Laurent series

1 Lemma. Let a ∈ C and let 0 ⩽ Ri < Ro ⩽ +∞. Denote the domain {z ∈ C :

Ri < |z − a| < Ro } by D.
(1) If a series of the form

X
+∞
an (z − a)n .
n=−∞

converges on D, then it converges compactly on D.

(2) If an equation

X
+∞ X
+∞
an (z − a) =
n
bn (z − a)n
n=−∞ n=−∞

holds for all z ∈ D, then an = bn for all n ∈ Z.

Proof. (1) The convergence of

X
+∞
an (z − a)n
n=−∞
P+∞ P+∞
means that the two series n=0 an (z − a)n and n=1 a−n (z − a)−n converge on
P+∞
D. So n=0 an un converges for u satisfying |u| < Ro and its radius of convergence
P+∞
is greater than or equal to Ro . Similarly, n=1 a−n un converges for u satisfying
|u| < R1i and its radius of convergence is greater than or equal to 1/Ri . These two
power series converge compactly on {u ∈ C : |u| < Ro } and {u ∈ C : |u| < 1/Ri },
respectively. P
+∞
Therefore n=0 an (z − a)n converges compactly on {z ∈ C : |z − a| < Ro } and
P+∞ −1 n
n=1 a−n [(z − a) ] converges compactly on {z ∈ C : |z − a|−1 < 1/Ri }. They
both converge compactly on

{z ∈ C : |z − a| < Ro }∩{z ∈ C : |z − a|−1 < 1/Ri }

= {z ∈ C : Ri < |z − a| < Ro } = D.
P+∞
Hence the given series n=−∞ an (z − a)n converges compactly on D.
(2) Suppose that

X
+∞ X
+∞
an (z − a) =
n
bn (z − a)n
n=−∞ n=−∞

for all z ∈ D. Let r be a real number strictly between Ri and Ro . By Part (1), these
series converge uniformly on the circle C given by z(t) = a + re2πit (0 ⩽ t ⩽ 1).
For each k ∈ Z, we have

X
+∞ X
+∞
an (z − a) n−k−1
= bn (z − a)n−k−1
n=−∞ n=−∞

1
and both sides converge uniformly on C. So termwise integration is legitimate and
yields
2πiak = 2πibk
since

ˆ  (z−a)n−k terminal point of C = 0 for k 6= n,
(z − a)n−k−1 dz = ´ n−k initial point of C
 1
C
C z−a
dz = 2πi for k = n.

Thus ak = bk for all k ∈ Z. □

1 Theorem. (Laurent expansion theorem) Let a ∈ C and let 0 ⩽ Ri < Ro ⩽ +∞.

Denote the domain {z ∈ C : Ri < |z − a| < Ro } by D.
If n ∈ Z and if f is holomorphic on the ring D, then, for every r ∈ R satisfying
Ri < r < Ro , the integral
ˆ
f (ζ)
dζ
∂B(a,r) (ζ − a)
n+1

has the same value.

For all n ∈ Z, write
ˆ
1 f (ζ)
an := dζ,
2πi ∂B(a,r) (ζ − a)n+1

where r is any real number such that Ri < r < Ro . Then for every z ∈ D, we have

X
+∞
f (z) = an (z − a)n .
n=−∞

Proof. Let r0 , r1 be two real numbers such that Ri < r0 < r1 < Ro . Let C0 , C1 be
the positively oriented circles of radii r0 , r1 (with common center a). Then C0 is
deformable to C1 on D and the function z 7−→ f (z)/(z − a)n+1 is holomorphic on
D. By the free end points version of Cauchy’s theorem (Theorem ??), we obtain
´ f (ζ) ´ f (ζ)
C0 (ζ−a)n+1
dζ = C1 (ζ−a) n+1 dζ.

The same assertion can be alternatively established by using the fixed-end-

points homotopy version of Cauchy’s theorem. For this purpose, we divide the
circles into two semicircles by a diameter of the larger circle, and consider the two
closed paths U , L each consisting of one of the small semicircles, one of the larger
semicircles, a portion of the diameter from the smaller semicircle to the larger one,
a large semicircle and a portion of the diameter from the larger semicircle to the
smaller one. Since f is holomorphic on D, the integral´ ´of f along U and along L
are both 0. Adding them, we get the difference C1 − C0 , because the integrals
´ ´
along line segments cancel during addition. Hence we obtain C0 = C1 .
Now we turn to the expansion of f (z) in positive and negative powers of z − a.
We fix a z in the domain D and set |z − a| = r, and then determine two real
numbers r0 and r1 such that Ri < r0 < r < r1 < R0 . Let C0 , C1 denote the

2
 Figure 1: Two applications of Cauchy’s theorem

positively oriented circles of radii r0 , r1 , with center a. We take two line segments
L1 , L2 which join the circles´ and which´ do not pass through z. Cauchy’s integral
1 1
formula yields f (z) = 2πi C1 + 2πi −C0 . We shall now compute the integrals
along C0 and C1 .

z z

a a

For any ζ ∈ |C0 |, we have | z−a

ζ−a
| = r0
< 1, and there is a real number M0
(ζ) ζ−a
n

r
M0 r0 n
such that |f (ζ)| ⩽ M0 for all ζ ∈ |C0 |. Then fz−a z−a ⩽ r r for all n. By
Weierstrass’s M-test, the series on the right side converges uniformly on |C0 | to
the function on the left side below:

X
+∞  n
f (ζ) 1 f (ζ) ζ − a
= .
z − a 1 − z−a
ζ−a
n=0
z − a z − a
3
So we can integrate along C0 term by term and get
ˆ +∞ ˆ
X  n
f (ζ) f (ζ) ζ − a
− dζ = dζ
C0 ζ −z n=0 C0
z−a z−a
+∞ ˆ
X 
f (ζ)
= −n
dζ (z − a)−(n+1)
n=0 C0 (ζ − a)
X
+∞ ˆ 
f (ζ)
= −n+1
dζ (z − a)−n ,
n=1 C0 (ζ − a)
ˆ X
+∞
1 f (ζ)
dζ = a−n (z − a)−n .
2πi −C0 ζ −z n=1

The computation of the integral along C1 follows similar lines. For any ζ ∈ |C1 |,
ζ−a | = r1 < 1, and there is a real number M1 such that |f (ζ)| ⩽ M1 for
we have | z−a r

all ζ ∈ |C1 |. By Weierstrass’s M-test again, we have

X
+∞  n
f (ζ) 1 f (ζ) z − a
= ,
ζ − a 1 − z−a
ζ−a n=0
ζ −a ζ −a

the convergence being uniform on |C1 |. Integrating term by term along C1 , we get
ˆ +∞ ˆ
X  n
f (ζ) f (ζ) z − a
dζ = dζ
C1 ζ −z n=0 C 1
ζ − a ζ − a
+∞ ˆ
X 
f (ζ)
= dζ (z − a)n ,
n=0 C 1
(ζ − a) n+1

ˆ X
+∞
1 f (ζ)
dζ = an (z − a)n .
2πi C1 ζ −z n=0

We get
ˆ ˆ
1 f (ζ) 1 f (ζ)
f (z) = dζ + dζ
2πi C1 ζ −z 2πi −C0 ζ −z
X
+∞ X
+∞
= an (z − a)n + a−n (z − a)−n
n=0 n=1
X
+∞
= an (z − a)n .
n=−∞

Definition. Let a ∈ C and let 0 ⩽ Ri < Ro ⩽ +∞. Let D := {z ∈ C : Ri <

|z − a| < Ro }.

4
 P+∞
If f is a holomorphic function on D, then the series f (z) = n=−∞ an (z − a)n
of the last theorem, with
ˆ
1 f (ζ)
an := dζ,
2πi ∂B(a,r) (ζ − a)n+1

is called the Laurent

P+∞ series or nthe Laurent expansion of f on D.
The series n=0 an (z − a) of nonnegative powers is called the regular part of
the Laurent expansion
P+∞ of f on −n D, or, more shortly, the regular part of f on D.
The series n=1 a−n (z − a) of negative powers is called the principal part of
the Laurent expansion of f on D, or, more shortly, the principal part of f on D.
The coefficient a−1 is called the residue of f at a and is denoted as Res f (z).
z=a

Definition. Let a ∈ C and let R be a positive real number or +∞. Write D :=

{z ∈P C : 0 < |z − a| < R} and let f : D −→ C be a function holomorphic on D.
+∞
Let n=−∞ an (z − a)n be the Laurent series of f on D.
If an = 0 for all n < 0, then a is said to be a removable singularity of f .
If there exists a negative integer n with an 6= 0, but if the number of n’s
satisfying n < 0 and an 6= 0 is finite, then a is called a pole of f . The largest
positive integer m for which a−m 6= 0 holds is called the order of the pole a of f .
If a is a pole of order 1 of f , then a is called a simple pole of f .
If there exist infinitely many n’s such that n < 0 and an 6= 0, then a is called
an essential singularity of f .

Behavior of holomorphic functions near isolated singularities

1 Proposition. Let D be a domain, a ∈ D, and let f be a function defined on
D \ {a}.
Then a is a removable singularity of f if and only if there is an r such that f
is bounded on B(a, r) \ {a}.
Proof. Assume that f has a removable singularity at a. Then there is a positive
real number R such that
X
+∞
f (z) = an (z − a)n for all z satisfying 0 < |z − a| < R
n=0

At z = a, this expansion need not be valid. We define a new function g :

B(a, R) −→ C by (
f (z) if z 6= a,
g(z) :=
a0 if z = a.
Then
X
+∞
g(z) = an (z − a)n for all z ∈ B(a, R)
n=0

and g is a holomorphic function on B(a, R). Put r := 21 R. Since g is continuous

on the compact set B[a, r], it is bounded on B[a, r], so g is bounded on B(a, r),
so f is bounded on B(a, r) \ {a}.

5
 Conversely, suppose that there are positive real numbers r, M such that
|f (z)| ⩽ M for all z ∈ B(a, r) \ {a}. Since f is holomorphic on B(a, r) \ {a} =
{z ∈ C : 0 < |z − a| < r}, we can develop f in a Laurent series about a, say

X
+∞
f (z) = an (z − a)n for all z ∈ B(a, r) \ {a}.
n=−∞

Then, for all n < 0 and for any positive real number r0 with r0 < r, we have
ˆ
1 f (ζ)
an = dζ
2πi |ζ−a|=r0 (ζ − a)n+1
1 M
|an | ⩽ −|n|+1
2πr0 ,
2π r
0
|n|
|an | ⩽ M r0

Since r0 can be chosen as small as we please, it follows that |an | = 0 for all n < 0.
Thus a is a removable singularity of f . □

2 Lemma. Let D be a domain and let f : D −→ C be a function defined on D.

If f is holomorphic on D \ {a} and f is continuous at a, then f is holomorphic on
D.

Proof. Continuity of f at a implies that f is bounded in a neighborhood of a,

and Proposition 1 yields that f has a removable singularity at a. Let a0 denote
the constant term in the Laurent expansion of f about a. Then the function
g : D −→ C, defined by
(
f (z) if z 6= a,
g(z) :=
a0 if z = a

is holomorphic on D. But since f is continuous at a, we have f (a) = a0 = g(a)

and f = g. Thus f is holomorphic at a. □

Compare to the absolute value function on R, which is continuous on R, dif-

ferentiable on R \ {0}, but which is not differentiable at 0.
3 Lemma. Let D be a domain and a ∈ D. Let f be a function defined on D. If
f is continuous at a, then there is a δ > 0 such that

1
|f (z)| ⩾ |f (a)| for all z ∈ B(a, δ).
2
Proof. Exercise. □

2 Proposition. Let D be a domain, a ∈ D and let f : D \ {a} −→ C be a

holomorphic function on D \ {a}. If a is a pole of f , then, for all positive real
numbers M , there is a δ > 0 such that 0 < |z − a| < δ =⇒ |f (z)| > M. In other
words, limz→a |f (z)| = +∞.

6
Proof. Let k denote the order of a as a pole of f . Then in some neighborhood of
g(z)
a except at a, we can write f as f (z) = (z−a) k , where g is a holomorphic function

in that neighborhood of a, and satisfies g(a) 6= 0.

Since g is continuous at a, there exists a real number δ such that |g(z)| ⩾ 12 |g(a)|
for all z ∈ B(a, δ).
′
Let
q M > 0 be given. Choose a positive real number δ with δ < δ and
k |g(a)|
δ< 2M . Then for every complex number z in B(a, δ), there holds

|g(z)| 2 |g(a)| 2 |g(a)|

1 1
|f (z)| = > > q = M.
|z − a|k δk k |g(a)| k
( 2M )

3 Proposition. (Casorati-Weierstrass theorem) Let D be a domain, a ∈ D and

let f : D \ {a} −→ C be a holomorphic function on D \ {a}. If a is an essential
singularity of f , then, for every complex number c, for every positive real number
r, for every positive real number ϵ, there is a z ∈ B(a, r) \ {a} such that

|f (z) − c| < ϵ.

Proof. Suppose this is not the case. Then there is a complex number c, a positive
real number r and a positive real number ϵ such that

|f (z) − c| ⩾ ϵ for all z ∈ B(a, r) \ {a}.

1
Thus f (z)−c ⩽ 1ϵ for all z ∈ B(a, r) \ {a}. Put g(z) := f (z)−c
1
for z ∈ B(a, r) \
{a}. Then g is holomorphic and bounded on B(a, r) \ {a}, so g has a removable
singularity at a. Therefore, for some suitable nonnegative integer k and some
holomorphic function h satisfying h(a) 6= 0, we have f (z)−c
1
= g(z) = (z − a)k h(z)
k h(z) . Since h(a) 6= 0, h does not vanish
1 1
in a neighborhood of a, so f (z) = c+ (z−a)
in a neighborhood of a. Thus 1/h is holomorphic at a, and f (z) has a removable
singularity or a pole at a.
This proves the contrapositive of the claim. □

The residue theorem and its applications

Residue formulas
Let f have a simple pole at a. Then the expansion of f in a Laurent series has
the form
1
f (z) = a−1 + a0 + a1 (z − a) + a2 (z − a)2 + · · · ,
z−a
so
(z − a)f (z) = a−1 + a0 (z − a) + a1 (z − a)2 + a2 (z − a)3 + · · ·
and
Res f (z) = lim (z − a)f (z).
z=a z→a

7
 g(z)
Assume that f (z) = h(z) , g, h are holomorphic at a and g(a) 6= 0, h(a) = 0,
′
h (a) 6= 0. Then h has a simple zero at a and f has a simple pole at a. We have

g(z) g(z)
lim (z − a)f (z) = lim (z − a) = lim h(z)−h(a)
z→a z→a h(z) z→a
z−a

and the above formula yields

g(z) g(a)
Res = ′ .
z=a h(z) h (a)

g(z)
Assume that f (z) = (z−a) m , g is holomorphic at a and g(a) 6= 0. Then f has

a pole of order m at a. We have

X
+∞ (n)
g (a)
g(z) = (z − a)n
n=0
n!

g(z) X
+∞ (n)
g (a)
f (z) = = (z − a)n−m
(z − a) m
n=0
n!

g (m−1) (a)
and the coefficient of (z − a)−1 = (z − a)(m−1)−m is equal to (m−1)! , so

g(z) g (m−1) (a)

Res = .
z=a (z − a)m (m − 1)!

Equivalently, residue of f at a pole a of order m is given by

1 dm−1
Res f (z) = [(z − a)m f (z)].
z=a (m − 1)! dz m−1

The Residue Theorem

2 Theorem. Let f be holomorphic on a domain D except for poles.
Let C be a simple closed curve in D that doesn’t pass through a pole and that
encloses (in the subset of D that lies on the left side of C) a finite number of poles
a1 , a2 , …, an of f .
Then
ˆ Xn
f = 2πi Res f.
C z=aj
j=1

Proof. About each of the points aj , we draw a circle, of such a small radius that
the circle centered at aj does not enclose any other pole of f , and the circle is
contained within D. We shall designate this circle by Cj .
Let a denote the initial and terminal point of C. We draw a path Pj from a
to a point on the circle Cj .

8
 By Cauchy’s theorem, we get
ˆ
f = 0,
A
ˆ ˆ ˆ ˆ !
X
k
f+ f − nj f− f = 0,
C j=1 Pj Cj Pj
ˆ X
k ˆ
f− nj f = 0,
C j=1 Cj
ˆ X
k ˆ
f= nj f,
C j=1 Cj
ˆ X
k
1
f= nj Res f
2πi C j=1
z=aj

and the claim is proved. □

Evaluation of real definite integrals

4 Lemma. Let θ1 , θ2 be real numbers, θ1 < θ2 + 2π, and let f be a continuous
function defined on the angular sector A := {0} ∪ {reθi ∈ C : θ1 ⩽ θ ⩽ θ2 }. Let
CR be the arc of the circle with center 0 and radius R comprised by the angles θ1
and θ2 , oriented positively. ´
If limz→∞ zf (z) = 0 on A, then limR→+∞ CR f (z) dz = 0.

Proof. Let M (R) := sup{|f (Reθi )| : θ1 ⩽ θ ⩽ θ2 }. By hypothesis, RM (R) → 0 as

R → +∞. From
ˆ
0⩽ f (z) dz ⩽ M (R) · length of CR = M (R)R(θ2 − θ1 ),
CR

we get ˆ
lim f (z) dz = 0.
R→+∞ CR

□
5 Lemma. Let θ1 , θ2 be real numbers, θ1 < θ2 + 2π, and let f be a continuous
function defined on the angular sector A := {0} ∪ {reθi ∈ C : θ1 ⩽ θ ⩽ θ2 }. Let
Cρ be the arc of the circle with center 0 and radius ρ comprised by the angles θ1
and θ2 , oriented positively. ´
If limz→0 zf (z) = 0 on A, then limρ→0 Cρ f (z) dz = 0.

Proof. Let M (ρ) := sup{|f (ρeθi )| : θ1 ⩽ θ ⩽ θ2 }. By hypothesis, ρM (ρ) → 0 as

ρ → 0. From
ˆ
0⩽ f (z) dz ⩽ M (ρ) · length of Cρ = M (ρ)ρ(θ2 − θ1 ),
Cρ

9
we get ˆ
lim f (z) dz = 0.
ρ→0 Cρ
□
´π
6 Lemma. For any r > 0, we have 0
e−r sin θ r dθ < π.

Proof. On the interval [0, 12 π], the function θ1 sin θ decreases (because its derivative
is non-positive since θ ⩽ tan θ when 0 ⩽ θ ⩽ 12 π). Therefore
sin θ 1
1⩾ ⩾ ,
θ π/2
2
θ ⩽ sin θ,
π
2
e−r sin θ ⩽ e− π rθ
for any r > 0 and for any 0 ⩽ θ ⩽ 12 π. We obtain
ˆ π ˆ π/2
−r sin θ
e r dθ = 2 e−r sin θ r dθ
0 0
ˆ π/2 2
⩽2 e− π rθ r dθ
0
ˆ
π π/2 − 2 rθ 2
=2· e π π r dθ
2 0
ˆ r
=π e−u du
0
ˆ +∞
<π e−u du = −π[e−u ]u=+∞
u=0
0
= π.
□
7 Lemma. Let 0 ⩽ θ1 < θ2 ⩽ π and let f be a continuous function defined on the
angular sector A := {0} ∪ {reθi ∈ C : θ1 ⩽ θ ⩽ θ2 }. Let CR be the arc of the circle
with center 0 and radius R comprised by the ´angles θ1 and θ2 , oriented positively.
If limz→∞ f (z) = 0 on A, then limR→+∞ CR f (z)eiz dz = 0.

Proof. Let M (R) := sup{|f (Reθi )| : θ1 ⩽ θ ⩽ θ2 }. By hypothesis, M (R) → 0 as

R → +∞. Using Lemma 6, we get
ˆ ˆ θ2
0⩽ iz
f (z)e dz = f (Reθi )eiR(cos θ+i sin θ) Reiθ i dθ
CR θ1
ˆ θ2
⩽ f (Reθi )eiR(cos θ+i sin θ) Reiθ i dθ
θ1
ˆ θ2
⩽ M (R)e−R sin θ R dθ ⩽ M (R)π,
θ1

10
so ˆ
lim f (z)eiz dz = 0.
R→+∞ CR

8 Lemma. Let θ1 , θ2 be real numbers, 0 < θ2 − θ1 ⩽ 2π. Suppose f has a simple

pole at z = a and let Cρ be the circular arc {a + ρeθi ∈ C : θ1 ⩽ θ ⩽ θ2 }, oriented
positively. Then ˆ
θ2 − θ1
lim f (z) dz = 2πi Res f (z).
ρ→0 C
ρ
2π z=a

Proof. Let b be the residue of f at a. There exists a function h, holomorphic at a,

b
and a sufficiently small r > 0 such that f (z) = z−a + h(z) for all z ∈ B(a, r) \ {a}.
We have
ˆ ˆ ˆ
dz
f (z) dz = b + h(z) dz for all ρ < r.
Cρ Cρ z − a Cρ

Let M denote the supremum of |h(z)| on B[a, r/2]. The integral of h is bounded
above by M (θ2 − θ1 )ρ if ρ < r/2, hence it tends to 0 as ρ → 0. Then we get
ˆ ˆ ˆ θ2
dz ρeθi i dθ
lim f (z) dz = b lim = b lim = b(θ2 − θ1 )i.
ρ→0 Cρ ρ→0 Cρ z−a ρ→0 θ1 ρeθi

*
* *
Using the lemmas above, we can evaluate a large number of integrals. When
the integration interval is given, we try to find a simple closed curve whose one
part in the given interval of integration in such a way that the integrals over the
remaining parts can be estimated or evaluated. The residue theorem enables us
to find the value of integral over the required interval.
This method will be illustrated in the case of three types of integrals.

´ 2π
Integrals of type 0
f (cos θ, sin θ) dθ

We assume that f (x, y) is a rational function of x, y. The substitution z = eθi

brings the integral to the form
ˆ  
z + z −1 z − z −1 dz
f , ,
∂B(0,1) 2 2i zi

since cos θ = 12 (eθi + e−θi ), sin θ = 2i

1
(eθi − e−θi ) and dz = eθi i dθ = zi dθ. The
integral is then equal to 2πi times the sum of the residues of the new integrand
inside the unit circle (assuming that the new integrand has no singularity on the
unit circle).

11
 ´ 2π
1 Example. To evaluate 0
dθ
a+sin θ , where a ∈ R, a > 1. We put z = eθi . Then
−1
dz = eθi i dθ = zi dθ and sin θ = z−z2i , dθ = dz/zi. We find
ˆ 2π ˆ ˆ ˆ
dθ 1 dz 2i 1 2
= = dz = dz,
0 a + sin θ C a+ z−z −1
2i
zi C 2ai + z − 1
z
zi C z2 + 2aiz − 1

C being the√ positively oriented unit circle. The√integrand has two singularities,
√
1
2 [−2ai± −4a 2 + 4], let us call them z = (−a+ a2 − 1)i, z = (−a− a2 − 1)i.
0 1
Since |z0 | < 1 and |z1 | > 1, we take only the residue at z0 into consideration. We
obtain
2 2 1 1
Res = = = √
z=z0 z 2 + 2aiz − 1 2z + 2ai z=z0 z0 + ai ( a − 1)i
2

and ˆ 2π
dθ 1 2π
= 2πi √ =√ .
0 a + sin θ ( a2 − 1)i a2 − 1
´ +∞
Integrals of type −∞
f (x) dx
Let f (x) be a rational function of x such that the degree of its denominator is
greater than or equal to 2+ the degree of its numerator. In this case, the integral
´ +∞ ´R
−∞
f (x) dx is equal to limR→+∞ −R f (x) dx and zf (z) → 0 as z → ∞. Applying
Lemma 4, we deduce that the integral is equal to 2πi times the sum of the residues
of f on the upper half plane. Here we assume that f has no singularity on the real
axis. If it does have some singularities, we modify the path as indicated in Figure 2.
When the singularities on the real axis happen to be simple poles, Lemma 8 states
that the integral along the semi-circle is half of 2πi times the residue.

−R 0 R
Figure 2: Detour around a pole

´ +∞ dx ´
1 +∞ dx
2 Example. We evaluate 0 1+x6 = 2 −∞ 1+x6 .
Let C be the path consisting of the interval L on the real axis from z = −R
to z = R, followed by the semi-circle S of radius R whose center is at the origin
that joins z = R to z = −R. The integral of f (z) := 1/(1 + z 6 ) along this simple
closed curve is 2πi times the sum of the residues inside it.

12
−R 0 R
Figure 3: The path in Example 2

The singularities of f are poles of order 1 at the points z = e2πik/12 , k odd.

The ones on the upper half-plane are a = eπi/6 , b = eπi/2 , c = e5πi/6 . The residues
are
1 z 1
Res f (z) = = = − a,
z=a 6z 5 z=a 6z 6 z=a 6
1
Res f (z) = − b,
z=b 6
1
Res f (z) = − c,
z=c 6
hence
ˆ +R ˆ ˆ
f (x) dx + f (z) dz = f (z) ds
−R S C
−1
= 2πi
(a + b + c)
6
1 √ √
= 2πi · (−1) ( 23 + 21 i + i − 23 + 12 i)
6
2π
=
3
´
for R > 1. We take limits as R → +∞. Then S f (z) dz → 0 by Lemma 4 and we
´ +∞
get −∞ f (x) dx = 2π3 , so
ˆ +∞
dx π
6
= .
0 1+x 3

´ +∞
Integrals of type −∞
f (z)eiz dz
We suppose that f (z) is holomorphic on the upper half-plane except for a finite
number of poles and that f (z) → 0 as z → ∞ through the upper half-plane.
Assume first that f has no singularity on the real axis. Then we can apply
Lemma 7. The problem is reduced to finding the residues of f on the upper half-
plane. If f has some singularities, we modify the path as indicated in Figure 2.
Lemma 8 may be useful in this case.
´ +∞ cos x ´
1 +∞ cos x
3 Example. We find the value of 0 2
x +1 dx = 2 −∞ x2 +1 dx. The factor
cos x suggests that we take e /(z + 1) as the integrand. Put f (z) := 1/(z 2 + 1)
zi 2

and let C be the same path as in Example 2.

13
−R 0 R ´ +∞ ezi
Figure 4: Path of integration for −∞ z dz

We have f (z) → 0 as z → ∞ through the upper half-plane and f (z) has only
one simple pole on the upper half-plane, namely a simple pole at z = i. We find
ˆ +R ˆ ˆ
1 1 1
2
exi dx + 2
ezi dz = ezi dz
−R x +1 S z +1 C z2 +1
ezi ezi π
= 2πi × Res = 2πi = ,
z=i z2 + 1 2z z=i e

and taking limits of both sides as R → +∞, we find

ˆ +∞
1 π
exi dx + 0 = ,
−∞ x2 +1 e

and taking real parts of both sides, we get

ˆ +∞
cos x π
dx = ,
−∞ x2 + 1 e
ˆ +∞
cos x π
2
dx = .
0 x +1 2e
´ +∞ sin x
4 Example. We compute 0 x dx. The integrand is an even function and
´ +∞ sin x ´ +∞ exi
we are led to consider −∞ x dx, which is the imaginary part of −∞ x dx. So
we set f (z) := 1/z and observe that f has no poles on the upper half-plane and
zi
f (z) → 0 as z → ∞ through the upper half-plane. The singularity z = 0 of ez
prompts us to make a small detour about the origin by means of a semicircle σ of
a small radius ρ whose center is the origin. We orient σ positively. Then one part
of the path of integration is −σ, not σ.

Thus we have
ˆ −ρ ˆ ˆ +R ˆ
ezi ezi ezi ezi
dz + dz + dz + dz = 0,
−R z −σ z ρ z S z

ˆ   ˆ ˆ
+R
ezi e−zi ezi ezi
− dz + dz = dz,
ρ z z S z σ z
14
and we let ρ → 0+. By virtue of Lemma 8, we obtain
ˆ ˆ
+R
sin x ezi π−0 ezi
lim 2i dx + dz = 2πi Res
ρ→0 ρ x S z 2π z=0 z

ˆ +R ˆ
sin x ezi
2i dx + dz = πi,
0 x S z
and now we let R → +∞. By Lemma 7, we find
ˆ +∞
sin x π
dx = .
0 x 2

15