

10