You are on page 1of 33

Section 5.

1 Cauchy’s Residue Theorem 101

Solutions to Exercises 5.1
1. Write
1+z 1
f(z) = = + 1.
z z
f has one simpe pole at z0 = 0. The Laurent series expansion of f(z) at z0 = 0 is already given.
The residue at 0 is the coefficient of 1z in the Laurent series a−1 . Thus a−1 = Res (f, 0) = 1.
5. We have one pole of order 3 at z0 = −3i.
Laurent series of f around z0 :
 3
z −1 1 3
= [(z + 3i) + (−3i − 1)]
z + 3i (z + 3i)3
1  
= 3
(z + 3i)3 + 3(z + 3i)2 (−1 − 3i) + 3(z + 3i)(−1 − 3i) + (−3i − 1)3
(z + 3i)
(−1 − 3i) (−1 − 3i) (−3i − 1)3
= 1+3 +3 +
z + 3i (z + 3i)2 (z + 3i)3
Thus a−1 = 3(−1 − 3i) = Res (−3i).
9. Write
z+1 1 z+1
f(z) = csc(πz) = .
z−1 sin πz z − 1
Simple poles at the integers, z = k, z 6= 1. For k 6= 1,
1 z +1
Res (f, k) = lim (z − k)
z→k sin πz z − 1
z+1 (z − k)
= lim lim
z→k z − 1 z→k sin πz
k+1 1
= lim (L’Hospital’s rule)
k − 1 z→k π cos πz
k+1 1 (−1)k k + 1
= = .
k − 1 π cos kπ π k−1
At z0 = 1, we have a pole of order 2. To simplify the computation of the residue, let’s rewrite f(z)
as follows:
1 z+1 1 (z − 1) + 2
=
sin πz z − 1 sin πz z−1
1 2
= +
sin πz (z − 1) sin πz
We have
1 2
Res (f, 1) = Res ( , 1) + Res ( , 1);
sin πz (z − 1) sin πz
1 1 1
Res ( , 1) = lim (z − 1) =− (Use l’Hospital’s rule.);
sin πz z→1 sin πz π

2 d 2(z − 1)
Res ( , 1) = lim
(z − 1) sin πz z→1 dz sin πz
sin πz − (z − 1)π cos πz
= 2 lim
z→1 (sin πz)2
π cos πz − π cos πz + (z − 1)π sin πz
= 2 lim =0
z→1 2π sin πz cos πz
102 Chapter 5 Residue Theory

So Res (f, 1) = − π1 .
13. The easiest way to compuet the integral is to apply Cauchy’s generalized formula with
z 2 + 3z − 1
f(z) = ,
z2 − 3
which is analytic inside and on C1(0). Hence
Z
z 2 + 3z − 1 1 2πi
2 − 3)
dz = 2πif(0) = 2πi( ) = .
C1 (0) z(z 3 3
Note that from this value, we conclude that
z 2 + 3z − 1 1
Res ( , 0) =
z(z 2 − 3) 3
because the integral is equal to 2πi times the residue at 0.
17. The function
1
f(z) =
z(z − 1)(z − 2) · · · (z − 10)
has simple poles at 0 and 1 inside C 32 (0). We have
1 1
Res (f(z), 0) = =
(0 − 1)(0 − 2) · · · (0 − 10) 10!
1 1
Res (f(z), 1) = =− .
1 (1 − 2) · · · (1 − 10) 9!
Hence
Z
dz
= 2πi ( Res (f(z), 0) + Res (f(z), 1))
C 3 (0) z(z − 1)(z − 2) · · · (z − 10)
2
 
1 1 18
= 2πi − =− πi
10! 9! 10!

21. The function 2
ez
f(z) =
z6
has a pole of order 6 at 0. To compute the residue at 0, we find the coefficient a1 in the Laurent
serie expansion about 0. We have

1 z2 1 X (z 2 )n
e = .
z6 z 6 n=0 n!
It is clear that this expansion has no terms with odd powers of z, positive or negative. Hence a−1 = 0
and so Z 2
ez
6
dz = 2πi Res (0) = 0.
C1 (0) z

25. Same approacha s in Exercise 21:

sin z 1 X z 2k+1
= (−1)k
z6 z 6 (2k + 1)!
k=0
 
1 z3 z5
= z − + − · · ·
z6 3! 5!
Section 5.1 Cauchy’s Residue Theorem 103

Coefficient of 1z : a−1 = 1
5! , so
Z
sin z 2πi
dz = 2πi Res (0) = .
C1 (0) z6 5!

1
29. (a) The Order of a pole of csc(πz) = sin πz
is the order of the zero of

1
= sin πz.
csc(πz)

Since the zeros of sin πz occur at the integers and are all simple zeros (see Example 1, Section 4.6),
it follows that csc πz has simple poles at the integers.
(b) For an integer k,
 z−k
Res csc πz, k = lim (z − k) csc πz = lim
z→k z→k sin πz
1
= lim (l’Hospital’s rule)
z→k π cos πz
(−1)k
= .
π
(c) Suppose that f is analytic at an integer k. Apply Proposition 1(iii), then

 (−1)k
Res f(z) csc(πz), k = f(k).
π

33. A Laurent series converges absolutely in its annulus of convergence. Thus to multiply two
Laurent series, we can use Cauchy products and sum the terms in any order. Write

X ∞
X
f(z)g(z) = an(z − z0 )n bn(z − z0 )n
n=−∞ n=−∞
X∞
= cn(z − z0 )n ,
n=−∞

where cn s obtained by collecting all the terms in (z − z0 )n, after expanding the product. Thus
X

cn = aj bn−j ;
j=−∞

in particular
X

c−1 = aj b−1−j ,
j=−∞

and hence

X

Res f(z)g(z), z0 = aj b−j−1.
j=−∞

37. (a) We have, Exercise 35(a), Section 4.5,
Z 
1 z
ζ− 1ζ ) dζ
J0 (z) = e2 .
2πi C1 (0) ζ
104 Chapter 5 Residue Theory

Thus Z ∞ Z ∞ Z 
1 s− 12 (ζ− 1ζ ) dζ
J0(t)e−st dt = e−t dt
0 2πi 0 C1 (0) ζ
(b) For ζ on C1 (0), we have
1
ζ− = ζ − ζ = 2i Im (ζ),
ζ
which is 0 if Im (ζ) = 0 (i.e., ζ = ±1) or is purely imaginary. In any case, for all ζ ∈ C1(0), and all
real s > 0 and t, we have

−t s− 1 (ζ− 1 ) −ts t (ζ− 1 )
e 2 ζ = e e 2 ζ
= e−ts .

So, by the inequality on integrals (Th.2, Sec. 3.2),

1 Z −t s− 1 (ζ− 1 ) 1
−t s− 12 (ζ− ζ1 ) dζ 2π
e ≤ max e 2 ζ
2πi C1 (0) ζ 2π ζ∈C1 (0) ζ

−t s− 1 (ζ− 1 )
=
max e 2 ζ (|ζ| = 1)
ζ∈C1 (0)
≤ e−ts

Thus the iterated integral in (a) is absolutely convergent because
Z ∞ Z 
1 Z ∞Z 
dζ dζ
1 1 1 1
e−t s− 2 (ζ− ζ ) dt ≤ e−t s− 2 (ζ− ζ ) dt
2πi 0 C1 (0) ζ 0 C1 (0) ζ
Z ∞
1
≤ e−ts dt = < ∞
0 s

(c) Interchange the order of integration, and evaluate the integral in t, and get
Z ∞ Z Z ∞ 
1 1 1 dζ
J0(t)e−st dt = e−t s− 2 (ζ− ζ ) dt
0 2πi C1 (0) 0 ζ
Z  ∞
1 1 −t s− 12 (ζ− 1ζ ) dζ
= e
2πi C1 (0) s − 12 (ζ − 1ζ ) ζ
0
Z
1 1
= dζ
πi C1 (0) −ζ 2 + 2sζ + 1

because, as t → ∞,
−t 
e s− 12 (ζ− 1ζ ) = e−ts → 0.

(d) We evaluate the integral using the residue theorem. We have simple poles at

−s ± s2 + 1 p
ζ= = s ± s2 + 1.
−1
√ √
Only√s − s2 + 1 is inside C0 (1). To see this, note that because s > 0 and s2 + 1 > 1, we have
s + s2 + 1 > 1. Also
p p p
s < s2 + 1 < 1 + s ⇒ −1 − s < − s2 + 1 < −s ⇒ −1 < s − s2 + 1 < 0.
Section 5.1 Cauchy’s Residue Theorem 105

By Proposition 1(ii)
 p 
1 1
Res 2
,s− s +1 =
−ζ 2 + 2sζ + 1 −2ζ + 2s ζ=s−√s2 +1
1
= √
2 s2 + 1

Thus, for s > 0,
Z ∞
2πi 1 1
J0 (t)e−st dt = √ = √ .
0 πi 2 s + 1
2 2
s +1

(e) Repeat the above steps making the appropriate changes.
Step (a’): Use the integral representation of Jn and get
Z ∞ Z ∞ Z 
1 s− 12 (ζ− ζ1 ) dζ
In = Jn (t)e −st
dt = e−t dt
0 2πi 0 C1 (0) ζ n+1

Step (b’) is exactly like (b) because, for ζ on C1(0), we have |ζ n+1 | = |ζ| = 1.
Step (c’): As in (c), we obtain
Z
1 1
In = dζ.
πi C1 (0) (−ζ 2 + 2sζ + 1)ζ n

Let η = ζ1 = ζ. Then dζ = − η12 dη. As ζ runs through C1(0) in the positive direction, η runs through
C1(0) in the negative direction. Hence

Z − η12 dη Z
1 1 ηndη
In = 1 = .
πi −C1 (0) (− η 2 + 2s η1 + 1) η1n πi C1 (0) η2 + 2sη − 1

Step (d’): We evaluate the integral using the residue theorem. We have simple poles at

−s ± s2 + 1 p
ζ= = −s ± s2 + 1.
−1

Only −s + s2 + 1 is inside C0(1). (Just arge as we did in (d).) By Proposition 1(ii)
 p 
ηn 2
ηn
Res , −s + s + 1 =
η2 + 2sη − 1 2η + 2s η=−s+√s2 +1
√ n
s2 + 1 − s
= √
2 s2 + 1

Thus, for s > 0,
Z ∞ p n
1
Jn(t)e−st dt = √ s2 + 1 − s .
0 s2 + 1
106 Chapter 5 Residue Theory

Solutions to Exercises 5.2
−i z+1/z
1. Let z = eiθ , dz = ieiθ dθ, dθ = z dz, cos θ = 2 . Then
Z 2π Z
dθ − zi dz
= (z+1/z)
0 2 − cos θ C1 (0) 2− 2
Z
dz
= −i z2 1
C1 (0) 2z − 2
− 2
Z
dz
= −i 2
− z2 + 2z − 12
C1 (0)
X  2 
z 1
= 2π Res − + 2z − , zj ,
j
2 2

1
where the sum of the residues extends over all the poles of 2 inside the unit disk. We have
2z− z2 − 12

z2 1
− + 2z − = 0 ⇔ z 2 − 4z + 1 = 0.
2 2
√ √
The roots are z = 2 ± 3, and only z1 = 2 − 3 is inside C1(0). We compute the residue using
Proposition 1(ii), Sec. 5.1:
 2 
z 1 1 1
Res − + 2z − , z1 = = √ .
2 2 −z1 + 2 3
Hence Z 2π
dθ 2π
= √ .
0 2 − cos θ 3
Section 5.2 Definite Integrals of Trigonometric Functions 107

−i dz z+1/z e2iθ +e−2iθ z 2 + 12
5. Let z = eiθ , dz = ieiθ dθ, dθ = z , cos θ = 2 , and cos 2θ = 2 = 2
z
. Then
Z 2π Z 1 2 1
cos 2θ 2 (z + z 2 dz
dθ = −i (z+1/z)
0 5 + 4 cos θ C1 (0) 5 + 4 z
2
Z 4
i z +1 dz
= −
2 C1 (0) 5z 2 3
+ 2z + 2z z
Z
i z4 + 1
= − dz
2
C1 (0) z 2(2z 2
+ 5z + 2)
X  
i z4 + 1
= − 2πi Res , zj
2 j
z 2 (2z 2 + 5z + 2)
X  
z4 + 1
= π Res , zj ,
z 2 (2z 2 + 5z + 2)
j

4
where the sum of the residues extends over all the poles of z2 (2zz2 +5z+2)
+1
inside the unit disk. We
have a pole of order 2 at 0 and possible more poles at the roots of 2z 2 + 5z + 2. Let’s compute the
residue at 0.
 
z4 + 1 d z4 + 1
Res , 0 = lim
z 2 (2z 2 + 5z + 2) z→0 dz (2z 2 + 5z + 2)

4z 3 (2z 2 + 5z + 2) − (z 4 + 1)(4z + 5) 5
= 2 2 =− .
(2z + 5z + 2) z=0 4

For the nonzero poles, solve
2z 2 + 5z + 2 = 0.
−5±3
The roots are z = 4 . Only z1 = − 12 is inside C1 (0). We compute the residue using Proposition
1(ii), Sec. 5.1:
 
z4 + 1 z14 + 1 1
Res , z1 =
z (2z 2 + 5z + 2)
2 z12 d
dz
(2z 2 + 5z + 2) z
1

( 12 )4 + 1 17
= 1 1 = .
4 (4(− 2 ) + 5) 12

Hence Z  

cos 2θ 17 5 π
dθ = π − = .
0 5 + 4 cos θ 12 4 6
108 Chapter 5 Residue Theory

−i dz z+1/z z−1/z
9. . Let z = eiθ , dz = ieiθ dθ, dθ = z
, cos θ = 2
, sin θ = 2i
. Then
Z 2π Z
dθ dz
= −i 3 2 3
0 7 + 2 cos θ + 3 sin θ C (0) z(7 + (z + 1/z) + 2i z − 2i
Z 1
dz
= −i 3 2 + 7z + (1 + 3 i)
C1 (0) (1 − i)z
X  2 2

1
= 2π Res , zj ,
j
(1 − 32 i)z 2 + 7z + (1 + 32 i)

1
where the sum of the residues extends over all the poles of (1− 3 i)z2 +7z+(1+ 3 inside the unit disk.
2 2 i)
Solve
3 3
(1 − i)z 2 + 7z + (1 + i) = 0.
2 2
You’ll find
q q
−7 ± 49 − 4(1 − 32 i)(1 + 32 i) −7 ± 49 − 4(1 + 94 )
z = =
2(1 − 32 i) 2(1 − 32 i)

−7 ± 36 −7 ± 6
= =
2 − 3i 2 − 3i
−13 −1
= or
2 − 3i 2 − 3i
−13(2 + 3i) −(2 + 3i)
= = −(2 + 3i) or .
13 13
We have √
√ −(2 + 3i)
|2 + 3i| = 13 > 1 and = 13 < 1.
13 13
−(2+3i)
So only z1 = 13 is inside C1(0). We compute the residue using Proposition 1(ii), Sec. 5.1:

1
Res (z1 ) = 3
2(1 − 2
i)z1 +7
1 1
= = .
3 −(2+3i) 6
−2(1 − 2 i) 13 +7

Hence Z 2π
dθ 1 π
= 2π = .
0 7 + 2 cos θ + 3 sin θ 6 3
Section 5.2 Definite Integrals of Trigonometric Functions 109

13. The solution will vary a little from what is in the text. Note the trick based on periodicity.
Step 1: Double angle formula
 
1 + cos 2θ 2a + b + b cos 2θ
a + b cos2 θ = a + b = ,
2 2
so
1 2
= .
a + b cos2 θ 2a + b + b cos 2θ
Step 2. Change variables in the integral: t = 2θ, dt = 2dθ. Then
Z 2π Z 4π
dθ dt
I= 2
= .
0 a + b cos θ 0 2a + b + b cos t
1
The function f(t) = 2a+b+b cos t is 2π-periodic. Hence its integral over intervals of length 2π are
equal. So
Z 2π Z 4π Z 2π
dt dt dt
I= + =2 .
0 2a + b + b cos t 2π 2a + b + b cos t 0 2a + b + b cos t
Step 4. Now use the method of Section 5.2 to evaluate the last integral. Let z = eit , dz = ieit dt,
dt = −izdz , cos t = z+1/z
2 . Then
Z 2π Z
dt dz
2 = −4i 2
0 2a + b + b cos t C1 (0) bz + (4a + 2b)z + b
X  
1
= 8π Res , zj ,
bz 2 + (4a + 2b)z + b
j

1
where the sum of the residues extends over all the poles of bz 2 +(4a+2b)z+b inside the unit disk. Solve

bz 2 + (4a + 2b)z + b = 0,
and get
p p
−(4a + 2b) ± (4a + 2b)2 − 4b2 −(2a + b) ± 2 a(a + b)
z = =
2b p b p
−(2a + b) + 2 a(a + b) −(2a + b) − 2 a(a + b)
= z1 = or z2 = .
b b
It is not hard to prove that |z1 | < 1 and |z2| > 1. Indeed, for z2 , we have
p p
2a + b 2 a(a + b) 2a 2 a(a + b)
|z2| = + =1+ + >1
b b b b
because a, b are > 0. Now the product of the roots of a quadratic equation αz 2 + βz + γ = 0 (α 6= 0)
γ
is always equal to α . Applying this in our case, we find that z1 · z2 = 1, and since |z2| > 1, we must
have |z1| < 1.
We compute the residue at z1 using Proposition 1(ii), Sec. 5.1:
 
1 1
Res 2
, z1 =
bz + (4a + 2b)z + b 2bz1 + (4a + 2b)
1
= p
−(2a + b)2 + 4 a(a + b) + (4a + 2b)
1
= p .
4 a(a + b)
110 Chapter 5 Residue Theory

Hence Z 2π
dθ 1 2π
= 2π p =p .
0 a + cos2 θ 4 a(a + b) a(a + b)
Section 5.3 Improper Integrals Involving Rational and Exponential Functions 111

Solutions to Exercises 5.3
1. Use the same contour as Example 1. Several steps in the solution are very similar to those
in Example 1; in particular, Steps 1, 2, and 4. Following the notation of Example 1, we have
IγR = I[−R, R] + IσR . Also, limR→∞ IσR = 0, and
Z ∞
dx
lim I[−R, R] = I = .
R→∞ −∞ x4 +1

These assertions are proved in Example 1 and will not be repeated here. So all we need to do is
evaluate IγR for large values of R and then let R → ∞. We have
X  
1
IγR = 2πi Res , zj ,
j
z4 + 1

where the sum ranges over all the residues of z41+1 in the upper half-plane. The function z41+1 have
four (simple) poles. These are the roots of z 4 + 1 = 0 or z 4 = −1. Using the result of Example 1,
we find the roots to be
1+i −1 + i −1 − i 1−i
z1 = √ , z2 = √ , z3 = √ , z4 = √ .
2 2 2 2
In exponential form,
π 3π 5π 7π
z1 = ei 4 , z2 = ei 4 , z3 = ei 4 , z4 = ei 4 .

Only z1 and z2 are in the upper half-plane, and so inside γR for large R > 0. Using Prop.1(ii),
Section 5.1,
 4 
z +1 1 1 1 3π
Res 1 z1 = = 3 = e−i 4 ;
, d 4
z + 1 4z1 4
dz z=z1
 4 
z +1 1 1 1 9π 1 π
Res 1 z2 = = 3 = e−i 4 = e−i 4 .
, d 4
z + 1 4z2 4 4
dz z=z2

So
1  −i 3π π

IγR = 2πi e 4 + e−i 4
4 
πi 3π 3π π π
= cos − i sin + cos − i sin
2 4 4 4 4
πi √ π
= (−i 2) = √ .
2 2
π
Letting R → ∞, we obtain I = √
2
.
112 Chapter 5 Residue Theory

5. The integral converges absolutely, as in Step 1 of Example 1. We will reason as in this example,
and omit some of the details. Here IγR = I[−R, R] + IσR ; limR→∞ IσR = 0, and
Z ∞
dx
lim I[−R, R] = I = =.
R→∞ −∞ (x2 + 1)3

All we need to do is evaluate IγR for large values of R and then let R → ∞. The function f(z) =
1
(z 2 +1)3 has poles of order 3 at z1 = i and z2 = −i, but only z1 is in the upper half-plane. You can
evaluate the integral IγR using the residue theorem; however, an equal good and perhaps faster way
in this case is to use Cauchy’s generalized integral formula (Sec. 3.6). Write
Z Z Z
1 1 1 dz
IγR = 2 3
dz = 3
dz = 3 3
.
γR (z + 1) γR [(z + i)(z − i)] γR (z + i) (z − i)

1
Let g(z) = (z+i)3 . According to Theorem 2, Sec. 3.6,

g00 (i)
IγR = 2πi .
2!
Compute:
g0 (z) = −3(z + i)−4 , g00(z) = 12(z + i)−5 ,
so
12
g00 (i) = 12(2i)−5 = (−i).
25
Finally,
12 3π
IγR = π 5
= .
2 8

Letting R → ∞, we obtain I = 8 .

9. Proceed as in Example 3, as follows. Let x = et , dx = et dt. Then
Z ∞ Z ∞ Z ∞
dx et ex
= dt = dx,
0 x3 + 1 3t
−∞ e + 1 −∞ e
3x + 1

where in the last integral we reverted to the variable x for convenience. Use a contour as in Fig.6.
Refer to Example 3 for notation. As in Example 3, limR→∞ |I2| = 0 and limR→∞ |I4| = 0. Let us
repeat the proof for I4 here. For I4 , z is on γ4 , so z = −R + iy, where 0 ≤ y ≤ 2π
3 . Hence
Z
ez
|I4 | = dz ≤ l(I4 ) · M,
γ4 e +1
3z

z

where l(I4 ) = 3
= length of vertical side γ4 , and M is the maximum value of e3ze +1 on γ4 . For z
on γ4 ,
ez e−R+iy e−R
=
e3z + 1 e3(−R+iy) + 1 ≤ 1 − e−3R ,

which tends to 0 as R → ∞. To justify the last inequality, note that by the reverse triangle inequality:
=1

3(−R+iy)

z }| {
e + 1 ≥ 1 − e−3R e3iy
= 1 − e−3R ,
Section 5.3 Improper Integrals Involving Rational and Exponential Functions 113

so
1 1

e3(−R+iy) + 1 ≤ 1 − e−3R ,

as desired.
2πi
As in Example 3, I1 → I as R → ∞. For I3 , we have z = x + 3 , x varies from R to −R,
dz = dx, so
Z −R 2πi
ex+ 3
I3 = dx
Re3x+2πi + 1
Z R
2πi ex 2πi
= −e− 3 3x + 1
dx = −e− 3 I1 .
−R e

Now

e3z + 1 = 0 ⇒ e3z = −1 = eiπ
⇒ 3z = iπ + 2kπi, (k = 0, ±1, ±2, . . .)

⇒ z = (2k + 1).
3
ez
Only one root, z = iπ
3 , is inside γR , and so the function e3z +1 has one pole at z = iπ
3 inside γR .
Hence (see Example 3 for a justification)
   
2πi ez πi
1 − e− 3 I = 2πi Res 3z
,
e +1 3
iπ iπ
e3 e3
= 2πi = 2πi 3 iπ
d 3z
dz (e + 1) z= iπ
3
3e 3
iπ iπ
e 3 e 3
= 2πi = −2πi .
−3 3

Solving for I,
−2π iπ 1 −2π 1
I= ie 3 · 2πi = i · −iπ πi
,
3 1−e 3 3 e 3 −e3
−iπ −iπ πi
where we have multiplied numerator and denominator by e 3 . Now e 3 −e 3 = −2i sin π3 . So

π π 2π
I= = √ = √ .
3 sin π3 3 23 3 3

13. Use the same reasoning as in Example 3. Let x = et , dx = et dt. Then
Z ∞ √ Z ∞ t Z ∞ 3x
x e2 t e2
I= dx = e dt = dx.
0 x3 + 1 3t
−∞ e + 1 −∞ e
3x + 1

Use a contour as in Fig.6. Refer to Example 3 for notation: limR→∞ |I2| = 0, limR→∞ |I4 | = 0, and
limR→∞ |I1| = I, the desired integral. For I3 , we have z = x + 2πi
3
, x varies from R to −R, dz = dx,
so
Z −R 3 x+ 3 2πi Z −R 3 x πi Z R 3
e2 2 3 e2 e e2x
I3 = dx = dx = dx = I1 .
R e3x+2πi + 1 R e3x + 1 −R e
3x + 1
114 Chapter 5 Residue Theory

Same poles as in Exercise 9. So
3
!
e2z πi
2I = 2πi Res 3z
,
e +1 3
3 πi
e2 3
= 2πi
d 3z + 1)
dz (e z= iπ 3
πi
e2 2πi(i) 2π
= 2πi iπ = = ;
3e −3 3
π
I =
3

17. As in Example 4: Let 2x = et, 2dx = et dt. Then
Z ∞ Z ∞ Z ∞
ln(2x) 1 t xex
I= dx = e2t
et dt = 2 dx.
0 x2 + 4 2 −∞ 4
+ 4 −∞ e2x + 16

Use a rectangular contour as in Fig.6, whose vertical sides have length π. Refer to Example 3 for
notation: limR→∞ |I2| = 0, limR→∞ |I4 | = 0, and limR→∞ |I1| = I, the desired integral. For I3,
z = x + iπ, x varies from R to −R, dz = dx, so
Z −R
(x + iπ)ex+iπ
I3 = 2 dx
R e2x+2πi + 16
Z R
(x + iπ)ex (−1)
= −2 dx
−R e2x + 16
=BR i
z }| {
Z R Z R
xex ex
= 2 2x
dx + 2πi dx = I1 + BR i,
−R e + 16 −R e2x + 16

where BR is a real constant, because the integrand is real-valued. We have

IγR = I1 + I2 + I3 + I4 = 2I1 + I2 + I4 + iBR .

Letting R → ∞, we get
lim IγR = 2I + iB,
R→∞

where Z ∞
ex
B = lim BR = dx.
R→∞ −∞ e2x + 16
At the same time  
2zez
IγR = 2πi Res , z0 ,
e2z + 16

where z0 is the root of e2z + 16 = 0 that lies inside γR (there is only one root, as you will see):

e2z + 16 = 0 ⇒ e2z = 16eiπ = eln(16)+iπ
⇒ 2z = 4 ln 2 + iπ + 2kπi
π
⇒ z = 2 ln 2 + i (2k + 1).
2
Section 5.3 Improper Integrals Involving Rational and Exponential Functions 115

Only z = 2 ln 2 + i π2 is inside the contour. So
 
2zez π
IγR = 2πi Res 2z
, 2 ln 2 + i
e + 16 2
π
2(2 ln 2 + i π2 )e2 ln 2+i 2
= π
2e2(2 ln 2+i 2 )
π
= (4 ln 2 + iπ)
4
Thus
π2
2I + 2πiB = π ln 2 + i .
4
Taking real and imaginary parts, we find

π ln 2 π
I= and B= .
2 8
This gives the value of the desired integral I and also of the integral
Z ∞
π ex
=B= 2x + 16
dx.
8 −∞ e

21. We use the contour γR in Figure 9 and follow the hint. Write IγR = I1 + I2 + I3. On γ1 , z = x,
dz = dx,
Z Z R Z ∞
1 1 1
3+1
dz = 3+1
dx → I = 3+1
dx, as R → ∞.
γ1 z 0 x 0 x
2π 2π
i 3 i
On γ3 , z = e 3 x, dz = e dx, x varies from R to 0:
Z Z 0
1 2π 1
dz = e 3 i dx

γ3 z3 + 1 R e3 3 i x3 + 1
Z R

i 1 2π
= −e 3 dx → −e 3 i I as R → ∞.
0 x3 +1

For I2 , we have z = Reit , 0 ≤ t ≤ 2π
3 :
Z
1 1
dz ≤ l(γ2 ) max 3
3 z on γ2 z + 1
γ2 z + 1
2π 1
≤ R· 3 → 0as R → ∞.
3 R −1
Now  
1
I1 + I2 + I3 = IγR = 2πi Res , z0 ,
z3 + 1
π 1
where z0 = ei 3 is the only pole of z 3 +1 inside γR . By Proposition 1(ii), Sec. 5.1,
  2π
1 π 1 e−i 3
Res , ei 3 = 2π = .
z3 + 1 3ei 3 3

So 2π
e−i 3
I1 + I2 + I3 = 2πi .
3
116 Chapter 5 Residue Theory

Letting R → ∞, we get
2π 2π

i e−i 3 2π e−i 3
I −e 3 I = 2πi 3 i
⇒ (1 − e )I = 2πi
3 3
−i 2π π
2πi e 3 e− 3 i
⇒ I= × π
3 (1 − e 2π
3 i) e− 3 i
2πi e−iπ π
⇒ I= π π =
3 (e− 3 i − e 3 i) 3 sin π3

⇒ I= √ .
3 3
Solutions to Exercises 5.4
1. Use a contour as in Fig. 2 and proceed exactly as in Example 1, replacing s by 4. No need to
repeat the details.
5. The degree of the denominator is 2 more than the degree of the numerator; so we can use the
contour in Fig. 2 and proceed as in Example 1.
Step 1: The integral is absolutely convergent.
2
x cos 2x x2 1

(x2 + 1)2 (x2 + 1)2 ≤ x2 + 1 ,

because
x2 (x2 + 1) − 1 1 1 1
2 2
= 2 2
= 2 − 2 2
≤ 2 .
(x + 1) (x + 1) x + 1 (x + 1) x +1
R∞
Since −∞ x21+1 dx < ∞ (you can actually compute the integral = π), we conclude that our integral
is absolutely convergent.
Step 2:
Z ∞ 2 Z ∞ 2 Z ∞ 2 Z ∞
x cos 2x x cos 2x x sin 2x x2 e2ix
2 2
dx = 2 2
dx + i 2 2
dx = 2 2
dx
−∞ (x + 1) −∞ (x + 1) −∞ (x + 1) −∞ (x + 1)

because Z ∞
x2 sin 2x
2 2
dx = 0,
−∞ (x + 1)
being the integral of an odd function over a symmetric interval.
Step 3: Let γR and σR be as in Fig. 2. We will show that
Z
z 2e2iz
2 2
dz → 0 as R → ∞.
σR (z + 1)
Z 2 2iz
z 2e2iz z e

|IσR | = dz ≤ l(σ ) max = πR · M.
2 2 R
z on σR (z 2 + 1)2

σR (z + 1)
As in (7) in Sec. 5.4, for z on σR , 2iz
e ≤ e−2R sin θ ≤ 1.
So
2 2iz 2
z e z2 z + 1 − 1

≤ 2 ≤
(z 2 + 1)2 (z + 1)2 (z 2 + 1)2

1 1

≤ 2 +
z + 1 (z 2 + 1)2
1 1
≤ + .
R2 − 1 (R2 − 1)2
Section 5.4 Improper Integrals of Products of Rational and Trigonometric Functions 117

So
πR πR
|IσR | = + ,
R2 − 1 (R2 − 1)2
and this goes to zero as R → ∞.
Step 4: We have Z  
z 2e2iz z 2 e2iz
dz = 2π Res , i ,
γR (z 2 + 1)2 (z 2 + 1)2
because we have only one pole of order 2 at i in the upper half-plane.
 2 2iz   2 2iz

z e d 2 z e
Res , i = lim (z − i) 2
(z 2 + 1)2 z→i dz (z + 1)2
 2 2iz 
d z e
=
dz (z + i) z=i
2

e−2
= i ,
4
−2 −2
after many (hard-to-type-but-easy-to-compute) steps that we omit. So IγR = 2πi(i e 4 ) = −π e 2 .
−2
Letting R → ∞ and using the fact that IγR → I, the desired integral, we find I = −π e 2 .
9. In the integral the degree of the denominator is only one more than the degree of the numerator.
So the integral converges in the principal value sense as in Examples 2 or 4. Let us check if the
denominator has roots on the real axis:

2 1 35
x +x+9= 0⇒x= − ±i .
2 2
We have no roots on the real axis, so we will proceed as in Example 2, and use Jordan’s Lemma.
Refer to Example 2 for further details of the solution. Consider
Z Z
z iπz
2
e dz = f(z)eiπz dz,
γR z + z + 9 γR

where γR is as in Fig. 5. By Corollary 1,
Z

f(z)e dz → 0, as R → ∞.
iπz

σR

Apply the residue theorem:
Z

f(z)eiπz dz = 2πi Res f(z)eiπz , z1 ,
γR

where z1 = − 12 + i 235 is the only (simple) pole of f(z)eiπz in the upper half-plane. By Proposition
1(ii), Sec. 5.1, we have

 zeiπz

Res f(z)eiπz , z1 =
d
dz
(z + z + 9)
2
z=z1

z1 e iπz1 −( 12 + 235 i) −π √35
= = √ e 2
2z1 + 1 35
√ √
35 35
e−π 2 e−π 2
= − √ −i
2 35 2
118 Chapter 5 Residue Theory

So
Z √
35

35
!
iπz e−π 2 e−π 2
f(z)e dz = 2πi − √ −i
γR 2 35 2

35
−π

35 e−π 2
= πe 2 + iπ √
35
But Z Z ∞ Z ∞
x cos πx x sin πx
lim f(z)eiπz dz = dx + i dx.
R→∞ γR −∞ x2 + x + 9 −∞ x2 + x + 9
Taking real and imaginary parts, we get
Z ∞ Z ∞

35
x cos πx √
35 x sin πx e−π 2
2
dx = πe−π 2 and 2
dx = π √ .
−∞ x + x + 9 −∞ x +x+9 35

13. Use an indented contour γr,R as in Fig. 9, with an indentation around 0, and consider the
integral Z Z
1 − eiz
Ir,R = dz = g(z) dz,
γr,R z2 γr,R
iz
where g(z) = 1−e
z 2 . Note that g(z) has a simple pole at 0. To see this, consider its Laurent series
expansion around 0:
1 
g(z) = 1 − 1 + (iz) + (iz)2 /2! + (iz)3 /3! + · · ·
z2
i 1 z
= − + +i − ···
z 2 3!
Moreover, Res (g(z), 0) = −i. By Corollary 2,
Z
lim g(z) = iπ(−i) = π.
r→0+ σr

(Keep in mind that σr has a positive orientation, so it is traversed in the opposit direction on γr,R .
See Fig. 9.) On the outer semi-circle, we have
Z
πR π
g(z) dz ≤ 2 max |1 − eiz | = max |1 − eiz |.
R z on σR R z on σR
σR

Using an estimate as in Example 1, Step 3, we find that for z on σR ,
iz
e ≤ e−R sin θ ≤ 1.

Hence maxz on σR |1 − eiz | ≤ 1 + 1 = 2. So |IσR | ≤ 2π
R → 0, as R → ∞. Now g(z) is analytic inside
and on the simple path γr,R . So by Cauchy’s theorem,
Z
g(z) dz = 0.
γr,R

So
→0 →π →I
Z zZ }| { zZ }| { zZ }| Z {
0= g(z) dz = g(z) dz − g(z) dz + g(z) dz + g(z) dz .
γr,R σR σr [−R, −r] [r, R]

As r → 0+ and R → ∞, we obtain
Z ∞
1 − cos x
P.V. dx = Re (I) = Re (π) = π.
−∞ x2
Section 5.4 Improper Integrals of Products of Rational and Trigonometric Functions 119

17. Use an indented contour as in Fig. 9 with an indentation around 0. We have
Z ∞  Z ∞ 
sin x eix
P.V. 2
dx = Im P.V. 2
dx = Im (I).
−∞ x(x + 1) −∞ x(x + 1)

Z Z
eiz
Ir,R = dz = g(z) dz,
γr,R z(z 2 + 1) γr,R

eiz
where g(z) = z(z 2 +1) . Note that g(z) has a simple pole at 0, and Res (g(z), 0) = 1. By Corollary 2,
Z
lim g(z) = iπ(1) = iπ.
r→0+ σr

(As in Exercise 13, σr has a positive orientation, so it is traversed in the opposit direction on γr,R .
See Fig. 9.) As in Example 1, |IσR | → 0, as R → ∞. Now g(z) has a simple pole at i inside γr,R .
So by the residue theorem,
Z
ei(i)
g(z) dz = 2π Res (g(z), i) = 2πi = −πe−1 .
γr,R i(2i)

So
→0 →iπ →I
Z zZ }| { zZ }| { zZ }| Z {
−1
−πe = g(z) dz = g(z) dz − g(z) dz + g(z) dz + g(z) dz .
γr,R σR σr [−R, −r] [r, R]

As r → 0+ and R → ∞, we obtain
−πe−1 = I − iπ.
Solving for I and taking imaginary parts, we find

I = i(π − πe−1 ) Im (I) = π − πe−1 ,

which is the value of the desired integral.
120 Chapter 5 Residue Theory

25. (a) For w > 0, consider Z
eiwz
J= dz,
C e2πz − 1
where C is the indented contour in Figure 14. The integrand is analytic inside and on C. So
Z
eiwz
(1) 0=J = 2πz
dz = I1 + I2 + I3 + I4 + I6 + I6 ,
C e −1

where Ij is the integral over the jth component of C, starting with the line segement [, R] and
moving around C counterclockwise. As  → 0+ and R → ∞, I1 → I, the desired integral.
For I3 , z = x + i, where x varies from R to :
Z  Z R
eiw(x+i) −w eiwx
I3 = 2π(x+i) − 1
dx = −e 2πx − 1
dx.
R e  e

As  → 0+ and R → ∞, I3 → −e−w I.
For I2 , z = R + iy, where y varies from 0 to 1:
iwz
e
|I3| ≤ 1 · max 2πz ;
z=R+iy e − 1
0≤y≤1

iwz iw(R+iy)
e e
=
e2πz − 1 e2π(R+iy) − 1
iwR
e
=
e2πR − 1 → 0, as R → ∞.

So I2 → 0, as R → ∞.
For I5 , z = iy, where y varies from 1 −  to :
Z  Z 1−
eiw(iy) e−wy
I5 = i dy = −i dy.
1− e
2π(iy) − 1
 e2πiy − 1

For I4 , the integral over the quarter circe from (,  + i) to (0, i − i), we apply Corollary 2, to
compute the limit
Z  iwz 
eiwz π e
lim dz = −i Res ,i
→0 γ e2πz − 1 2 e2πz − 1
4

π eiw(i) e−w
= −i 2π(i)
= −i .
2 2πe 4
For I6 , the integral over the quarter circe from (0, i) to (, 0), we apply Corollary 2, to compute
the limit
Z  iwz 
eiwz π e
lim dz = − Res ,0
→0 γ e2πz − 1 2 e2πz − 1
6

π eiw(0) i
= − i 2π(0)
=− .
2 2πe 4
Plug these findings in (1) and take the limit as R → ∞ then as  → 0, and get
Z ∞ Z ∞ Z 1− 
eiwx −w eiwx e−wy e−w i
lim 2πx
dx − e 2πx
dx − i 2πiy
dy −i − =0
→0  e − 1  e − 1  e − 1 4 4
Section 5.4 Improper Integrals of Products of Rational and Trigonometric Functions 121

or Z Z Z 
∞ ∞ 1−
eiwx eiwx e−wy e−w 1
lim dx − e−w dx − i dy = i( + ).
→0  e2πx − 1  e2πx − 1  e2πiy −1 4 4
(Note: The limits of each individual improper integral does not exist. But the limit of the sum,
as shown above does exist. So, we must work wth the limit of the three terms together.) Take
imaginary parts on both sides:
Z ∞ Z ∞ Z 1− 
eiwx eiwx e−wy e−w + 1
(2) lim Im 2πx
dx − e−w 2πx
dx − i 2πiy
dy = .
→0  e −1  e −1  e −1 4

Now
Z ∞ Z ∞Z 1− 
eiwx eiwx e−wy
Im dx − e−w dx − i dy
 e2πx − 1  e2πx − 1  e2πiy − 1
Z ∞ Z 1−  
sin wx −w (−i)e−wy
(3) = (1 − e ) dx + Im dy;
 e2πx − 1  e2πiy − 1

   
(−i)e−wy (−i)e−wy
Im = − Re
e2πiy − 1 e2πiy − 1
 −wy 
e
= − Re
e2πiy − 1
   
−wy 1 −wy e−iπy
= −e Re = −e Re
e2πiy − 1 eiπy − e−iπy
  −wy
cos πy − i sin πy e
= −e−wy Re = .
2i sin πy 2

Plugging this in (3) and using (2), we get
Z ∞ Z 1− 
sin wx e−wy e−w + 1
lim (1 − e−w ) dx + dy = .
→0  e2πx − 1  2 4

Evaluate the second integral and get
Z ∞ 
sin wx −w e−w(1−) − e−w e−w + 1
lim (1 − e ) dx + = .
→0  e2πx − 1 2(−w) 4
Z ∞
sin wx −w e−w(1−) − e−w e−w + 1
lim (1 − e ) dx + lim = .
→0  e2πx − 1 →0 2(−w) 4
Z ∞
sin wx 1 − e−w e−w + 1
(1 − e−w ) lim 2πx
dx + = .
→0  e −1 2w 4
So
Z ∞
−w sin wx e−w + 1 1 − e−w
(1 − e ) lim dx = − ;
→0  e2πx − 1 4 2w
Z ∞  −w 
sin wx 1 e + 1 1 − e−w
lim dx = −
→0  e2πx − 1 1 − e−w 4 2w
Z ∞ w
sin wx −1 1 e + 1
dx = + .
0 e2πx − 1 2w 4 ew − 1
122 Chapter 5 Residue Theory

(b) From (10), Section 4.4 (recall B0 = 1):

X B2n 2n
z coth z = 22n z , |z| < π
n=0
(2n)!
X∞
z z B2n z 2n z
coth = 22n ( ) , | |<π
2 2 n=0
(2n)! 2 2
z
ze +e
2 − z2 X∞
B2n 2n
z z = z , |z| < 2π
2 e 2 − e− 2 n=0
(2n)!
B0 X B2n 2n−1
z z ∞
1 e 2 + e− 2
z z = + z , |z| < 2π
2 e 2 − e− 2 z n=1
(2n)!
z z
X∞
1 e 2 + e− 2 B0 B2n 2n−1
z z − = z , |z| < 2π
2e −e
2 − 2 z (2n)!
n=1

1 ez + 1 1 1 X B2n 2n−1
z
− = z , |z| < 2π
4 e − 1 2z 2 n=1 (2n)!

(c) Replace sin wx in the integral in (a) by its Taylor series

X w2k−1x2k−1
sin wx = (−1)k−1 ,
(2k − 1)!
k=1

ue (b), and interchange order of integration, and get
Z X 1 X B2n 2n−1
∞ ∞ ∞
1 (wx)2n+1
(−1)n dx = w
0 e2πx − 1 n=0 (2n + 1)! 2 n=1 (2n)!
X∞ Z ∞ ∞
(−1)n x2n+1 1 X B2n 2n−1
w2n+1 dx = w
n=0
(2n + 1)! 0 e2πx − 1 2 n=1 (2n)!

X (−1)n−1 Z ∞

x2n−1 1 X B2n 2n−1
w2n−1 2πx
dx = w .
n=1
(2n − 1)! 0 e −1 2 n=1 (2n)!

Comparing the coefficients of w, we obtain
Z ∞ 2n−1
x (−1)n−1
dx = B2n (n = 1, 2, . . .).
0 e2πx − 1 4n
Section 5.5 Advanced Integrals by Residues 123

Solutions to Exercises 5.5
1. As in Example 1, take w > 0 (the integral is an even fnction of w),
Z ∞ Z ∞ Z ∞
1 2
− x2 1 2
− x2 iwx x2
√ e cos wx dx = √ e e dx (because e− 2 sin wx dx = 0)
2π −∞ 2π −∞ −∞
Z ∞
1 1 2 1 2
= √ e− 2 (x−iw) − 2 w dx
2π −∞
=J
zZ }| { 1 2

− 12 w 2 1 − 12 (x−iw)2 e− 2 w
= e √ e dx = √ J.
2π −∞ 2π
To evaluate J, consider the integral
Z
1 2
I= e− 2 (z−iw) dz,
γR

where γR is a rectangualr contour as in Fig. 1, with length of the vertical sides equal to w. By
Cauchy’s theorem, I = 0 for all R.
Let Ij denote the integral on γj (see Example 1). Using the estimate in Example 1, we see that
I2 and I4 tend to 0 as R → ∞. On γ3 , z = z + iw, where x varies from R ro −R:
Z −R Z R
1 2 1 2
e− 2 (x+iw−iw) dx = − e− 2 x dx,
R −R

and this tends to − 2π as R → ∞, by (1), Sec. 5.4. Since I = 0 for all R, it follows that

J = lim Iγ1 = − lim Iγ3 = 2π.
R→∞ R→∞

Consequently, Z ∞
1 x2 w2
√ e− 2 cos wx dx = e− 2
2π −∞

for all w > 0. Since the integral is even inw, the formula holds for w < 0. For w = 0 the formula
follows from (1), Sec. 5.5.
124 Chapter 5 Residue Theory

5. . (a) Let f(x) = cos 2x, then f 0 (x) = −2 sin 2x and f 00 (x) = −4 cos 2x. If x ∈ (0, π4 ), then
f 00 (x) < 0 and the graph concaves down. So any chord joining two points on the graph of y = cos 2x
above the interval (0, π4 ) lies under the graph of y = cos 2x. take the two points on the graph, (0, 1)
and ( π4 , 0). The equation of the line joining them is y = − π4 x + 1. Since it is under the graph of
y = cos 2x for x ∈ (0, π4 ), we obtain

4 π
− x + 1 ≤ cos 2x for 0 ≤ x ≤
π 4
2 2
(b) Let Ij denote the integral of e−z over the path γj in Fig. 13. Since e−z is entre, by Cauchy’s
theorem, I1 + I2 + I3 = 0.
(c) For I1 , z = x,
Z R Z ∞ √
−x2 −x2 π
I1 = e dx → e dx = , as R → ∞,
0 0 2
by (1), Sec. 5.5. On γ2 , z = Reiθ z 2 = R2(cos 2θ + i sin 2θ),
2
−z −R2 (cos 2θ+i sin 2θ) 2 2 4
e = e = e−R cos 2θ ≤ e−R (1− π θ).

Parametrize the integral I2 and estimate:
Z π
4
−R2 e2iθ iθ
|I2 | = e Rie dθ
0
Z π4 Z π
−R2 e2iθ iθ 4 2 4
≤ R e e dθ ≤ R e−R (1− π θ)

0 0
Z π
2 4 2 4
≤ Re−R eR πθ dθ
0
π
−R2 1 π R2 4 θ 4 π −R2 h R2 i
= Re e π
= e e − 1 ,
R2 4 0 4R

which tends to 0 as R → ∞.
π π
(d) On γ3 , z = xei 4 , where x varies from R to 0, dz = ei 4 dx. So
Z R Z R
π 2 iπ π 2
I3 = −ei 4 e−x e 2
dx = −ei 4 e−x idx
0 0
Z R
π
= −ei 4 (cos x2 − i sin x2 )dx.
0

As R → ∞, I3 converges to Z ∞
π
−ei 4 (cos x2 − i sin x2 )dx.
0

(e) Let R → ∞ in the sum I1 + I2 + I3 = 0 and get
π R∞

π
2 − ei 4 0 (cos x2 − i sin x2 )dx = 0;
π R∞

ei 4 0 (cos x2 − i sin x2)dx = 2π ;
R∞ 2 2

π −i π
0
(cos x − i sin x )dx = 2 e
4 ;

R∞ √ √ √ 
0
(cos x2 − i sin x2)dx = 2π 22 − i 22 .

The desired result follows upon taking real and imaginary parts.
Section 5.5 Advanced Integrals by Residues 125


9. We will integrate the function f(z) = z(z−1)1√z−2 around the contour in Figure 16. Here z − 2
is defined with the branch of the logarithm with a branch cut on the positive real axis. It is
mulitple valued √
on the semi-axis
√ x > 2. Aproaching the real axis from above and to right of 2,
we have√limz→x z √ − 2 = x − 2. Aproaching the real axis from below and to right of 2, we have
limz→x z − 2 = − x − 2. Write
I = I1 + I2 + I3 + I4 ,
where I1 is the integral over the small circular part; I2 is the integral over the interval above the
x-axis to the right of 2; I3 is the integral over the larger circular path; I4 is the integral over the
interval (neg. orientation) below the x-axis to the right of 2. We have I1 → 0 as r → 0 and I3 → 0
as R → ∞. (See Example 3 for similar details.) We have I2 → I as R → ∞ and I4 → I as R → ∞,
where I is the desired integral. So

2I = 2πi [ Res (0) + Res (1)] ;
1 1
Res (0) = lim √ = −√
z→0 (z − 1) z − 2 −2
1 1
= − 1 log (−2) = − 1 (ln 2+iπ)
e2 0 e2
1 i
= − √ = √ ;
i 2 2
1
Res (1) = √ = −i;
−1
   
i 1
I = iπ √ − i = π 1 − √ .
2 2

13. (a) In Figure 19, let
γ1 denote the small circular path around 0 (negative direction);
γ2 the line segment from r to 1 − r, above the x-axis (positive direction);
γ3 the small semi-circular path around 1 above the x-axis (negative direction);
γ4 the line segment from 1 + r to R − 1, above the x-axis (positive direction);
γ5 the large circular path around 0 (positive direction);
γ6 the line segment from R to 1 + r, below the x-axis (negative direction);
γ7 the small semi-circular path around 1 below the x-axis (negative direction);
γ8 the line segment from 1 − r to r, below the x-axis (negative direction).
We integrate the function
zp
f(z) =
z(1 − z)
on the contour γ, where z p = ep log0 z (branch cut along positive x-axis). By Cauchy’s theorem,
Z X
8
f(z) dz = 0 ⇒ Ij = 0.
γ j=1

Review the integrals I3 , I4, I7 , and I8 from Example 3, then you can show in a similar way that I1,
I3 , and I7 tend to 0 as r → 0. Also I5 → 0 as R → ∞. We will give some details. For I1 ,
Z
zp
|I1| = dz
γ1 z(1 − z)
rp rp
= 2πr = 2π → 0 as r → 0.
r(1 − r) (1 − r)
126 Chapter 5 Residue Theory

For I2 and I5 , we have
Z ∞
xp
I2 + I5 → dx, as r → 0 and R → ∞.
0 x(1 − x)
For I3 and I6 , we have
Z ∞ Z ∞
ep log0 x xp
I3 + I6 → − dx = −e2pπi dx, as r → 0 and R → ∞.
0 x(1 − x) 0 x(1 − x)
To evaluate I4 and I7 , we use a trick that will allow us to apply Lemma 2, Sec. 5.4. Note that on
γ4 , log0 z = Log z, and on γ7 , log0 z = log π2 z. This allows us to replace log0 by a branch of the log
which is analytic in a neighborhood of the contour of integration and ths allows us to apply Lemma
2 of Sec. 5.4. According to this lemma, as r → 0,
Z
ep Log z
I4 = → iπep Log (1) = iπ;
γ4 z(1 − z)
and Z p log π z
e 2 p log π (1)
I7 = → iπe 2 = iπe2πpi .
γ4 z(1 − z)
So as r → 0, Z
ep Log z
I4 + I7 = → iπ(1 + e2πpi ).
γ4 z(1 − z)
Adding the integrals together and then taking limits, we get

(1 − e2pπi )I + iπ(1 + e2πpi ) = 0
2πpi
I = −iπ 1+e
1−e2πpi
−πpi πpi
I = −iπ ee−πpi +e
−eπpi
= π cot pπ.

(b) Use x = et , do the substitution, then replace t by x, and get, from (a),
Z ∞
epx
P.V. x
dx = π cot pπ (0 < p < 1).
−∞ 1 − e

(c) Change variables x = 2u, dx = 2 du, then
Z ∞
e2pu
π cot pπ = 2P.V. du
−∞ 1 − e2u
Z ∞ Z ∞ (2p−1)u
e2pu e−u e
= 2P.V. −u − eu
du = −P.V. du
−∞ e −∞ sinh u
Z ∞
w+1 ewu
−π cot(π ) = P.V. du (w = 2p − 1).
2 −∞ sinh u

But
w+1 πw
− cot(π ) = tan ,
2 2
so Z ∞
ewu πw
P.V. du = tan .
−∞ sinh u 2
Replace u by x and get Z ∞
ewx πw
P.V. dx = π tan .
−∞ sinh x 2
Section 5.5 Advanced Integrals by Residues 127

(d) If |a| < b, take t = bx and w = a, then
Z ∞
eax π πa
P.V. dx = tan .
−∞ sinh bx b 2b

(e) Replace a by −a in (d) and get
Z ∞
e−ax π πa
P.V. dx = − tan .
−∞ sinh bx b 2b

Subtract from (d) and divide by 2:
Z ∞
sinh ax π πa
dx = tan (b > |a|).
−∞ sinh bx b 2b

Note that the integral is convergent so there is no need to use the principal value.
17. We use a contour like th eone in Fig. 17: Let
γ1 denote the small circular path around 0 (negative direction);
γ2 the line segment from r to R, above the x-axis (positive direction);
γ3 the large circular path around 0 (positive direction);
γ4 the line segment from R to r, below the x-axis (negative direction). We integrate the function

z
f(z) = 2
z +z+1
√ 1
on the contour γ, where z = e 2 log0 z (branch cut along positive x-axis). By the residue theorem,
Z X 4
X X
f(z) dz = 2πi Res (f, zj ) ⇒ Ij = 2πi Res (f, zj ),
γ j j=1 j

where the sum is over all the residues of f in the region inside γ. The poles of f in this region are
at the roots of z 2 + z + 1 = 0 or
√ √ √
−1 ± −3 −1 3 −1 3
z= ; z1 = +i , z2 = −i .
2 2 2 2 2
We have
q √ 2π
|z1 | = ( −1 2 3 2
2 ) + ( 2 ) = 1, z1 = ei 3

log0 (z1 ) = ln |z1| + i arg 0 (z1 ) = 0 + i 2π
3
= i 2π
3
√ 1
z1 e 2 log 0 (z1 )
Res (z1 ) = 2z1 +1 = 2z1 +1
π
ei 3
Res (z1 ) = 2π .
2ei 3 +1

Similarly,
q √ 4π
|z2 | = ( −1 2 − 3 2
2 ) + ( 2 ) = 1, z2 = ei 3

log0 (z2 ) = ln |z2| + i arg 0 (z2 ) = 0 + i 4π
3
= i 4π
3
√ 1
z2 e 2 log 0 (z2 )
Res (z2 ) = 2z2 +1 = 2z2 +1

ei 3
Res (z2 ) = 4π .
2ei 3 +1
128 Chapter 5 Residue Theory

Let us now compute the integrals. For I1 ,
Z √
z
|I1| = 2
dz
γ1 z + z + 1

r
= 2πr → 0 as r → 0.
1 − r2 − r
For I2 , we have √ Z ∞
x
I2 →
2
dx = I, as r → 0 and R → ∞.
0 x +x+1
√ 1 √
A simple estimate shows that I3 → 0 as R → ∞. For I4 , z = e 2 (ln |z|+2πi) = xeiπ . So
Z ∞ 1 Z ∞ √
e 2 log0 x x
I4 → − dx = −eπi dx = I, as r → 0 and R → ∞.
0 x2 + x + 1 0 x2 +x+1

Adding the integrals together and then taking limits, we get
π 2π
!
ei 3 ei 3
2I = 2πi 2π + 4π
2ei 3 +1 2ei 3 +1
!
iπ i 2π
e 3 e 3
I = πi i 2π
+ i 4π
2e 3 +1 2e 3 +1
!
i 5π iπ i 4π 2π
2e 3 +e 3 + 2e 3 + ei 3 π
= πi i 6π i 2π i 4π
=√ .
4e 3 + 2e 3 + 2e 3 +1 3

Use
π 1 i√ 2π 1 i√
ei 3 = + 3, ei 3 =− + 3,
2 2 2 2
3π 4π 1 i√
ei 3 = −1, ei 3 =− − 3,
2 2
5π 1 i√ 6π
ei 3 = − 3, ei 3 =1
2 2
Section 5.6 Summing Series by Residues 129

Solutions to Exercises 5.6
1. Note that f(z) = z21+9 has two simples poles at ±3i. Since f(z) does not have poles on the
integers, we may apply Proposition 1 and get

X 1
= −π [ Res (f(z) cot πz, 3i) + Res (f(z) cot πz, −3i)]
k2 + 9
k=−∞
 
cot πz cot πz
= −π lim + lim
z→3i z + 3i z→−3i z − 3i
 
1 1
= −π cot(3πi) − cot(−3πi)
6i 6i
π π
= [i cot(3πi)] = coth(3π),
3 3
because cot(iz) = −i coth(z). You can prove the last identity by using (25) and (26) of Section 1.6.
5. Reason as in Exercise 1 with f(z) = 4z21−1 , which has two simples poles at ± 12 . Since f(z) does
not have poles on the integrers, we may apply Proposition 1 and get

X
∞  
1 1 1
= −π Res (f(z) cot πz, ) + Res (f(z) cot πz, − )
4k2 − 1 2 2
k=−∞
" #
cot πz cot πz
= −π lim 1
+ lim 1
z→ 12 4(z + 2 ) z→− 12 4(z − 2 )
πh π π i
= − cot( ) − cot(− ) = 0,
2 2 2
because cot( π2 ) = 0.
1
9. Reason as in Exercise 1 with f(z) = (z−2)(z−1)+1 , which has two simples poles where z 2 −3z+3 = 0
or √ √ √
3 ± −3 3+i 3 3−i 3
z= ; z1 = , z2 = .
2 2 2
Since f(z) does not have poles on the integrers, we may apply Proposition 1 and get

X 1
= −π [ Res (f(z) cot πz, z1 ) + Res (f(z) cot πz, z2 )]
(k − 2)(k − 1) + 1
k=−∞

Let’s compute:
cot πz cot πz1
Res (z1 ) = lim =
(z − z2 )
z→z1 (z1 − z2 )

z1 − z2 = i 3,
√ ! √ ! √ !
3+i 3 3 3
cot(πz1 ) = cot π = − tan iπ = −i tanh π .
2 2 2


(Prove and use the identities cot(z + 2 ) = − tan(z) and tan(iz) = i tanh z.) So
√  √ 
−i tanh 23 π tanh 23 π
Res (z1 ) = √ =− √ .
i 3 3
130 Chapter 5 Residue Theory

A similar computation shows that Res (z2 ) = Res (z1 ), hence
∞ √ !
X 1 2π 3
= √ tanh π
(k − 2)(k − 1) + 1 3 2
k=−∞

13. Just reapeat the proof of Proposition 1. The only difference is that here w no longer have
Res (f(z) cot πz, 0) = f(z). This residue has to be computed anew.
1
17. (a) Apply the result of Execrise 13 with f(z) = z 2n :

X

1 cot(πz) 
= −π Res ,0 ,
k=−∞
k2n z 2n
k6=0

because f has only one pole of order 2n at 0. The sum on the left is even, so
X

1 π cot(πz) 
2n
= − Res ,0 ,
k 2 z 2n
k=1

(b) Recall the Taylor series expansion of z cot z from Exercise 31, Section 4.4,

X

22k B2k 2k
z cot z = (−1)k z
(2k)!
k=0
X∞
22k B2k π2k 2k
(πz) cot(πz) = (−1)k z
(2k)!
k=0

X 22k B2k π2k−1 2k−1
cot(πz) = (−1)k z .
(2k)!
k=0

So

cot(πz) X 2k
k 2 B2k π
2k−1
= (−1) z +2k−2n−1.
z 2n (2k)!
k=0

The residue at 0 is a−1 , the coefficient of 1z , which is obtained from the series above when 2k−2n−1 =
−1 or n = k. Hence
cot(πz)  2n
n 2 B2n π
2n−1
Res , 0 = (−1) .
z 2n (2n)!
Using (a), we get
X

1 2n−1
B2nπ2n
n−1 2
= (−1) .
k2n (2n)!
k=1
Section 5.7 The Counting Theorem and Rouché Theorem 131

Solutions to Exercises 5.7
1. The roots of the polynomial are easy to find using the quadratic formula:

z 2 + 2z + 2 = 0 ⇒ z = −1 ± i.

Thus no roots are in the first quadrant. This, of course, does not answer the exercise. We must
arrive at this answer using the method of Example 1, with the help of the argument principle.
First, we must argue that f has no roots of the positive x − axis. This is clear, because if x > 0
then x2 + x + 2 > 2 and so it cannot possibly be equal to 0. Second, we must argue that there are
no roots on the upper imaginary axis. f(iy) = −y2 + 2y + 2 = 2 − y2 + 2y. If y = 0, f(0) = 2 6= 0.
If y > 0, then Im (f(y)) = 2y 6= 0. In all cases, f(iy) 6= 0 if y ≥ 0.
The number of zeros of the polynomial f(z) = z 2 + 2z + 2 is equal to the number of times the
image of σR wraps around the origin, where γR is the circular path in the first quadrant, in Fig. 4.
This path consists of the interval [0, R], the circular arc σR , and the interval on the imaginary axis
from iR to 0. To find the image on γR , we consider the image of each component separately.
Since f(x) is real for real x, we conclude that the image of the interval [0, R] is also an interval,
and it is easy to see that this interval is [2, R2 + 2R + 2]. So its initial point is w0 = 2 and its
terminal point is w1 = R2 + 2R + 2.
The image of the arc σR starts at the point w1 = R2 +2R+2 and ends at f(iR) = −R2 +2iR+2 =
w2, which is the image of the terminal point of σR . We have Im (w2 )) = 2R and Re (w2 ) = 2−R2 < 0
if R is very large. Hence the point f(w2 ) is in the second quadrant. Also, for very large R, and
|z| = R, the mapping z 7→ f(z) is approximately like z 7→ z 2 . So f(z) takes σR and maps it
”approximately” to the semi-circle (the map w = z 2 doubles the angles), with initial point w1 and
terminal point w2.
We now come to the third part of the image of γR . We know that it starts at w2 and end at w0.
As this image path go from w2 to w0 , does it wrap around zero or not? To answer this question, we
consider f(iy) = 2 − R2 + 2iy. Since Im (f(iy) > 0 if y > 0, we conclude that the image point of iy
remains in the upper half-plane as it moves from w2 to w0. Consequently, the image curve does not
wrap around 0; and hence the polynomial has no roots in the first quadrant-as expected.
5. Follow the steps in the solution of Exercise 1, but here the roots of z 4 + 8z 2 + 16z + 20 = 0 are
not so easy to find, so we will not give them.
Argue that f has no roots of the positive x − axis. This is clear, because if x > 0 then
x4 + 8x2 + 16z + 20 > 20 and so it cannot possibly be equal to 0. Second, we must argue that there
are no roots on the upper imaginary axis. f(iy) = y4 − 8y2 + 16iy + 20 = y4 − 8y2 + 20 − 8iy. If
y = 0, f(0) = 20 6= 0. If y > 0, then Im (f(y)) = −8y 6= 0. In all cases, f(iy) 6= 0 if y ≥ 0.
The number of zeros of the polynomial f(z) = z 4 + 8z 2 + 16z + 20 is equal to the number of
times the image of σR wraps around the origin, where γR is the circular path in the first quadrant,
in Fig. 4. This path consists of the interval [0, R], the circular arc σR , and the interval on the
imaginary axis from iR to 0. To find the image on γR , we consider the image of each component
separately.
Since f(x) is real for real x, we conclude that the image of the interval [0, R] is also an interval,
and it is easy to see that this interval is [20, R4 + 8R2 + 16R + 20]. So its initial point is w0 = 2
and its terminal point is w1 = R4 + 8R2 + 16R + 20.
The image of the arc σR starts at the point w1 = R4 + 8R2 + 16R + 20 on the real axis and ends
at f(iR) = R4 − 8R2 + 20 + 16iR = w2, which is the image of the terminal point of σR . We have
Im (w2)) = 16R > 0 and Re (w2 ) = R4 − 8R2 + 20 > 0 if R is very large. Hence the point f(w2 ) is
in the first quadrant. Also, for very large R, and |z| = R, the mapping z 7→ f(z) is approximately
like z 7→ z 2 . So f(z) takes σR and maps it ”approximately” to a circle (the map w = z 4 multiplies
angles by 4), with initial point w1 and terminal point w2. So far, the image of [0, R] and σR wraps
one around the origin.
132 Chapter 5 Residue Theory

We now come to the third part of the image of γR . We know that it starts at w2 and end at
w0. As this image path go from w2 to w0, does it close the loop around 0 or does it unwrap it? To
answer this question, we consider f(iy) = R4 − 8R2 + 20 + 16iy. Since Im (f(iy) = 16y > 0 if y > 0,
we conclude that the image point of iy remains in the upper half-plane as it moves from w2 to w0.
Consequently, the image curve wraps around 0; and hence the polynomial has one root in the first
quadrant.
9. Apply Rouché’s theorem with f(z) = 11, g(z) = 7z 3 + 3z 2. On |z| = 1, |f(z)| = 11 and
|g(z)| ≤ 7 + 3 = 10. Since |f| > |g| on |z| = 1, we conclude that N (f) = N (f + g) inside C1(0).
Since N (f) = 0 we conclude that the polynomial 7z 3 + 3z 2 + 11 has no roots in the unit disk.
13. Apply Rouché’s theorem with f(z) = −3z, g(z) = ez . On |z| = 1, |f(z)| = 3 and

|g(z)| = ecos t+i sin t = ecos t ≤ e.

Since |f| > |g| on |z| = 1, we conclude that N (f) = N (f + g) inside C1(0). Since N (f) = 1 we
conclude that the function ez − 3z has one root in the unit disk.
17. Apply Rouché’s theorem with f(z) = 5, g(z) = z 5 + 3z. On |z| = 1, |f(z)| = 5 and |g(z)| ≤ 4.
Since |f| > |g| on |z| = 1, we conclude that N (f) = N (f + g) inside C1(0). Since N (f) = 0 we
1
conclude that the function z 5 + 3z + 5 has no roots in the unit disk. Hence z5 +3z+5 is analytic in
the unit disk and by Cauchy’s theorem
Z
dz
5
= 0.
C1 (0) z + 3z + 5

21. (a) Let g(z) = z and f(z) = z n − 1. Then
Z Z Xn
1 nz n−1 1 f 0 (z)
z dz = g(z) dz = g(zj ),
2πi CR (0) zn − 1 2πi CR (0) f(z) j=1
Pn
where zj are the roots of f(z). These are precisely the n roots of unity. Hence j=1 g(zj ) =
Pn 1 1
j=1 zj = S. Let ζ = z
, dζ = − z 2 dz, as z runs through CR (0) in the positive direction, ζ runs
through C1/R(0) in the negative direction. So
Z Z Z
1 nz n−1 1 n dζ n dζ
z n
dz = n 1 2
= .
2πi CR (0) z −1 2πi −C 1 (0) ζ ( ζ n − 1) −ζ 2πi −C 1 (0) (1 − ζ n )ζ
R R

(b) Evaluate the second integral in (a) using Cauchy’s generalized integral formula and conclude
that Z
n dζ d 1 −nζ n−1
= = = 0.
2πi −C 1 (0) (1 − ζ n )ζ dζ 1 − ζ n ζ=0 (1 − ζ n )2 ζ=0
R

Using (a), we find S = 0.
A different way to evaluate S is as follows: From (a),
Z
n zn
S = dz
2πi CR (0) z n − 1
Z Z
n zn − 1 n 1
= n
dz + n
dz
2πi CR (0) z − 1 2πi CR (0) z − 1
Z Z
n n 1
= 1 · dz + n
dz
2πi CR (0) 2πi CR (0) z − 1
= 0 + 0 = 0.
Section 5.7 The Counting Theorem and Rouché Theorem 133

The first integral is 0 by Cauchy’s theorem. The second integral is zero because CR (0) contains all
the roots of p(z) = z n − 1 (see Exercise 38, Sec. 3.4).
25. Suppose that f is not identically 0 in Ω, and let Br (z0 ) ⊂ Ω be a disk such that f(z) 6= 0 for
all z on Cr (z0 ). Since |f| is continuous, it attains its maximum and minimum on Cr (z0 ). So there
is a point z1 on Cr (z0 ) such that |f(z1 )| = m = min |f| on Cr (z0 ). Since |f(z1 )| 6= 0, we find that
m > 0.
Apply uniform convergence to get an index N such that n > N implies that |fn − f| < m ≤ |f|
on Cr (z0 ). Now, |f| ≥ m on Cr (z0 ), so we can apply Rouché’s theorem and conclude that N (f) =
N (fn − f + f) = N (fn ) for all n ≥ N . That is, the number of zeros of fn inside Cr (z0 ) is equal to
the number of zeros of f inside Cr (z0 ).
29. Follow the steps outlined in Exercise 28. Let Ω denote the open unit disk and let fn (z) =
z 5 + z 4 + 6z 2 + 3z + 11 + n1 . Then fn(z) converge uniformly to p(z) = z 5 + z 4 + 6z 2 + 3z + 11 on any
closed disk in Ω. Since p(z) is clearly not identically zero, it follows from Hurwitz’s theorem that
N (fn ) = N (p) in |z| < 1 for all sufficiently large n. But, an application of Rouché’s theorem shows
that fn (z) has no zeros in N1 (0) or on C1(0). (Take f(z) = 11 + n1 and g(z) = z 5 + z + 6z 2 + 3z.)
Hence p(z) has no roots in N1 (0).
Note: Hurwitz’s theorem does not apply to p(z) on the boundary of N1 (0). On this boundary,
p(z) may have roots. Consider another simpler example: p(z) = z 2 + 1, fn (z) = z 2 + 1 + n1 . Then
fn converges to p uniformly on any closed set. Clearly, p(z) has two roots, ±i, on the boundary of
the unit disk, but fn (z) has no roots inside or on the boundary of the unit disk.