Vous êtes sur la page 1sur 19

Section 8.

8
C08S08.001: The integral converges because
  z

−3/2 −1/2 2 2 √
x dx = lim − 2x = √ − lim √ = 2 .
2 z→∞
2 2 z→∞ z

Computer algebra programs generally have little difficulty with improper integrals. The Maple V version
5.1 command

int(x∧(−3/2), x=2..infinity);

and the Mathematica 3.0 command

Integrate[ x∧(−3/2), { x, 2, Infinity } ];



both immediately produce the response 2 . In Derive 2.56 (which is menu-driven), the commands

Author [Enter]
x∧(−3/2) [Enter]
Calculus [Enter]
Integrate [Enter]
Lower Limit 2 Upper Limit inf [Enter]

produce the display


 ∞
x−3/2 dx;
2

then the command Evaluate produces the immediate result 2.

C08S08.002: This improper integral diverges:


 ∞  z
x−2/3 dx = lim 3x1/3 = −3 + lim 3z 1/3 = +∞.
1 z→∞ z→∞
1

 4  4
−3/2 −1/2 2 2
C08S08.003: Divergent: x dx = lim+ 2x = − √ + lim+ √ = +∞.
0 z→0 z 4 z→0 z
 8  8
−2/3
C08S08.004: Convergent: x dx = lim+ 3x 1/3
= 6 − 0 = 6.
0 z→0 z

 ∞  z
1
C08S08.005: Divergent: dx = lim ln(x + 1) = − ln 2 + lim ln(z + 1) = +∞.
1 x+1 z→∞
1
z→∞

  z

−1/2
√ √
C08S08.006: Divergent: (x + 1) dx = lim 2 x+1 = −4 + lim x + 1 = +∞.
3 z→∞ z→∞
3

 ∞  ∞
2 2
C08S08.007: Convergent: (x − 1)−3/2 dx = lim −2(x − 1)−1/2 = √ − lim √ = 1.
5 z→∞
5 4 z→∞ z−1

1
  z
4 √ √
C08S08.008: Convergent: (4 − x)−1/2 dx = lim −2(4 − x)1/2 =2· 4 −2· 0 = 4.
0 z→4− 0

 9  z
−3/2 −1/2 2 2
C08S08.009: Divergent: (9 − x) dx = lim 2(9 − x) = − √ + lim √ = +∞.
0 z→9− 0 9 z→9 −
9−z
 3  z
1 1
C08S08.010: Divergent: dx = lim = +∞.
0 (x − 3)2 z→3− 3 − x 0

 −2  −2
1 1 1 1 1
C08S08.011: Convergent: dx = lim − = − + lim =− .
−∞ (x + 1)3 z→−∞ 2(x + 1)2 z 2 z→−∞ 2(z + 1)2 2
  0
0
1 √ √ √
C08S08.012: Divergent: √ dx = lim −2 4 − x = −2 · 4 + lim 2 4 − z = +∞.
−∞ 4−x Z→−∞
z
z→−∞

 8  z  
−1/3 3 2/3 3 3 2/3 3 9
C08S08.013: Convergent: x dx = lim x = − + lim z =− +6= .
−1 z→8− 2 −1 2 z→8− 2 2 2
 4  4
1
C08S08.014: Convergent: dx = lim 3(x + 4)1/3
= 6 − 0 = 6.
−4 (x + 4)2/3 z→−4+ z

 ∞  z
−1/3 3
C08S08.015: Divergent: (x − 1) dx = lim (x − 1)2/3 = +∞.
2 z→∞ 2 2

C08S08.016: Convergent:
 ∞  0  ∞
x −3/2
dx = 2
x(x + 4) dx + x(x2 + 4)−3/2 dx
−∞ (x2 + 4)3/2 −∞ 0
 0  w
1 1 1 1
= lim −√ + lim − √ = − + 0 − 0 + = 0.
z→−∞ 2
x +4 z
w→∞ 2
x +4 0 2 2

C08S08.017: Divergent:

 ∞  0  ∞  0  w
x x x 1 2 1 2
dx = dx + dx = lim ln(x + 4) + lim ln(x + 4) ,
−∞ x2 + 4 −∞ x2 + 4 0 x2 + 4 z→−∞ 2
z
w→∞ 2
0

and neither of the last two limits exists.


 ∞  z
−(x+1) −(x+1) 1
C08S08.018: Convergent: e dx = lim −e = 0 + e−1 = .
0 z→∞
0 e
 √  
1
exp ( x ) √  1
C08S08.019: Convergent: √ dx = lim+ 2 exp x = 2e − 2 = 2(e − 1).
0 x z→0 z

C08S08.020: Divergent:

 2  1  2  z  2
x x x 1 1
dx = dx + dx = lim ln |x2
− 1| + lim ln |x2
− 1| ,
0 x2 − 1 0 x2 − 1 1 x2 − 1 z→1− 2 0 w→1+ 2 w

and neither of the last two limits exists (both are −∞).

2
 ∞  z
3x + 1 −3x 1
C08S08.021: Convergent: xe−3x dx = lim − e = . To find the antiderivative:
0 z→∞ 9 0 9

Let u=x and dv = e−3x dx.


1
Then du = dx and v = − e−3x . Hence
3
 
−3x 1 1 −3x 1 1 3x + 1 −3x
xe dx = − xe−3x + e dx = − xe−3x − e−3x + C = − e + C.
3 3 3 9 9
 2  2
1 2x 1 4 1
C08S08.022: Convergent: e2x dx = lim e = e − 0 = e4 ≈ 27.2990750166.
−∞ z→−∞ 2 z 2 2
 ∞  z  
  1   1 1
C08S08.023: Convergent: x exp −x2 dx = lim − exp −x2 =0− − = .
0 z→∞ 2 0 2 2
 ∞   ∞
  0    
C08S08.024: Given: I = |x| exp −x2 dx = (−x) exp −x2 dx + x exp −x2 dx.
−∞ −∞ 0

Now substitute u = −x, x = −u, and dx = −du in the second of the three integrals, noting that −∞ < x  0
corresponds to 0  u < +∞. Thus

  ∞
0    
I= (−u) exp −u2 du + x exp −x2 dx
∞ 0
 ∞  ∞  ∞
      1
= u exp −u2 du + x exp −x2 dx = 2 x exp −x2 dx = 2 · = 1
0 0 0 2
by the result in Problem 23. Thus the improper integral I converges.
 ∞  z
1 π
C08S08.025: Convergent: dx = lim arctan x = lim arctan z = .
0 1 + x2 z→∞
0
z→∞ 2
 ∞  z
x 1 2 1
C08S08.026: Divergent: dx = lim ln(1 + x ) = lim ln(1 + z 2 ) = +∞.
0 1 + x2 z→∞ 2
0
z→∞ 2

 2nπ
C08S08.027: cos x dx = 0 if n is a positive integer, but
0
 2nπ+(π/2)  2nπ+(π/2)
cos x dx = sin x = 1.
0 0
 z
Therefore lim cos x dx does not exist; this improper integral diverges.
z→∞ 0

C08S08.028: If n is a positive integer, then


   2nπ
2nπ 2nπ
1 − cos 2x 1
2
sin x dx = dx = (x − sin x cos x) = nπ.
0 0 2 2 0

 z
Therefore the given improper integral diverges because lim sin2 x dx = +∞.
z→∞ 0

3
 ∞  z
ln x 1 1
C08S08.029: Divergent: dx = lim (ln x)2 = lim (ln z)2 = +∞.
1 x z→∞ 2
1
z→∞ 2

 109
ln x
(But the divergence is relatively slow: dx ≈ 214.7268734744.)
1 x
 ∞  z
1
C08S08.030: Divergent: dx = lim ln(ln x) = +∞.
2 x ln x z→∞
2

 109
1
(But the divergence is extremely slow: dx ≈ 3.3977699432.)
2 x ln x
 ∞  z
1 1 1
C08S08.031: Convergent: 2
dx = lim − = .
2 x(ln x) z→∞ ln x 2 ln 2
 ∞  z  
ln x 1 + ln x 1 + ln z 1
C08S08.032: Convergent: dx = lim − = 1 − lim = 1 − lim = 1.
1 x2 z→∞ x 1
z→∞ z z→∞ z

The next-to-last equality results from an application of l’Hôpital’s rule. The antiderivative was obtained by
integration by parts:

1
Let u = ln x and dv = dx;
x2
1 1
then du = dx and v = − .
x x
 
ln x 1 1 1 1
Therefore dx = − ln x + dx = − − ln x + C.
x2 x x2 x x
  π/2
π/2
cos x √
C08S08.033: Convergent: √ dx = lim 2 sin x = 2 − 0 = 2.
0 sin x z→0+ z

 π/2  z
sin x 3
C08S08.034: Divergent: dx = lim = +∞.
0 (cos x)3/2 z→π/2− (cos x)1/3 0

C08S08.035: First note that—by l’Hôpital’s rule—

1
ln z z
lim z ln z = lim = lim+ = lim+ (−z) = 0.
z→0+ z→0+ 1 z→0 1 z→0
− 2
z z
Therefore the given improper integral converges because
 1  1
ln x dx = lim −x + x ln x = −1 + lim (z − z ln z) = −1 + 0 − 0 = −1.
0 z→0+ z z→0+

See Example 1 of Section 8.3 for the computation of the antiderivative.


 1  1  
ln x 1 1
C08S08.036: Divergent: dx = lim (ln x)2 = − lim (ln z)2 = −∞.
0 x z→0+ 2 z z→0+ 2

1 1 1
C08S08.037: Let u = ln x and dv = 2
dx. Then du = dx; choose v = − . Hence
x x x

4
 
ln x 1 1 1 1
dx = − ln x + dx = − ln x − + C.
x2 x x2 x x
Therefore the given improper integral diverges because
 1  1  
ln x 1 1 + ln z
dx = lim − (1 + ln x) = −1 + lim = −∞.
0 x2 z→0+ x z z→0+ z

(If you use l’Hôpital’s rule on the last limit, you’ll get the wrong answer!)

C08S08.038: Let u = e−x and dv = cos x dx. Then du = −e−x dx; choose v = sin x. Then
 
K = e cos x dx = e sin x + e−x sin x dx.
−x −x

Now let u = e−x and dv = sin x dx. Then du = −e−x dx; choose v = − cos x. Therefore

K = e−x sin x − e−x cos x − e−x cos x dx;

2K = e−x (sin x − cos x) + 2C;



1
K = e−x cos x dx = e−x (sin x − cos x) + C.
2

Therefore the given improper integral converges because


 ∞  z  
1 −x 1 e−z (sin z − cos z) 1
e−x cos x dx = lim e (sin x − cos x) = + lim = .
0 z→∞ 2 0 2 z→∞ 2 2

C08S08.039: The first improper integral diverges because

       
1
1 1
1 1 x 1 z
dx = − dx = lim+
ln = − ln 2 − lim ln = +∞.
0 x + x2 0 x x+1 z→0 x+1 z z→0+ z+1

The second integral converges because


 ∞  
1 1 z
dx = − ln + lim ln = 0 + ln 2 = ln 2.
1 x + x2 2 z→∞ z+1

C08S08.040: First, the method of partial fractions yields

1 A B Cx + D 1 1
= + 2+ 2 = 2− 2 .
x2 +x4 x x x +1 x x +1

The first improper integral diverges because


 1  1
1 1
dx = lim − − arctan x = +∞.
0 x2 + x4 z→0+ x z

The second integral converges because


 ∞  z
1 1 π π 4−π
dx = lim − − arctan x =− +1+ = ≈ 0.2146018366.
1 x2 + x4 z→∞ x 1 2 4 4

5
C08S08.041: Let x = u2 . Then dx = 2u du, and thus
  
1 2u 2 √
dx = du = du = 2 arctan u + C = 2 arctan x + C.
x + x3/2
1/2 x + x2 1+u 2

Hence both improper integrals converge, because


  
1
1 √ 1 π π
1/2 3/2
dx = lim+ 2 arctan x =2· −2·0=
0 x +x z→0 z 4 2

and
 ∞  
1 √ z π π π
dx = lim 2 arctan x =2· −2· = .
1 x1/2 + x3/2 z→∞
1 2 4 2

C08S08.042: Let x = u3 ; then dx = 3u2 du. Hence

  

1 3u2 3 1/3
dx = du = du = 3 arctan u + C = 3 arctan x + C.
x + x4/3
2/3 u2 + u4 1 + u2
Therefore
 1 
1
1 3π
2/3 4/3
dx = lim 3 arctan x1/3
= 3 arctan(1) − 3 arctan(0) =
0 x +x z→0 +
z 4

and
 ∞ 
z
1 3π 3π 3π
2/3 4/3
dx = lim 3 arctan x1/3
= − = .
1 x +x z→∞
1 2 4 4

Thus both improper integrals converge.

C08S08.043: If k = 1, then
 1  1
1
dx = lim ln x = +∞.
0 xk z→0+ z

If k
= 1, then
 1
 1  1   
 if k < 1,
x1−k
1 z 1−k
1−k
x−k dx = lim = − lim =
0 x→0+ 1−k z 1−k z→0 + 1 −k 

+∞ if k > 1.

Therefore the given improper integral converges precisely when k < 1.

C08S08.044: If k = 1, then
 ∞  z
1
dx = lim ln x = +∞.
1 xk z→∞
1

If k
= 1, then

6

  z    +∞ if k < 1,

x 1−k
1 1−k
z 
x−k dx = lim =− + lim =
1 z→∞ 1−k 1 1−k z→∞ 1 − k 
 1
if k > 1.
k−1

Therefore the given improper integral converges exactly when k > 1.

C08S08.045: If k = −1, then


 1  1  1  
ln x 1 1
k
x ln x dx = dx = lim+ (ln x)2
= − lim+ (ln z) = −∞.
2
0 0 x z→0 2 z z→0 2

If k
= −1, then use integration by parts:

Let u = ln x and dv = xk dx;

1 xk+1
then du = dx and v = .
x k+1
Hence

 
xk+1 ln x xk
xk ln x dx = − dx
k+1 k+1

xk+1 ln x xk+1 xk+1


= − + C = [(k + 1)(ln x) − 1] + C.
k+1 (k + 1)2 (k + 1)2
Therefore if k
= −1, then

 1  1
xk+1
I= xk ln x dx = lim+ {(k + 1)(ln x) − 1}
0 z→0 (k + 1)2 z
 k+1 
1 z
=− − lim {(k + 1)(ln z) − 1} .
(k + 1)2 z→0+ (k + 1)2

Suppose first that k < −1. Write k = −1 −  where  > 0. Then


 1
ln x
I= dx.
0 x1+

But 0 < x  1, so x1+ = x · x  x. Therefore

ln x ln x
 .
x1+ x
Therefore I diverges to −∞ if k  −1.
Now suppose that k > −1. Recall that
 k+1 
1 z
I=− − lim {(k + 1)(ln z) − 1} .
(k + 1)2 z→0+ (k + 1)2

Now

7
lim z k+1 ln z = lim+ z ln z (where  > 0)
z→0+ z→0
   
ln z 1 1 1 z
= lim = lim =− lim =− lim = 0.
z→0+ z − z→0+ −z − −1 z  z→0+ z − z→0+ 
Also,

lim z k+1 = lim z = 0.


z→0+ + z→0

1
Therefore I = − if k > −1.
(k + 1)2
 1
1
Consequently xk ln x dx converges exactly when k > −1, and its value for such k is − .
0 (k + 1)2
 ∞
1
C08S08.046: Let f (k) = dx. Then
1 x(ln x)k
 ∞  z
ln x 1 1
f (−1) = dx = lim (ln x)2 = lim (ln z)2 = +∞.
1 x z→∞ 2
1
z→∞ 2

If k < −1, then write k = −1 −  where  > 0. Then


 ∞  ∞
(ln x)1+ ln x
f (k) = f (−1 − ) = dx  dx = +∞.
1 x 1 x

Therefore the given improper integral diverges to +∞ if k  −1.


If k > −1, then −k < 1. So
 ∞  z
(ln x)−k (ln x)−k+1 (ln z)1−k
f (k) = dx = lim = lim . (1)
1 x z→∞ −k + 1 1 z→∞ 1 − k

If k < 1, then −1 < k < 1, so 0 < 1 − k < 2. Therefore the integral in Eq. (1) diverges to +∞ in this case.
If k = 1, then
 ∞  ∞
1
f (1) = dx = ln(ln x) = +∞.
1 x ln x 1

Finally, if k > 1, then 1 − k < 0. Therefore, by Eq. (1),


 z
(ln x)1−k 1 1
f (k) = lim = lim − lim = +∞.
z→∞ 1−k 1
z→∞ (1 − k)(ln z)k−1 w→1+ (1 − k)(ln w)k−1
 ∞
1
In conclusion, dx diverges to +∞ for all real numbers k.
1 x(ln x)k
 ∞
C08S08.047: Given: Γ(t) = xt−1 e−x dx for t > 0. Thus
0
 ∞
Γ(t + 1) = xt e−x dx.
0

Let u = xt and dv = e−x dx. Then du = txt−1 dx and v = −e−x . Thus

8
 ∞  ∞  
t −x t−1 −x zt
Γ(t + 1) = − (x e ) +t x e dx = − lim z + 0 + tΓ(t) = tΓ(t).
0 z→∞ e
x=0

To evaluate the last limit, we used the result in Problem 61 of Section 7.2.

C08S08.048: Prove that Γ(n + 1) = n! for every positive integer n.


Proof: By Example 5 of Section 8.8, Γ(1 + 1) = Γ(2) = 1!, so the theorem holds when n = 1. Assume
that Γ(k + 1) = k! for some integer k  1. Then

Γ(k + 2) = (k + 1)Γ(k + 1) = k!(k + 1) = (k + 1)!.

Thus whenever the theorem holds for the positive integer k, it also holds for k + 1. Therefore, by induction,
Γ(n + 1) = n! for every positive integer n. 

C08S08.049: The area is


 ∞  z
1
A= dx = lim ln x = lim ln z = +∞.
1 x z→∞
1
z→∞

C08S08.050: The volume is


  π z


π π
V = 2
dx = lim − = π − lim = π.
1 x z→∞ x 1 z→∞ z

dy 1
C08S08.051: Because = − 2 , we have arc length element
dx x
 1/2 √
1 x4 + 1
ds = 1 + 4 dx = dx.
x x2
Therefore the area of the surface of Gabriel’s horn satisfies the inequality

 √   √ 
∞ z
1 x4 + 1 x4 + 1
S= 2π · · dx = 2π lim dx
1 x x2 z→∞ 1 x3
  √    z 
z
x4
 2π lim dx = 2π lim ln x = +∞.
z→∞ 1 x3 z→∞
1

C08S08.052: First,
 ∞  z  
1+x 1 2 1 2
dx = lim arctan x + ln(1 + x ) = lim arctan z + ln(1 + z ) = +∞.
0 1 + x2 z→∞ 2 0
z→∞ 2

 ∞
1+x
Therefore dx diverges. But
−∞ 1 + x2
 t  t
1+x 1 2
lim dx = lim arctan x + ln(1 + x )
t→∞ −t 1 + x2 t→∞ 2 −t
 
1 1
= lim arctan t − arctan(−t) + ln(1 + t2 ) − ln(1 + t2 ) = lim 2 arctan t = π.
t→∞ 2 2 t→∞

9
This technique of assigning a “plausible” (or, perhaps, “balanced”) value to a divergent improper integral
of the form
 ∞
f (x) dx (1)
−∞

is the evaluation of the so-called Cauchy principal value of the integral in Eq. (1).

C08S08.053: We will use the definition


 ∞
Γ(t) = xt−1 e−x dx
x=0

for t > 0 and the result of Problem 48 to the effect that Γ(n + 1) = n! if n is a positive integer. For fixed
nonnegative integers m and n, let
 1
J(m, n) = xm (ln x)n dx.
0

Then
 1
1 0!(−1)0 n!(−1)n
J(m, 0) = xm dx = = =
0 m+1 (m + 1)1 (m + 1)n+1

where n = 0. Therefore

n!(−1)n
J(m, n) =
(m + 1)n+1

if n = 0 and m  0. Assume that

k!(−1)k
J(m, k) =
(m + 1)k+1

for some integer k  0 and all integers m  0. Then


 1
J(m, k + 1) = xm (ln x)k+1 dx.
0

k+1 m
Let u = (ln x) and dv = x dx. Then

(k + 1)(ln x)k xm+1


du = dx and v= .
x m+1
Therefore
 1  1
(ln x)k+1 xm+1 k+1
J(m, k + 1) = − xm (ln x)k dx.
m+1 0 m +1 0

The value of the evaluation bracket is zero because

 k+1  k+1
(ln x)k+1 ln x k+1
lim = lim+ = lim+ = 0k+1 = 0.
x→0+ 1 x→0 x(m+1)/(k+1) x→0 (m + 1)x(m+1)/(k+1)
xm+1
Therefore

10
k+1 k+1 k!(−1)k (−1)k+1 (k + 1)!
J(m, k + 1) = − · J(m, k) = − · = .
m+1 m + 1 (m + 1)k+1 (m + 1)k+2
Therefore, by induction,
 1
n!(−1)n
xm (ln x)n dx =
0 (m + 1)n+1
for all positive integers m and n.

C08S08.054: As we saw in Section 8.8, the present value of 10 + t thousand dollars t years in the future
is (10 + t)e−t/10 thousand dollars if the interest rate is 10%. So the total present value of the perpetuity is
 ∞
P = (10 + t)e−t/10 dt.
0

Let u = 10 + t and dv = e−t/10 dt. Then du = dt and we may choose v = −10e−t/10 . Thus
 ∞  ∞  ∞
−t/10 −t/10 −t/10
P = −10(10 + t)e + 10 e dt = 100 − 100 e = 200;
0 0 0

that is, $200,000.

C08S08.055: We assume that a > 0. A short segment of the rod “at” position x  0 and of length dx
has mass δ dx, and thereby exerts on m the force

Gmδ
dx.
(x + a)2
Therefore the total force exerted by the rod on m is
 ∞  ∞
Gmδ Gmδ Gmδ
F = dx = − = .
0 (x + a)2 x+a 0 a

What if a = 0? Then the total force is


 ∞  ∞
Gmδ Gmδ Gmδ
F = dx = − = lim = +∞.
0 x2 x 0 z→0 + z

You will obtain the same result if a < 0.

C08S08.056: A small segment of the rod “at” location y and with length dy has mass δ dy, so exerts on
m the force

Gmδ
dy.
a2 + y 2
The vertical components of all such forces cancel (perhaps the reason that Cauchy developed the idea of the
principal value of certain improper integrals—see the solution to Problem 52), so the total force exerted by
the rod on the mass m is the sum (i.e., integral) of the horizontal components of such forces:
 ∞  ∞  ∞
Gmδ cos θ a 1
F = 2 + y2
dy = Gmδ 2 + y 2 )3/2
dy = 2Gmaδ 2 + y 2 )3/2
dy.
y=−∞ a −∞ (a 0 (a

Then a trigonometric substitution, or integral formula 52 of the endpapers, yields

11
 ∞    
y 2Gmδ y 2Gmδ 1 2Gmδ
F = 2Gmaδ  = lim  = lim  = .
a2 a2 + y 2 a y→∞ a2 + y2 a y→∞ 1+ (a/y)2 a
0

C08S08.057: In the integral


   ∞
1
Γ = x−1/2 e−x dx
2 0

we substitute x = u2 , so that dx = 2u du. Then


   ∞  ∞  ∞
1 1 −u2
e−u du = 2 e−x dx.
2 2
Γ = ·e · 2u du = 2
2 0 u 0 0

 ∞
1√
e−x dx =
2
C08S08.058: We will use π . The volume of revolution around the x-axis is
0 2
 ∞  ∞
  2 2  
V = π exp −x dx = π exp −2x2 dx.
0 0
√ √  √ 
Let u = x 2 , so that x = u/ 2 and dx = 1/ 2 du. Then
 ∞
√ √ √ √
2  2 π 2 π π 3/2 2
V =π exp −u du = · = ≈ 1.968701243.
0 2 2 2 4

C08S08.059: The volume of revolution around the y-axis is


  
∞     ∞
V = 2πx exp −x2 dx = π − exp −x2 = π.
0 0

C08S08.060:  We  will use the result of Problem 47, Γ(x + 1) = xΓ(x) for all x > 0, and Eq. (9), which

tells us that Γ 12 = π . If n = 1, then
     
3 1 1 1 1√ 1 · 3 · 5 · · · (2n − 1) √
Γ =Γ 1+ = Γ = π = π.
2 2 2 2 2 2n

So the desired result holds when n = 1. Assume that


 
1 1 · 3 · 5 · · · (2k − 1) √
Γ k+ = π
2 2k

for some integer k  1. Then


     
1 1 1
Γ k+1+ = k+ Γ k+
2 2 2

2k + 1 1 · 3 · 5 · · · (2k − 1) √ 1 · 3 · 5 · · · (2k − 1) · (2k + 1) √


= · π = π.
2 2k 2k+1
Therefore, by induction, the desired result holds for every integer n  1.

C08S08.061: Part (a): Suppose that k > 1 and let

12
 ∞  
I= xk exp −x2 dx.
0
 
Let u = xk−1 and dv = x exp −x2 dx. Then

1  
du = (k − 1)xk−2 dx and v=− exp −x2 .
2
Therefore
 ∞  ∞
xk−1   k−1  
I= − exp −x2 + xk−2 exp −x2 dx.
2 0 2 0

The evaluation bracket is zero by Problem 62 of Section 7.2. This concludes the proof in part (a).
Part (b): Now suppose that n is a positive integer. If n = 1, then
 ∞  ∞  
    1 1
xn−1 exp −x2 dx = exp −x2 dx = Γ
0 0 2 2

by Eq. (9) of the text. Assume that for some integer k  1,


 ∞  
  1 k
xk−1
exp −x 2
dx = Γ
0 2 2

and in addition that


 ∞   1
m
xm−1 exp −x2 dx = Γ
0 2 2

for every integer m such that 1  m  k. Then


 ∞  ∞
  k−1  
x exp −x
k 2
dx = xk−2 exp −x2 dx
0 2 0
     
k−1 1 k−1 1 k−1 1 k+1
= · ·Γ = Γ 1+ = Γ .
2 2 2 2 2 2 2
Therefore, by induction,
 ∞   1
n
xn−1 exp −x2 dx = Γ
0 2 2

for every positive integer n.

C08S08.062: Answer:
 ∞  ∞
500000 −3t/50 500000
P = 10000e−3t/50 dt = − e = ;
0 3 0 3

that is, $166,666.67.


x
C08S08.063: Substitute t = √ in the given interval. This routinely gives Eq. (10).
2

C08S08.064: Begin with

13
 
u
erf √
2
as given in Eq. (15), then use the substitution of the previous solution. This leads routinely to Eq. (16).

C08S08.065: We defined
 b
Ib = x5 e−x dx
0

and asked Mathematica 3.0 to evaluate Ib for increasing large positive values of b. The results:

b Ib (approximately)

10 111.949684454516

20 119.991370939137

30 119.999997291182

40 119.999999999505

50 119.999999999999933

60 119.99999999999999999258

(Results were essentially the same with Maple V version 5.1.) It seems very likely that

1
k = lim Ib = 2.
b→∞ 60
Indeed, because

x5 e−x dx = −(x5 + 5x4 + 20x3 + 60x2 + 120x + 120)e−x + C,

it follows that

b5 + 5b4 + 20b3 + 60b2 + 120b + 12


Ib = 120 − → 120
eb
as b → +∞, and this proves that k = 2.

C08S08.066: We defined
π
k(b) =  b
(1)
sin x
dx
0 x

and asked Mathematica 3.0 to evaluate k(b) for various increasing large values of b. Here are the results:

b k(b) (approximately)

10 1.8944114397618596

14
20 2.0291357941544347

40 1.9795980537404538

80 1.9980480443011534

160 1.9922776386463579

320 2.0035967916498463

640 2.0012584221020270

1280 1.9998026067119984

2560 1.9995416983267274

5120 2.0001739448256683

10240 1.9999973452978946

As in the previous problem, it seems likely that k = 2. Unlike the previous problem, it appears that k(b)
oscillates around the limiting value k = 2 as b → +∞. This is plausible given the oscillatory behavior of the
denominator in Eq. (1). Methods of Laplace transforms can be used to prove that
 ∞
sin x π
dx =
0 x 2

(exactly), and—given that result—this proves that k = 2.

C08S08.067: We defined
π
√ 
kb =
b
1
2 · dx
0 x2 +2

and asked Mathematica 3.0 to evaluate kb for various large increasing values of b. Here are the results:

b kb (approximately)

10 2.1964469705919963

20 2.0941114933683164

40 2.0460327426191207

80 2.0227515961720504

160 2.0113173416565956

320 2.0056428162376127

640 2.0028174473301223

1280 2.0014077338319199

2560 2.0007036195035935

15
5120 2.0003517479044710

10240 2.0001758584911264

There is good evidence that k = 2. In fact, because


 b    b  
1 1 x 1 b
Ib = 2
dx = √ arctan √ = √ arctan √ ,
0 x +2 2 2 0 2 2

it follows that
√ √
2 2 1 π 1
lim Ib = ·√ · = ,
b→∞ π π 2 2 2

and therefore k = 2 exactly.

C08S08.068: Because
 ∞  ∞  ∞  ∞
1 − e−3x 1 1
dx  dx  dx = ln x = +∞,
0 x 0 x 1 x 1

no such integer k can exist.

C08S08.069: We defined

π
k(b) =  b 
e· exp −x2 cos 2x dx
0

and asked Mathematica 3.0 to evaluate k(b) for some large positive values of b. The results:

k(10) = k(100) ≈ 3.54490770181103205459633496668229

to the number of decimal places shown. We conclude that there is no such integer k.

C08S08.070: We defined

π
k(b) =
√  ∞
2 · sin(x2 ) dx
0

and asked Mathematica 3.0 to approximate k(b) for some large positive values of b. The results:

k(102 ) ≈ 1.9849201201159858,

k(106 ) ≈ 2.0000012629663674, and

k(109 ) ≈ 2.0000000001888944.

It seems plausible that k = 2. Can you prove it?

C08S08.071: In the notation of this section, we have µ = 100 and σ = 15. For part (a), we let a = 10/σ.
Then

16
 √ 
1 a   2
P = √ exp − 12 x2 dx = erf ≈ 0.4950149249.
2π −a 3

Thus just under 50% of students have IQs between 90 and 110. For part (b), we let a = 25/σ. Then
 ∞   
1  1 2 1 5
P = √ exp − 2 x dx = 1 − erf √ ≈ 0.0477903523.
2π a 2 3 2

Thus just under 5% of students have IQs of 125 or higher.

C08S08.072: Let µ = 69 and σ = 3. For part (a), we let

2 2 3
a = − = − and b = = 1.
σ 3 σ
Then we compute
   √ 
1 b   1 1 1 2
P = √ exp − 12 x2 dx = erf √ + erf ≈ 0.5888522085.
2π a 2 2 2 3

Thus just under 59% of adult males are between 5 7 and 6 . Part (b): Let a = 7/σ = 7/3. Then
 ∞   
1  1 2 1 7
P = √ exp − 2 x dx = 1 − erf √ ≈ 0.0098153286.
2π a 2 3 2

Therefore just under 1% of adult males are 6 4 or taller.



C08S08.073: Here we take p = q = 1/2 and N = 1000; we have µ = N p = 450 and σ = N pq = 15. So
we let a = 25/σ = 5/3 and compute
 a  
1   5
P = √ exp − 12 x2 dx = erf √ ≈ 0.9044192954.
2π −a 3 2

Thus there is over a 90% probability of 425 to 475 heads. For part (b), we let a = 50/σ = 10/3 and compute
 ∞   √ 
1  1 2 1 5 2
P = √ exp − 2 x dx = 1 − erf ≈ 0.0004290603.
2π a 2 3

Thus there is only a probability of 0.04%—les than one chance in 2000—of 500 or more heads.

C08S08.074: We are given p = 3/5, q = 2/5, and N = 600. Then µ = N p = 360 and σ = N pq = 12.
Part (a): We take a = 15/σ = 5/4 and compute
 a  
1  1 2 5
P = √ exp − 2 x dx = erf √ ≈ 0.7887004527.
2π −a 4 2

Thus the probability of obtaining 345 to 375 heads is just under 0.79 (sometimes called a 79% probability).
For part (b), we let b = 10/15 and compute
 −b  √  
1  1 2 1 2
P = √ exp − 2 x dx = 1 − erf ≈ 0.2524925375.
2π −∞ 2 3

Thus there is just over a 25% probability of 350 or fewer heads.

17
√ √
C08S08.075: Here we√have p = q = 1/2 and N = 50. Then µ = N p = 25 and σ = N pq = 5/ 2 . Part
(a): We let a = 5/σ = 2 and compute
 ∞
1   1
P = √ exp − 12 x2 dx = [1 − erf(1)] ≈ 0.0786496035.
2π a 2
Thus there is slightly more than a 1 in 13 chance of passing by pure guessing. Part (b): We let a = 10/σ
and compute
 ∞
1   1
P = √ exp − 12 x2 dx = [1 − erf(2)] ≈ 0.0023388675.
2π a 2
Thus there is less than 1 chance in 425 of making a C by pure guessing.

C08S08.076: Here we
√ take p = 0.99, q = 0.01, and N = 500.
√ In the notation of Section 7.8, µ = N P = 495
√ 3
and σ = N pq = 10 55 ≈ 2.22486. We let a = 5/σ = 1033 55 ≈ 2.24733 and compute
   
1 −a   1 5 √
P = √ exp − 12 x2 dx = 1 − erf 10 ≈ 0.0123093807.
2π −∞ 2 33
Therefore there is only about one chance in 81 that ten or more are defective in a batch of 500.

C08S08.077: Let p√= 0.55, q = 0.45, and N = 750. In the notation of Section 7.8, we have µ = N p = 412.5

and σ = N pq = 34 330 ≈ 13.62443. Now 59% of the N = 750 voters amounts to 442.5, so we let

442.5 − µ 4 √
a= = 330 ≈ 2.20183
σ 33
and evaluate
  
1 a   4 √
P = √ exp − 12 x2 dx = erf 165 ≈ 0.9723295720.
2π −a 33
Thus there is a 97.23% probability that between 51% and 59% will say that they are Democratic voters, and
thus that between 41% and 49% will say that they are Republican voters.

C08S08.078: For each integer n  0, let


 ∞
(ln x)n
In = dx.
1 x2
Then
 ∞  ∞
1 1
I0 = dx = − = 1 = 0!,
1 x2 x 1

and therefore In = n! if n = 0. Next, suppose that Ik = k! for some integer k  0. Then


 ∞
(ln x)k+1
Ik+1 = dx.
1 x2
Integrate by parts: Let

1
u = (ln x)k+1 , dv = dx.
x2
(k + 1)(ln x)k 1
Then du = dx, v=− .
x x

18
Thus
 ∞  ∞
(ln x)k+1 (ln x)k
Ik+1 = − + (k + 1) dx = 0 + (k + 1)Ik = (k + 1)Ik = (k + 1)(k!) = (k + 1)!.
x 1 1 x2
Therefore, by induction, In = n! for each integer n  0.

19

Vous aimerez peut-être aussi