The Notebook Series

The integral theorems of

complex analysis with applications to
the evaluation of real integrals

by

R.S. Johnson
Professor of Applied Mathematics
School of Mathematics & Statistics
University of Newcastle upon Tyne

R.S.Johnson 2007

CONTENTS
List of integrals .
3
Preface .
4
1. Introduction .
1.1 Complex integration ..
Exercises 1

2. The integral theorems .

5
5
7

2.1 Green's theorem 8

2.2 Cauchy's integral theorem
8
2.3 Cauchy's integral formula ..
11
2.4 The (Cauchy) residue theorem
15
Exercises 2 .. 18

3. Evaluation of simple, improper real integrals .

19
3.1 Estimating integrals on semi-circular arcs .
20
3.2 Real integrals of type 1 ..
22
3.3 Real integrals of type 2 ..24
Exercises 3 27

4. Indented contours, contours with branch cuts

and other special contours
4.1 Cauchy principal value ..
4.2 The indented contour ...
4.3 Contours with branch cuts .
4.4 Special contours .
Exercises 4 .

28
28
31
34
39
42

5. Integration of rational functions

of trigonometric functions .
43
Exercises 5 .

Answers .
Index

46

47
48

List of Integrals
This is a list of the integrals and associated calculations that are discussed in this
Notebook.
Integral of f ( z ) = 2 z iz 2 along z = (t ) = t 2 + it , from t = 0 to t = 1.. p.6

z
z

f ( z ) dz for f ( z ) is: (a) z 2 ; (b) 1 ( z 2) ; (c) z 1 ..p.9

z + ez
1
dz p.13
dz ..p.11;
C 1+ z
C z ( z + 3)

Given f ( z ) = z 2 on C: z = r ei , 0 2 , find f ( z ) in interior of C.p.13

ze

z2 ez

2
C z z 1 ( z + 3)

f ( z) =

dz where C is z = 2 .p.17

satisfies zf ( z ) K ( R ) 0 on z = R , and identify K ( R ) ....p.20

1 + z2
1
f ( z) =
satisfies f ( z ) K ( R ) 0 on z = R , and identify K ( R ) ....p.22
1+ z

Evaluate:

0 1+ x

( x + a ) + b

z
z
2

Show that

Evaluate:

ln x

e x

dx

13
1 x

dx .p.22;

ze
z

dx

1 + x

cos( kx )
2

x2

dx .p.24;

2 3

x sin x

2
x + 4

.p.23

dx . ..p.26

exists ....p.29

sin x
dx p.31;
x

z
z

sin x

2
x 1 + x

dx ..p.33

x k
d
x
p.34;
dx with 0 < k < 1 ....p.37
2
2
1+ x
x
+
a
0
0

d
x
p.39;
cos x 2 dx ..p.41
x
0
1+ e
2

d
where 0 < k < 1 p.44;
1 + k sin

ze

e j

cos 2

4
0 1 5 cos

d . p.45

Preface
This text is intended to provide an overview of the main integral theorems that
are an essential element of complex analysis. This topic is covered in the degree
programmes offered in the School of Mathematics & Statistics at Newcastle
University; in this Notebook we take the opportunity to describe some of these
important ideas more fully. The material has been written so that it can be used as an
adjunct to a basic course in complex analysis, but it is not linked directly to a specific
module (although it naturally builds on the introduction provided by the standard
Stage 2 module on this topic). We present the material, particularly the applications to
integration, so that it can be used as a tool in a number of different modules; on the
other hand, it might simply help the reader to gain a broader experience of these
mathematical ideas. The aim is to go beyond the methods and techniques that are
presented in our modules, but all the standard ideas are discussed (and can be
accessed through the comprehensive index).
It is assumed that the reader has a basic knowledge of, and practical
experience in, complex numbers and line and double integrals (including Greens
theorem). To have attended an introductory course on complex analysis would be an
advantage, but this is not essential. We do not attempt to include any physical or
applied mathematical applications of these methods; this is properly left to a specific
module that might be offered in a conventional applied mathematics or engineering
mathematics or physics programme. Nevertheless, the applications that we do present
to the evaluation of real integrals will, we hope, be useful in many different
contexts.
The approach adopted here is, first, to provide proofs of the three main
integral theorems, and then, most particularly, to describe various applications to the
evaluation or real integrals, developed through a number of carefully worked
examples we present 23. A small number of exercises, with answers, are also
offered, although it must be emphasised that this notebook is not designed to be a
comprehensive text in the conventional sense.

Robin Johnson, February 2007

1. Introduction
Complex analysis, and particularly the theory associated with the integral
theorems, is an altogether amazing and beautiful branch of mathematics that
comfortably straddles both pure and applied mathematics. It provides the opportunity
to analyse and present in a very formal way, as well as to develop a powerful tool in
mathematical methods. The integral theorems take a staggeringly simple form, which
seems to run counter to all the experience gained by students familiar with
conventional integration methods. The consequence is that the results are very
straightforward to use, even though they describe deep and far-reaching ideas. In this
Notebook, we shall present, and prove, the three fundamental integral theorems:
Cauchys Integral Theorem and Integral Formula, and the Residue Theorem. These
results are then used to evaluate various types of improper integrals (using direct
methods, indented contours and regions with branch cuts) as well as integrals of
functions that are periodic on [0, 2 ] . As part of the essential background, we need to
define carefully what we mean by the integral of complex-valued functions along
curves in the complex plane; this is where we start.

1.1 Complex integration

We consider the differentiable function of a complex variable (i.e. an analytic
function)
f ( z ) = f ( x + iy ) = u ( x , y ) + iv ( x , y ) ,
for which therefore the Cauchy-Riemann relations hold:
u v
u
v
=
and
= .
x y
y
x
The aim is to define what it means to integrate f ( z ) along a curve in the complex
plane, but first we consider a simplified version of the essential problem, namely,
f (t ) = u(t ) + iv (t ) , where t is a suitable parameter. Thus we form, for t [a , b] ,

b
a

f (t ) dt =

u(t ) + iv (t ) dt

= u(t ) dt + i v (t ) dt ,
a

by invoking the linearity of the integral operator (and noting that i is a constant
independent of t).
Now suppose that, given f ( z ) , and a curve, C, described by z = (t ) , a t b ,
we wish to integrate f along the curve i.e. form a line integral in the complex plane.
We define this by using the familiar rule for the change of variable:

f ( z ) dz =

f (t )

d
dt .
dt

The curves, C, that we use may be simple, open curves i.e. they are not closed and do
not intersect, or more usually they will be Jordan curves i.e. simple, closed
curves.

Example 1
Evaluate the integral of f ( z ) = 2 z iz 2 along the curve z = (t ) = t 2 + it , from t = 0
to t = 1.
We have

e
j
= 2 x + 2 xy + ie y 2 x 2 + 2 y j ,

f ( z ) = f ( x + iy ) = 2( x + iy ) i x 2 y 2 + 2ixy

and z = x (t ) + iy (t ) = (t ) = t 2 + it i.e. x (t ) = t 2 and y (t ) = t on the curve. Thus

z
z

f ( z ) dz =

2t 2 (1 + t ) + i t 2 t 4 + 2t

0
1

j b2t + ig dt

4t 3 (1 + t ) t 2 t 4 + 2t + i2t 2 (1 + t ) + i2t t 2 t 4 + 2t dt

1
= t 4 + t 5 13 t 3 t 2 + i 2t 3 + t 4 13 t 6 = 23 + i 83 .
0

Comment: we observe that

f ( z ) dz =

ze

2 z iz 2 dz = z 2 13 iz 3 + C and so the value

of the integral from z = 0 (i.e. t = 0 ) to z = 1 + i (i.e. t = 1) becomes

z 2 13 iz 3

1+ i
0

= (1 + i) 2 13 i(1 + i) 3 = 2i 13 i( 2 + 2i) = 23 + i 83 .

This recovers the previous result because, in this example, the function

f ( z ) = 2 z iz 2 is an analytic function and so

f ( z ) dz has its conventional meaning.

We return to the original complex integral, and treat it as in Example 1, but now in
general:
I=

z
zm

f ( z ) dz =

C
t1

t0

u ( x , y ) + i v ( x , y ) dz

u x (t ), y (t ) + iv x (t ), y (t ) (t ) dt

7
on the curve z = (t ) , t0 t t1 . Further, let us write explicitly (t ) = x (t ) + iy (t ) ,
then

I=

zm

t1

u x (t ), y (t ) + iv x (t ), y (t )

r x(t ) + iy (t ) dt .

t0

Finally, this can be recast as line integrals in x and y:

zm

t1

I=

u x (t ), y (t ) x (t ) v x (t ), y (t ) y (t ) dt

t0

+i
=

zm

t1

u x (t ), y (t ) y (t ) + v x (t ), y (t ) x (t ) dt

t0

u ( x , y ) dx v ( x , y ) dy + i v ( x , y ) dx + u ( x , y ) dy .

This representation of the integral along a curve in the complex plane is the starting
point for the integral theorems.

Exercises 1
1. Evaluate the integral of the function f ( z ) = z 3 z 2 + i( z 2) along the curve

z = (t ) = 1 t + i t + t 2 from t = 0 to t = 1.
2. Confirm, by direct integration of

f ( z ) dz followed by evaluation, your answer

obtained in Q.1.

3. Repeat Q.1 for the function f ( z ) = z (the conjugate of z) from z = 0 to z = 1 + i

along the curves: (a) z = t + it , 0 t 1 ; (b) z = t 2 it 3 , 0 t 1 . (You should find
that the answers are different: f = z is not an analytic function of z.)

****************
**********

2. The integral theorems

The three theorems all involve Jordan curves (so simple closed curves, sometimes
called contours), but for three different types of function. The first case is the integral
of a function that is analytic inside and on the Jordan curve; in the second, the
function takes the form f ( z ) ( z z0 ) where f ( z ) satisfies the conditions of the first
case and z = z0 is a point inside the Jordan curve. The third and arguably the most
powerful result is essentially a generalisation of the preceding one, to a finite
number of singular points (usually poles) inside the contour. The first (the Cauchy
Integral Theorem) can be proved by a direct application of Green theorem, so we
provide a brief reminder of this.

2.1 Greens theorem

Let us be given a Jordan curve, labelled , which is mapped counter-clockwise;
the region interior to is labelled R. Further, we are given two functions, u( x , y ) and
v ( x , y ) , which possess continuous first partial derivatives in R and on . Although
we can work separately with u or v, it is usual to combine the pair particularly in the
light of the complex-valued integral that we obtained in 1.1. The theorem is then
expressed as
v u
u( x , y ) dx + v ( x , y ) dy =

dxdy ,
x y

zz FGH

IJ
K

which can be interpreted as a two-dimensional version of Gauss (divergence)

theorem. This is obtained by taking the divergence of the vector function (v , u) and,
of course, restricting the geometry to the 2D plane (but remember that Greens
theorem predates Gauss!). The circle on the line integral is used to denote a simple,
closed contour, normally mapped counter-clockwise.

2.2 Cauchys integral theorem

We shall provide a proof the classical one of this theorem. The function f ( z )
is necessarily an analytic function in the region R, and on the Jordan curve, C, that
bounds this region. (It is common practice to label curves in the complex plane as C,
whereas curves in the real plane are labelled .) Furthermore, we shall make the
additional assumption that f ( z ) is continuous in R and on C; we shall comment on
this second requirement later. We write
f ( z ) = f ( x + iy ) = u ( x , y ) + iv ( x , y )
and then

f ( z ) dz =

u ( x , y ) dx v ( x , y ) dy + i v ( x , y ) dx + u ( x , y ) dy ;
C

see 1.1. The two real line integrals that we have now generated are rewritten using
Greens theorem (all the conditions for which are satisfied):

u( x , y ) dx v ( x , y ) dy =

v ( x , y ) dx + u( x , y ) dy =

and

zz FGH

zz FGH

u v

dxdy .
x y

IJ
K

v u

dxdy
x y

IJ
K

But f ( z ) is an analytic function, so the Cauchy-Riemann relations hold i.e. ux = v y

and u y = v x throughout R (using subscripts to denote partial derivatives); so the two
double integrals above are zero, and hence

f ( z ) dz = 0 ,

which is Cauchys Integral Theorem (1825).

Example 2
The contour C is a circle of radius 1, centre at the origin, mapped counter-clockwise;
evaluate, where possible,
(c) z 1 .

f ( z ) dz , given that f ( z ) is: (a) z 2 ; (b) 1 ( z 2) ;

(a) The function f ( z ) = z 2 is analytic everywhere in the complex plane, so

immediately

z 2 dz = 0 .

1
is not analytic at z = 2 , but is analytic everywhere else,
z2
i.e. it is analytic for all 0 z 1 , so again dz ( z 2) = 0 .

(b) The function f ( z ) =

(c) Now the function f ( z ) = z 1 is not analytic at z = 0 , which is inside C, so we are

not able to use Cauchys integral theorem; we cannot (yet) find the value of the
integral.
Cauchys integral theorem requires only that f ( z ) be analytic (and f ( z )
continuous for our proof, but see later) inside and on the Jordan curve, C: any valid
Jordan curve will therefore suffice. This implies that, given any particular C, we may
deform C into any other Jordan curve, provided that inside and on the new curve,
f ( z ) satisfies the same conditions as just mentioned; on all such curves, we have

f ( z ) dz = 0 . Thus, even if f ( z ) is not analytic at points in the complex plane, any

contour that avoids them will still produce the zero value for the integral; we sketch
some examples below.

10

P
P

C2
C1

C3

The function in this example is not analytic (i.e. it is singular) at the point P in the plane; Cauchys
integral theorem applies on all three contours ( C1 , C2 , C3 ).

Indeed, we may deform the contour in a more precise fashion, as shown below:

where the two straight-line segments, L1 and L2, are parallel and equal in length. We
now close the gap between these two lines, and ensure that the inner contour so
produced encircles the singularity at P; when the lines coincide, the line integrals on
each cancel. This is simply because the integral (which exists the function is
analytic on C) in one direction is minus the value of the integral in the other. In the
limit, we obtain:

C2

C1

11

and we still have

f ( z ) dz = 0 where C = C1 + C2 , and the region, R, is that between

C1 and C2 . The totality of the contour, and its enclosed region, is conveniently
interpreted this way: as the contour is mapped out, so the region (R) is always on the
left. This is a fundamentally important choice of deformed contour, as we shall see in
2.2.

Example 3
The contour C is a circle of radius 2, mapped counter-clockwise, together with the
circle of radius 1, mapped clockwise, both centred at the origin; the region R is the
annulus between them. Evaluate

f ( z ) dz where f ( z ) = 1 z ( z + 3) .

The function f ( z ) is singular it has simple poles only at z = 0 and z = 3; it is

analytic everywhere else. These two points sit outside the annulus 1 z 2 , so
Cauchys integral theorem gives

f ( z ) dz = 0 .

We conclude this introductory section by making two general observations. Our

proof of Cauchys integral theorem requires that f ( z ) is analytic and that f ( z ) is
continuous on and inside C. However, E.J.-B. Goursat (1858-1936) proved in 1900
that Cauchys integral theorem is valid even if the condition on f ( z ) is relaxed: it is
sufficient that f ( z ) be continuous, and that f ( z ) exists, inside and on C.
The second point relates to a converse of Cauchys integral theorem. If f ( z ) is
continuous throughout a domain, D, in the complex plane, and if

f ( z ) dz = 0 on

every Jordan curve, C, in D, then f ( z ) is analytic in D. This is known as Moreras

theorem; G. Morera (1856-1907), an Italian mathematician, who proved this result in
1889.

2.2 Cauchys integral formula

We are given a function, f ( z ) , analytic inside and on the Jordan curve, C,
mapped counter-clockwise, and a point z = z0 interior to C; we consider the integral

f ( z)
dz .
z z0

The contour that we use for the purposes of evaluation is C as just defined, plus a
circle C0 , mapped clockwise, of radius with its centre at z = z0 . The circle must sit
wholly within C, which is always possible for z0 an interior point and a sufficiently
small (but non-zero) radius; the configuration is sketched in the figure below.

12

C0

Now by Cauchys integral theorem we have

f ( z)
dz = 0
z z0

C + C0

or

f ( z)
dz =
z z0

C0

f ( z)
dz ,
z z0

where the circle, C0 , is mapped clockwise; we describe the circle using the
parametric form z = z0 + e i for 2 0 . Thus we may write

C0

f ( z)
dz =
z z0

0 f z + e i
0

e i

ze
0

=i

j iei d

f z0 + e i d .

This integral can be evaluated and any evaluation will suffice by allowing 0
(which is allowed because f ( z ) is a continuous function), which gives

zb
z
0

b g

f z0 d = 2 i f z0

and so

b g

f ( z)
dz = 2 i f z0 ,
z z0

which is Cauchys Integral Formula (1831).

13

Example 4
Evaluate

z + ez
dz , where C is the circle z = 2 , mapped counter-clockwise.
C 1+ z

This is evaluated by a direct application of Cauchye integral formula: z = 1 is

inside C, so we have

z + ez
dz = 2 i e 1 1 .
C 1+ z

An illuminating and intriguing interpretation of Cauchys integral formula is made

more obvious when we write it as

f ( z) =

1
f ( )
d .
2 i z
C

That is, given an analytic function defined on C, f ( z ) is then known at every point
inside C. This result has no counterpart in the theory of real functions.

Example 5
Given that f ( z ) = z 2 on the contour C, defined by z = r ei , 0 2 , determine
f ( z ) throughout the interior of C.
We form f ( z ) =

1
2
d , where z is any interior point; thus we obtain on
2 i z
C

= re i :

2
2 r e i
1
f ( z) =
r iei d ,

i
2 i r e z
0

e j

and we note that z r ei for every (because z is interior to the circle). It is

convenient to rewrite the integrand as
r 3e3i
r ei z
and then we have

F
GH

2 2i

=r e

+ r ze

I
JK

z 2 r e i
i r 2 e 2i + r z e i + i
d
r
e

z
0

z 2 r ei
r ei z

14

= 21 r 2 e2i + r z ei + z 2 log r ei z

= z 2 log r ei z

j0

j0

= 2 i z 2 ,

by virtue of the jump in value of the logarithmic function across its branch cut. Thus
f ( z ) = z 2 on C and throughout the interior of C.

Before we turn to the most powerful and useful of these integral theorems, we
need one more result, which puts into a clearer perspective the identity

C0

dz
= 2 i ,
z z0

where C0 is a circle (or, indeed, any contour in this result) that encircles z = z0 . We
now consider the evaluation of
In =

zb

gn

z z0 dz

C0

where n = 0, 1, 2, 3, ... and C0 is the circle z = z0 + r e i , 0 2 ; note that the

case (omitted here) of n = 1 is evaluated by Cauchys integral formula. Thus we
have

In =

n in

r e

r ie d = i r

1+ n

LM ei(1+n) OP2
MN i(1 + n) PQ0

r 1+ n i2 (1+ n)
=
e
1 = 0 ,
1+ n
for every n 1 . Indeed, this makes clear just how special n = 1 is: in this case,
treating the problem as a conventional integral yields a logarithmic term, which
requires a branch cut (and a consequent jump in value) in order to evaluate it. In
summary, we have

In =

zb

C0

gn

z z0 dz =

RS2 i for n = 1
T0 for all other ns,

where C0 is a circle, centre z0 , mapped counter-clockwise.

15

2.3 The (Cauchy) residue theorem

The function, f ( z ) , is now assumed to have a finite number of singular points
inside the Jordan curve C; at each point, valid within an appropriate annulus about the
point ( z = z0 , say), it is assumed that f ( z ) can be expressed as a Laurent series i.e.
f ( z) =

an b z z0 g

n=0

+
n =1

bn

bz z0 gn

This always exists for a function that is analytic except at a finite number of discrete
singular points, each annulus being centred around each point, and not enclosing
another one. A function that possesses a Laurent series about a point at which the
terms in bn , i.e. the negative powers, do not terminate is said to have an essential
singularity at this point. A function that has Laurent series that terminates in the bn s
for every singularity has only poles (of a given order) and such a function is normally
called a meromorphic function. That is, a meromorphic function has no essential
singularities, but it does have poles; cf. analytic, which implies no singularities of any
sort. [Meromorphic comes from Greek ( and ), and means, literally,
part of the form/appearance, which is to be compared with holomorphic which is
sometimes used in place of analytic meaning whole of the appearance.]
The new theorem relates to the value of

f ( z ) dz , where C is as described

above; this situation is represented in the figure below.

C
z2

z1
z3

In this example, the contour C encloses three singular points.

To proceed, we use Cauchys integral theorem on a deformed contour; this is

constructed as in 2.2 so we deform around z1 (say), enclose this by an almostcomplete circle, and then close the circle. The contour otherwise is deformed around
all the other singular points, ensuring that they remain outside the contour. This is
represented in the figure below:

16

z1

z2

C1

C1

z3

The contour C1 encloses only z1 , which is itself is enclosed by a circle, C1 .

Cauchys integral theorem then gives

f ( z ) dz +

C1

f ( z ) dz = 0 ,

C1

where, as we have seen before, C1 is mapped counter-clockwise, but C1 is mapped

clockwise. Now choosing C1 to be inside the annulus around z = z1 , inside which the
Laurent expansion exists, enables us to write

R
|
n
n U
f
(
z
)
d
z
=
a
z

z
+
b
z

z
b
g
b
g
S

zC Cz T|n=0 n 1 n=1 n 1 |VW| dz

1

= 2 ib1 ,

g1

is absent, for then b1 = 0 ; b1 is called the residue of

and zero if the term in z z1
f ( z ) at z = z1 . Thus we have the evaluation

f ( z ) dz = 2 ib1 .

C1

Let us relabel the residue, so that it corresponds to the coefficient b1 at z = z1 , by

writing it as b11 ; the corresponding residue at z = zn is then b1n . This process of
forming circles around each singular point is continued by next encircling z2 , and
then z3 , and so on, each one contributing a term 2 i residue . Combining all the
contributions from the singular points inside C gives us
N
F
I
f
(
z
)
d
z
=
2

i
b

1
n
G
zC
H n =1 JK

17
for N singular points inside C; this is the Residue Theorem, sometimes called the
Cauchy Residue Theorem (1846).
It is clear that the residue theorem subsumes both Cauchys integral theorem and
integral formula. For, on the one hand, if the function is analytic so no singular
points anywhere then all the bn s will be zero for the Laurent expansions about
every point; hence the value of the contour integral will be zero: Cauchys integral
theorem. On the other hand, if the function to be integrated takes the form
f ( z ) = g ( z ) z z0 , where g ( z ) is analytic inside and on the contour and z = z0 is

b g

an interior point, then there is a one singular point inside C with a residue g z0 ,
which recovers Cauchys integral formula.

Example 6
Evaluate

ze

z2 ez

2
C z z 1 ( z + 3)

dz where C is the Jordan curve z = 2 , mapped counter-

clockwise.
2

z2 ez
The function f ( z ) =
has (simple) poles at z = 0, 1 inside C; the
z ( z 1)( z + 1)( z + 3)
pole at z = 3 is outside C and therefore does not contribute. In the neighbourhood of
each pole we have
at z = 0 : f ( z ) =...

1
1
... and so the residue here is ;
3z
3

at z = 1: f ( z ) =...

1 e
1 e
... and so the residue is
;
8
8( z 1)

at z = 1: f ( z ) =...

1 e
1 e
... and here the residue is
.
4
4( z + 1)

(These results can be obtained by observation, since no formal expansion is required

to determine the relevant terms.) The residue theorem then gives

ze

z2 ez

2
C z z 1 ( z + 3)

dz = 2 i

FG 1 + 1 e + 1 e IJ = FG 17 3 eIJ i .
H 3 8 4 K H 12 4 K

18

Exercises 2
1. Evaluate
(b) z = 3 .

2. Evaluate

sin z
dz where C, mapped counter-clockwise, is the circle: (a) z = 1 ;
C z+2

ze z cos z

C z z2 4

dz where C, mapped counter-clockwise, is the circle: (a)

z = 1 ; (b) z = r > 2 .

****************
**********

19

3. Evaluation of simple, improper real integrals

The first simple and direct application of complex integration to the evaluation of
real integrals is to improper integrals of the type

f ( x ) dx or

f ( x ) dx if f ( x ) is an even function.

In order to evaluate these integrals, we consider

f ( z ) dz

for a suitable choice of the contour, C. Since we eventually require the integral along
the real line, this (initially in the form R to R) must be included as part of C. The
most convenient way to accomplish this (but not exclusively so, as we shall see later)
is to use a contour which is the boundary of a semi-circular region of radius R,
normally taken to be in the upper half-plane:

The integral in the complex plane can therefore be written

f ( z ) dz =

f ( z ) dz +

f ( z ) dz

sc

where sc denotes the integral along the semi-circular arc; further, on the real line we
have z = x , so we may write

The procedure is to evaluate

f ( z ) dz =

f ( x ) dx +

f ( z ) dz .

sc

f ( z ) dz , using the residue theorem with the radius of

the arc sufficiently large to enclose all the singular points in the upper half-plane, to
estimate the integral along the arc and then to let R . In practice, the useful
results occur only if

sc

aspect of the problem.

f ( z ) dz 0 as R ; we now investigate this important

20

3.1 Estimating integrals on semi-circular arcs

We shall examine two cases: zf ( z ) 0 uniformly as R , and eikz f ( z )
( k > 0 and real) with f ( z ) 0 uniformly as R , both on the semi-circular arc.
By uniformly we mean the following: if g ( z ) K ( R ) , where R = z , and if
K ( R ) 0 as R , we say that g ( z ) 0 uniformly as R .
(a) Type 1
We are given that zf ( z ) K ( R ) with K ( R ) 0 as R ; we now consider
the integral

f ( z ) dz and construct an estimate for it:

sc

f ( z ) dz

sc

f ( z ) dz =

sc

f ( z ) R d .

But zf ( z ) = z f ( z ) = R f ( z ) K ( R) , and so we obtain

R f ( z ) d K d = K 0 as R ;

thus

f ( z ) dz 0 as R .

sc

Example 7
Show that f ( z ) =

We have zf ( z ) =

satisfies zf ( z ) K ( R) 0 on the semi-circular arc, as

1 + z2
R , and identify K ( R ) .

1 + z2
inequality, we have

, and on the semi-circular arc z = R ; but by the triangle

z2
1 + z2 + 1 z2 = R2

1 + z2
and so

z
1+ z

z
1 + z2

R
2

R 1

= K ( R) 0 as R .

21
(b) Type 2
This time we are given f ( z ) K ( R) 0 on the semi-circular arc, as R ;

the integral under consideration is

eikz f ( z ) dz where k is real and positive. (If k is

sc

complex-valued, then the imaginary part can be subsumed into the definition of f ( z ) ,
but the condition on the new f ( z ) must be unchanged.) We proceed in a similar
fashion to that adopted for type 1:

ikz

f ( z ) dz

sc

ikz

f ( z ) dz =

sc

e ikz f ( z ) R d ,

and note that

eikz f ( z ) = e ik ( x + iy ) f ( z ) = e ky f ( z ) e ky K ( R) .
Thus we may write

ikz

f ( z ) dz RK e ky d ,

sc

and on the semi-circular arc y = R sin , so we now require an estimate for

kR sin

d = 2

But a standard result is that

e kR sin d .

sin

or sin

, so that we have

e kR sin e 2 kR ,
and hence
2

kR sin

e 2 kR d =
e 2 kR
=
1 e kR .
0
2 kR
2 kR

When we combine these results, we obtain

sc

eikz f ( z ) dz 2 KR

1 e kR j = e1 e kR j K 0 as R ,
e
2 kR
k

22

and so we have proved that

eikz f ( z ) dz 0 as R . (This is sometimes referred

sc

to as Jordans lemma.)

Example 7
1
satisfies f ( z ) K ( R) 0 on the semicircular arc, as
1+ z
R , and identify K ( R ) .

Show that f ( z ) =

This is very straightforward, based directly on the triangle inequality:

1 + z + 1 z = R so

1
1
1
=

= K ( R) 0 as R .
1+ z 1+ z R 1

3.2 Real integrals of type 1

We explain the essential ingredients of the method by evaluating an improper
integral that is not elementary, although it does take a fairly simple form:

x2

4
0 1+ x

dx .

The complex integral that we consider is

f ( z ) dz with f ( z ) =

z2
1 + z4

je

this function has a denominator z 4 + 1 = z 2 + i z 2 i which has zeros in the upper

half-plane at z = 1 (1 + i), 1 1 + i . Thus we have, for R > 1, the following picture

2

The semi-circular region has two poles inside it, at z = 1 ( 1 + i ) .

2

and then it is convenient to write

23

z2

z2
.
=
1
1
1
1
1 + z4
z
(1 + i) z
(1 i) z
( 1 + i ) z
( 1 i )

LM
N

PQOLMN

PQOLMN

OPLM
QN

PQO

The residues at the two (simple) poles inside C are now easily obtained:
at z =

1
2

at z =

1
2

(1 + i) :

1 (1 + i) 2
2

di 2 id

id i

2 +i 2 i 2
1 ( 1 + i ) 2
2

( 1 + i ) :

d 2 id

i4 2

id i

2 +i 2 i 2

(1 + i) =

1
i4 2

1
4 2

(1 i) ;

( 1 + i ) =

1
4 2

(1 + i) .

We may express the contour integral as

z2

x2
z2
d
z
=
d
x
+
dz ,
4
4
4
1
+
z
1
+
x
1
+
z
C
R
sc
R3

0 as R : the condition for the

1 + z4 R4 1
type 1 integral is satisfied. We also have, by an application of the residue theorem,
and 1 + z 4 R 4 1 so that z

z2

C 1+ z

z2

dz = 2 i

LM 1 (1 i 1 i)OP =
N4 2
Q

Thus, letting R , we obtain

x2

1 + x

dx =

and then, because the integrand is an even function, we finally have the evaluation

x2

dx =
.
4
2 2
1
+
x
0

We now try one further example of this type.

Example 8

Evaluate

ze

dx

1+ x

2 3

24
We consider the integral (with C the boundary of the standard semi-circular region)

ze

C 1+ z

ze

dz

2 3

dx

1+ x

we see that 1 + z 2 R 2 1 and so

2 3

2 3

ze

dz

sc 1 + z

e1 + z j e R 1j
2

2 3

0 as R . We also

have z 2 + 1 = ( z + i)( z i) , which gives a pole (of order 3) at z = i in the upper halfplane; thus we write, with = z i ,
3
1
3 ( 2 i )
1 i 3
2
1
=
(
i
+
)
=

2
( z + i) 3 ( z i) 3 3
3
i 1
=
1 + 23 i 23 2 +.... ,
3
8

and thus the residue at z = i is 3 i . Hence, taking R , we obtain

32

ze

dx

1+ x

2 3

FG 3i IJ = 3 .
H 32 K 8

= 2 i

We are now in a position to consider some type 2 integrals.

3.3 Real integrals of type 2

This type of improper integral is nicely represented by this problem: find the value
of

cos( kx )

2
2
( x + a ) + b

dx ,

where k, a and b are real constants, and we take k > 0 , b > 0 . Although we could use
cos( kz ) (provided that the relevant conditions hold on the semi-circular arc), it is far
neater and more straightforward to replace cos( kx ) by eikz (and eventually take the
real part). Thus we consider

e ikz

eikx
e ikz
d
z
=
d
x
+
dz ,
2
2
2
2
2
2
(
z
+
a
)
+
b
(
x
+
a
)
+
b
(
z
+
a
)
+
b
C
R
sc

25
and on the semi-circular arc
2
( z + a ) 2 + b 2 = z 2 + 2az + a 2 + b2 z 2a z a 2 b2 = R 2 2aR a 2 b 2 ;

i.e.

f ( z) =

0 as R ,
( z + a ) 2 + b 2 R 2 2aR a 2 b 2

and the type 2 conditions are satisfied.

We have that ( z + a ) 2 + b 2 = ( z + a + ib)( z + a ib) which is zero in the upper
half-plane at z = a + ib , so we have, for R > a 2 + b 2 ,

The semi-circular contour has one pole inside, at z = a + ib .

and then the residue at z = a + ib is

e ik ( a + ib)
i
= e kb iak .
-a + ib + a + ib
2b

Hence we find

FG
H

IJ
K

i kb iak
kb iak
d
2

i
e
=
e
z
=

2
2
2b
b
(
z
+
a
)
+
b
C
e ikz

and so, letting R , we obtain

eikx

( x + a )

+b

dx =

e kb iak ;

when we then take the real part of this equation, we obtain

cos( kx )

dx = e kb cos(ak ) .
2
2
b
( x + a ) + b
In passing, we can note that in this example and this is typical of problems
interpreted by introducing eikz we also obtain the integral

26

sin( kx )

( x + a )

+b

dx =

e kb sin(ak ) ,

so two integrals are evaluated for the price of one calculation.

Let us now tackle another example of this type.

Example 9

Evaluate

x sin x

2
x + 4

dx .

To evaluate this, we consider the integral

zeiz

2
Cz +4

dz =

xeix

2
R x + 4

and note that z 2 + 4 z 2 4 = R 2 4 so that

dx +

ze i z

2
sc z + 4

z
2

z +4

dz ,

R
2

R 4

0 as R . But

we have z 2 + 4 = ( z + 2i)( z 2i) which is zero in the upper half-plane at z = 2i ; we

have the residue at z = 2i as

2ie 2 1 2
= 2e .
4i

Thus

zeiz
2

Cz +4

dz = 2 i 21 e 2 = i e 2 ,

and then, with R , we get

xe ix
2

x +4

dx = i e 2 ;

on taking the imaginary part, this produces the required evaluation:

x sin x
2

x +4

dx = e 2 .

Comment: In this example, we see that the real part gives

27

x cos x

2
x + 4

dx = 0

which is no surprise because the integrand is an odd function. Thus, although we can
use this method to find

x sin x
2

0 x +4

we are unable to find

x cos x

2
0 x +4

dx ( = 21 e 2 )

dx (even though this is expected to exist and be non-

zero).

Exercises 3
Evaluate these real integrals:

(a)

ze

1 + x

dx
2

je x + 4j
2

; (b)

ze

dx

0 1+ x

2 2

; (c)

****************
**********

cos x

4
x + 4

dx .

28

4. Indented contours, contours with branch cuts

and other special contours
The examples described in the previous chapter have enabled us to introduce the
basic principles that apply to the evaluation of real integrals, using these techniques,
although all the problems have involved a semi-circular region in the complex plane.
However, any contour could be chosen and the consequences explored, which may
lead to a suitable method of evaluation but it may not! In this chapter we will
present a few examples that require different choices of contour, some which turn out
to be an adjustment of the classical semi-circular one, but others are very different.
Indeed, we shall find that, in order to evaluate certain integrals, we may use the semicircle, but with indentations. On other occasions, the integrand itself can be defined
only by the inclusion of branch cuts, and this must be accommodated by the chosen
contour. Finally we shall show that, for other evaluations, some very special contours
must be used. However, we typically encounter some technical problems (associated
with the definition and existence of integrals, even though the original real integral is
well behaved); this aspect needs to be addressed first.

4.1 Cauchy principal value

In order to tackle integrals such as

sin x
dx
x

we shall discuss two difficulties that stem from the consideration of the integral of

eiz z . For example, it is immediately clear that sin x x is integrable in the

neighbourhood of x = 0 (for sin x x 1 as x 0 ), whereas eiz z does not exist at
z = 0 . The original real integral does exist, although confirming this by examining the
behaviour at infinity is not straightforward; thus the simple estimate sin x 1 leads to
the integral of 1 x , yielding ln x which diverges as x .
The familiar choice of a semi-circular region does work for this example, using
the integrand eiz z , which satisfies the requirements of a type 2 integral (3.3) at
infinity. However, we must define what we mean by the integral of a function that
possesses the property of this one near z = 0 (which, we must expect, should not be
critical to the evaluation of the real integral because this does exist). We first address
the problem of developing a suitable definition, and then we will see how we can
incorporate this within a formulation of an integral in the complex plane.
The situation that we must clarify is best described by reminding ourselves of the
definition of an improper integral, where the failure to be proper is because the
integrand is not defined at a point in the range of integration. A simple example is the
integral

y k dy , for 0 < k < 1 and x > 0 ,

29
where y k does not exist at y = 0 . The value of this integral (if it exists) is defined
by

F x k I
G z y dyJJ ;
0 GH
K
lim

this gives

F L y1 k O x I
F 1
J
lim G M
= lim G
P
0 GH MN 1 k PQ JK 0 H 1 k
+

x1 k 1 k

IJ = FG 1 IJ x1 k
K H 1 k K

because 1 k 0 with 0 < k < 1 : the integral exists. Let us now suppose that we
require the integral of f ( x ) , for x a , b , where f ( x0 ) , a < x0 < b , is undefined;
the integral exists if

F b
I
F x I
lim G z f ( x ) dxJ + lim G z f ( x ) dxJ
JK 0 GH x + JK
0 GH a
0

is finite. The use of two parameters is essential here, making clear that the processes

0+ and 0+ are independent.

Example 10

Show that the real integral

dx

exists (where, of course, it is necessary that all

13
1 x

values of x k , for suitable k, are taken to be real).

The integral is defined as

F 2 1 3 I
F 1 3 I
lim G z x
dxJ + lim G z x
JK 0 GH dxJJK
0 GH 1
I
2I
F
F
= lim G 23 x 2 3 J + lim G 23 x 2 3 J
1 K 0 H
K
0 H
+

e2

e2

= lim 3 ( ) 2 3 3 ( 1) 2 3 + lim 3 2 2 3 3 2 3
0+

0+

= 23 2 2 3 1 .

dx
fails because
1 x
the integrand, and hence the integral, are not real for x < 0 . In this case we can allow
only the one-sided limit
Comment: We should note that the corresponding argument for

30

F 2 dx I
lim G z
J.
0 GH x JK
+

Let us now apply this same approach to the integral

dx
,
x

so we consider

F 2 dx I
F dx I
+
lim
Gz J
Gz J
0 GH 1 x JK 0 GH x JK
I
= lim F ln x
H 1 K +
lim

FH ln x 2 IK
0
0
= lim bln ln 1g + lim bln 2 ln g .
0
0
+

lim

However, ln and ln increase indefinitely in size as 0+ and 0+ ,

respectively: the integral does not exist, according to the familiar definition (and this
result should be no surprise). But there is something rather special about this example;
if we allowed = , then we obtain

lim ln ln 1 + ln 2 ln = ln 2 ,

0+

and the integral exists! Nevertheless, we should be aware that even this manoeuvre
does not always work; consider

dx

1 x

which we write as

2
F dx 2 dx I
F
1O
1O I
L
L
F 1 3 1I
lim G z 2 + z 2 J = lim G M P + M P J = lim G + J ;
J
N x Q 1 N x Q K 0 H 2 K
0 GH 1 x
x K 0 H
+

this does not exist the two terms in do not cancel.

The Cauchy Principal Value, when it exists, is defined by
b
F x
I
lim G z f ( x ) dx + z f ( x ) dxJ
JK
0 GH a
x +
0

31

and this value is usually represented by a bar through the integral sign,

f ( x ) dx , or

by writing

PV f ( x ) dx .
a

Of course, a function may possess more than one point where it does not exist, so the
principal-value definition must be applied to each one. We also record that the
definition can be extended to an integral that is improper because the limits extend to
infinity. So the integral

x dx clearly does not exist,

and neither does

x dx ; on the other hand, the Cauchy principal value of this latter

integral is defined as

FR I
F R IJ = lim e 1 R2 1 R 2 j = 0 ,
lim G z x dxJ = lim G 21 x 2
JK RH R K R 2 2
R GH
R
which does exist.

4.2 The indented contour

We return to the real integral that we introduced above, and use this as a vehicle to
describe and explain how contours are indented. Thus in order to evaluate

sin x
dx we consider
x

e iz
dz ,
z

which satisfies the type 2 conditions at infinity, but is undefined at z = 0 . Thus we

take as the contour, C, a semi-circular arc together with the diameter along the real
axis indented by a (small) semi-circular arc that allows us to avoid z = 0 , as shown in
the figure below.

The contour comprising two semi-circular arcs (radii R and ) and

an almost-complete diameter connecting them.

32
According to Cauchys integral theorem, we have

e iz
dz = 0 ,
z

because the only singularity of eiz z is a simple pole at z = 0 which lies outside the
contour. But we may write the integral on C as

e ix
e ix
dx +
dx +
x
x

sc

eiz
e iz
dz +
dz = 0 ,
z
z
sc

where sc labels the integral on the semi-circular arc of radius (mapped

clockwise), and sc is our familiar label for the larger semi-circle. We know that

sc

e iz
dz 0 as R
z

(cf. Example 7), but we do not know the behaviour on sc (although we might
surmise that is it 1 2 i residue = 1 2 i 1 = i ).
2

On the smaller semi-circular arc, we have z = ei , 0 , so we obtain

0 exp i e i
e iz
dz =
ie i d
i

z
e
sc

= i exp i e i d

i d = i as 0 ;

its value is indeed i ! Thus we may write, once we have taken 0 and R ,

eix
dx = i ,
x

the principal value being necessary because we have taken 0 about x = 0 . The
imaginary part of this equation yields

33

sin x
dx = ,
x

which we may write like this because the integral does exist in the conventional sense.
On the other hand, when we take the real part, the principal-value notation must be
retained to give

cos x
dx = 0 ;
x

this integral certainly does not exist in the conventional sense it is not integrable at
x = 0 but it does in the PV sense.

Example 11

Evaluate

sin x

2
x 1 + x

dx .

ze

We consider the integral

eiz

2
C z 1+ z

dz , which requires an indented contour around

z = 0 , exactly as in the example above. However, we also have simple poles at

z = i , and we may note that, on the semi-circular arc of radius R:

z 1 + z2

1
1
0 as R .
R R2 1

The residue at z = i is e 1 i.2i = 1 e 1 , and so we obtain

2

ze

Cz

e iz

1+ z

dz = 2i 21 e 1 = i e 1 .

However, the integral along C can be written as

where

x 1+ x

ze

sc z

e iz

1+ z

ze

e ix

dx +

eix

x 1+ x

dx +

sc

eiz

z 1+ z

dz +

ze

sc z

eiz

1+ z

dz = i e 1 ,

dz 0 as R . On z = ei , 0 , we obtain

34

exp i ei
i

e1 + 2e2i j

i e d = i

0 exp i e i

j d i as 0 .

2 2i

1+ e

Thus, letting R and 0 , we see that

e ix

x 1+ x

dx = i i e 1 ,

and then the imaginary part yields

sin x

x 1+ x

Comment: This same complex integral gives

part), although the integral

cos x

2
x 1 + x

dx = 1 e 1 .

cos x

2
x 1 + x

dx = 0 (by taking the real

dx does not exist in the conventional sense.

4.3 Contours with branch cuts

The logarithmic function, log z , is the function that is the most familiar one with a
branch cut. It is usual to take the cut along the negative real axis, thereby defining the
principal value as

b g

Logz = PV log z which is log z for < arg( z ) .

A suitable interpretation is necessary for the evaluation of integrals such as

ln x
2

2
0 x +a

dx ,

where a is a positive, real constant, when treated as a corresponding integral in the

complex plane. The natural way to attempt an evaluation is to replace ln x by log z ,
and then by Logz (to make it single valued), together with a suitable contour; so we
consider
Logz
dz .
2
2
z
+
a
C

35
The choice of C requires some care, when we note the existence of the branch cut

necessary for Logz . We have a simple pole at z = ia = a ei 2 in the upper half-plane,

and we anticipate that the integral along the semi-circular arc tends to zero as the
radius increases, so we use the C shown below.

The two semi-circular arcs extend from = 0 to = ; the radii of the arcs are R and , and
they are joined along = by the straight line L. The branch cut is along the negative real axis.

First, we write
Logz
z2 + a 2

Logz
,
( z + ia )( z ia )

and then the residue at z = ia becomes

i 2

Log(ia ) Log ae
i
ln a
=
=
ln a + i 2 =
i
;
2 ia
2 ia
2a
4a
2a

the residue theorem now gives

FG
H

IJ
K

ln a

2
d
z
=
2

i
=
ln
a
+
i
.
2
2
4a
2a
a
2a
z
+
a
C
Logz

The contour integral can be expressed as

Logz
2

Cz +a

dz =

ln x
2

x +a

dx +

Logz
2

sc z + a

dz +

Logz
2

Lz +a

dz +

Logz
2

2
sc z + a

dz ,

where sc and sc denote, here, the almost-complete semi-circular arcs (see the figure
above), and L is the line z = r ei( ) , r R . On the larger semi-circular arc we
have

R Logz R (ln R) 2 + 2

=
0 as R ,
z2 + a 2 R2 a2
R2 a2
zLogz

36
(because 0 ) which therefore satisfies the type 1 condition. On the smaller
semi-circular arc, we have

0 Log

e e i j iei d = i 0 bln + i gei d 0 as 0 ;

2 2 i

2 2 i

a + e

+ a2

thus, with R and 0 , we are left with

Logz

2
2
Cz +a

dz =

ln x

2
2
0 x +a

dx +

Logz

2
2
bc z + a

F = ln a + i 2 I .
GH a
2a JK

dz

But on the branch cut (bc) we have z = r ei , > r 0 , so we may write

Logz
2

bc z + a

dz =

ln r + i
2

r +a

( 1) dr =

and this latter integral is elementary:

ln r
2

0 r +a

dr + i

LM
N

dr
2

2
0 r +a

gOPQ

= arctan r a
=
.
2
2
a
2a
r
+
a
0
0
dr

Thus, finally, we have

ln x
2
2
d
x
+
d
r
+
i
=
2
d
x
+
i
=
ln
a
+
i
,
2
2
2
2
2
2
2a
2a a
2a
x
a
r
a
x
a
+
+
+
0
0
0
ln x

ln r

and so the required evaluation is

ln x
2

0 x +a

dx =

2a

ln a .

One perhaps rather surprising outcome of this calculation is the special case
a = 1:

ln x

ln x

2
0 x +1
1

and then an interpretation of this is

0 x +1

dx = 0 ,

dx =

ln x

2
1 x +1

dx .

Another fairly common appearance of branch cuts (because it is directly

associated with the logarithmic function) is in the evaluation of z k for arbitrary k; we
will investigate this case in the next example.

37

Example 12
k

Evaluate

x
dx for real k with 0 < k < 1 . (This integral is related to the Beta
1+ x

function.)
In order to evaluate this integral, we consider

zk
dz , for a suitable choice of the
1+ z

contour C; we must take the principal value of z k defined by

z k = exp k (ln r + i ) ,
with the cut along = 0 (because we have a pole at z = 1 i.e. on = ), so we use
0 < 2 . The contour we use is

where the outer, almost-complete circle is of radius R and the corresponding inner one

is of radius (for 0 < < 1 ); the two straight lines, L1 and L2, are z = r e i for

> 0 and r R . The singular point is at z = ei = 1 , which lies between the

two almost-complete circles. This type of curve is often called a keyhole contour. The
residue at the (simple) pole is exp k (ln 1 + i ) = e i k , and then the residue theorem
gives
zk
dz = 2 ie i k .
1+ z

But we may write

38

zk
dz =
1+ z

CR

zk
zk
dz +
dz +
1+ z
1+ z
c

zk
dz +
1+ z

L1

L2

zk
dz ,
1+ z

where CR denotes the almost-complete circle of radius R (mapped counterclockwise), and c is the corresponding inner circular arc (mapped clockwise). On
CR we have
z

zk
R1 k

0 as R ,
1+ z
R 1

and so

CR

zk
dz 0 as R ;
1+ z

also, on c , where z = ei , 2 , we may write this integral as

k e i k
ei(1 k )
1 k
i
ie
d
=
i
d 0 as 0 .

i
i

1
+
e
1
+
e
2
2
Finally, on L1 and L2 (and we note the directions along these lines), we have

r k e i k ( 2 ) i( 2 )
i
e
d
r
+
e
dr
i
i ( 2 )
1
+
r
e
1
+
r
e

R k i k

=e

i (1 k )

r k

dr e

i
1 + re

1 e i 2 k

i (1 k )( 2 )

r k

i ( 2 )
1+ re

dr

j z 1r + r dr as 0 .
R k

Then, collecting all these results together, and taking 0 , 0 and R , we

obtain

zk
dz = 1 e i2 k
1+ z

j z 1x+ x dx = 2 ieik .
0

Thus

ee

i k

i k

j z 1x+ x dx = 2 i ,
0

and with ei k e i k = 2i sin( k ) , the required value is

39

dx =
1+ x
sin( k )

(for 0 < k < 1 ).

4.4 Special contours

We conclude with two examples that require special choices for the contour. We
have become rather familiar with the semi-circular contour, or some suitable
refinement of it; indeed, it is by far the most common choice, but for some functions it
is altogether inappropriate.

(a) A rectangular region

We consider the problem of evaluating

e x

dx ,

x
1+ e

where might be a complex constant, but such that 0 < ( ) < 1 ; it is clear that this
condition on is necessary in order to guarantee the existence of the integral. (Note
the behaviour of the integrand as x .) We shall take the case of the real integral,
so will be real here. We introduce the integral

e z

z
C 1+ e

dz ,

and observe that 1 + e z = 0 for z = i 1 + 2n , for n = 0, 1, 2, ... , so that a semicircular contour which is extended to infinity will necessarily enclose an increasing
number of poles. Thus we select a rectangle that encloses only one pole, at z = i ,
and so we take C as the rectangle ( 2 R 2 ) shown in the figure below.
2

The only pole inside C is at z = i , with the residue obtained by writing z = i + :

e z
1+ e

i +

1+ e

i +

i +

1 e

ei 1 + +...
=
,
1 1 + +...

40

and so the residue at z = i (i.e. at = 0 ) is ei . Thus, by the residue theorem, we

have

e z

C 1+ e

dz = 2 i ei = 2 iei .

Now we may write the contour integral as

e x

R 1+ e

dx +

2 ( R + i y )

0 1+ e

idy +

R +i y

However, we note that

R 1+ e

2 R + i y

R x + 2 i

R+i y
0 1+ e

x + 2 i

idy

dx +

2 1 + e

R +i y

R+i y

e R
eR 1

R+i y

dy

2 0 as R

for 0 < < 1 ; similarly

R+i y

idy
1 + e R +i y

Finally, we also have

R x + 2 i

dx = e
1 + e x + 2 i

R+i y

R +i y
0 1+ e

e R
1 e R

2 i

dy

2 0 as R .

e x

x
R 1+ e

dx ,

and so, taking R , we obtain

e1 e j z 1 + e x dx = 2 iei
2 i

e x

or e

j z 1 + e x dx = 2 i

e x

idy = 2 iei .

R +i y
0 1+ e

41

b g z 1e+ e x dx = 2 i .

i.e. 2i sin

Hence we have the evaluation

e x

dx =
,
x
sin
1+
e

b g

where is real, with 0 < < 1 .

This result should be no surprise: cf. Example 12. Let us write y = e x , with
< x < (so that 0 < y < ), then we obtain

1
k
y 1
y
y

d
x
=
d
y
=
d
y
=
d
y
=
x
1+ y y
1+ y
1+ y
sin( k )
1 + e
0
0
0

e x

from Example 12, where k = 1 (so 0 < k < 1 ) and also

sin( k ) = sin (1 ) = sin( ) .

(b) A sector of a circle

This is the problem of evaluating

e j

cos x 2 dx , and the obvious choice for a

e j

function in the complex plane is exp iz 2 (so that the real part for z real is the
required function); thus we consider

e i z dz .

The most convenient and straightforward method for the evaluation of this integral
(which we may note is the integral of an entire function) is to incorporate the standard
result

y2
e
dy
0
i 4

= 21 . This can be accomplished by taking, as one part of C, the

line z = r e
, for then i z 2 = i r 2 e i 2 = r 2 ; thus we choose to use a C as shown in
the figure below.

4
R

42
By Cauchys integral theorem, we have

e i z dz = 0 ,

but we may write

iz2

dz = e

i x2

dx +

2 2i

exp i R e

jRie

d + exp ir 2 ei 2 ei 4 dr ( = 0) .
R

Here, we see that

4

iR

2 2i

e exp i R e

j d R z e

R sin 2

d R

2
e 4 R d

because ei = 1 , exp i R 2 e2 i = exp i R 2 cos 2 + i sin 2

= exp R 2 sin 2 and

sin 2 4 for 0 4 . Thus we have

4

iR

F
H

e i exp i R 2 e2i d 41 R 1 1 e R

I 0 as R ,
K

and so when we take R , we are left with

i x2

dx e

i 4

r 2

dr = 0 or

i x2

dx =

1 (1 + i)
2

2
e r dr = 1 (1 + i) 21 .

The real part of this equation then gives

1
cos x 2 dx =
,
2 2

e j

which is the required value. (Note that the corresponding integral for sin yields, from
the imaginary part, the same value:

e j

sin x 2 dx = 2 2 .)

Exercises 4

Evaluate these real integrals: (a)

(ln x ) 2

dx ; (b)
2
1
+
x
0
0

x sin x
x3

****************
**********

dx .

43

5. Integration of rational functions of

trigonometric functions
The final type of integral that we consider takes a rather different form; indeed,
the problem and the approach to its solution harks back to standard methods of
elementary integration: substitution. Essentially all we do is to introduce a routine
change of variable the substitution and then integrate, except that here we generate
an integral in the complex plane that can be evaluated by using the residue theorem.
We should comment that all such integrals can be evaluated by conventional means,
i.e. by using a standard (real) substitution, but the definite integrals that arise are far
more easily computed by these new methods. We shall consider definite integrals of
the form
2

f (sin ,cos ) d ,

where f is a rational function of its arguments; a simple example is

2

sin
d .
3 + 2 cos

The method involves using the familiar identification z = ei , so that 0 2

will map out the unit circle, z = 1 , in the counter-clockwise direction; we will label
this contour C0 . We also have

j e
sin = 21i eei e i j = 21i e z z 1 j ,

dz
= ie i = i z , cos = 21 e i + e i = 21 z + z 1 ,
d

and so we may write

2

f (sin ,cos ) d =

C0

e e

j e

f 21i z z 1 , 21 z + z 1

jj FGH i dzz IJK .

But since f is rational, it remains rational under this transformation to z, and hence we
can readily identify the poles (and the residues) inside C0 , the unit circle.

Example 13
2

Evaluate

d
where 0 < k < 1 (and k is real). (The condition on k is
1 + k sin

necessary if the integral is to exist.)

44

We introduce z = ei with sin = 21i z z 1 , so we have

2

d
=
1 + k sin
=

z
z

1
C0 1 2 i k

ez z

FG i dz IJ
jH zK

dz

2
1
C0 i z + 2 k z 1

F
H

b g

2
k

dz
2

2i
C0 z + k z 1

I
K

where z 2 + 2i k z 1 = 0 at z = ki 1 1 k 2 . The root with the positive square

root corresponds to a point inside the unit circle; the other point lies outside. Thus we
write
1
2

z + 2ki z 1

F z + ki ki
H

1 k 2

I F z + ki + ki
KH

1 k 2

I
K

and so the residue (of this function) at z = i + i 1 k 2 is

k

1
ki + ki 1 k 2 + ki + ki 1 k 2
Thus we finally obtain

F
GH

ik
1
.
2 1 k 2

I
JK

2
k
1
2
d
= 2 i i
=
.
2
1 + k sin k
2 1 k 2
1
k

It is convenient, and often very useful, to note that sin n and cosn can be
expressed as rational functions of sin and cos , so more involved trigonometric
terms can appear in the integrand, without causing as we shall see undue algebraic
complications (which might have been expected). So, given ei = cos + i sin , we
have

ei n = cos + i sin

gn = cos n + i sin n ,

which is the very familiar de Moivres theorem; then we may write

j e

j e

cosn = 21 e in + e i n = 21 z n + z n
and

j
j

sin n = 21i e in e i n = 21i z n z n ,

45
which considerably simplifies the process of substitution. [A. de Moivre, 1667-1754,
French mathematician who developed the field of analytical trigonometry; he was
severely persecuted for his Protestant faith.]

Example 14
2

Evaluate

ze

cos 2

2
4
0 1 5 cos

d .

We introduce z = ei with cos = 21 z + z 1 and cos2 = 21 z 2 + z 2 ; thus we

obtain
2

ze

cos 2

4
0 1 5 cos

d =
2

C0

j FG i dz IJ
2 H
zK
1 25 e z + z 1 j
1 z 2 + z 2
2

i
2

z 3 + z 1

2
C0 4 z 5 z + 1
25
2

dz .

e j

Now z 2 25 z + 1 = ( z 2) z 21 , so we have one root of this quadratic expression that

lies inside the unit circle (at z = 1 2 ). Thus it is convenient to write
2

ze

cos 2

25
1 + z4

d
=

i
dz
2
2
8
2
4
1
0 1 5 cos
C0 z ( z 2) z 2

e j

which has poles at z = 0 and at z = 1 2 inside the contour; the residue at z = 0 is

25
1
25
i
= i.
2
2
8 ( 2 ) ( 1 2 )
8

The residue at z = 1 2 is obtained by writing z = 1 + , then we obtain

2

e
je

4
1 + 21 +
25
i
8 1 + 3 + 2 2
2
2

jb

ge

25 8 17 1
i 2 16 + 2 +... 1 2 +... 1 + 43 +...
8 9

46

=...

25 1 1 17 17
i
+
+ ... ,
9 2 8 12

and so the residue at z = 1 2 is

25 1 17 17
125
i 2 8 + 12 =
i.
9
216

Thus we finally have the evaluation

2

ze

FG
H

cos 2

IJ
K

25 125
275

.
d
=
2
i

i
+
i
=
2
8
216
54
4
0 1 5 cos

Exercises 5
Evaluate these real integrals:
2

(a)

ze

2 2
0 1 + 3 cos

; (b)

d
( a 2 + b 2 < 1 ).
1 + a sin + b cos

****************
**********

47

Answers
Exercises 1
97 13
i .
12
6
1
3. (a) 1 ; (b)
(1 + i) .
10

1. & 2.

Exercises 2
1. (a) 0; (b) 2 i sin 2 .

2. (a) 21 i ; (b) 21 i 1 + 2 sinh 2 cos 2 .

Exercises 3
(a)

; (b)

; (c)

4e

bsin 1 + cos1g .

Exercises 4
(a)

3
8

; (b)

Exercises 5
(a)

5
; (b)
8

1 a + b

48

Index
analytic function
Beta function
branch cut
Cauchy principal value
Cauchy residue theorem
Cauchy-Riemann relations
Cauchy's integral formula
Cauchy's integral theorem
contour

5
37
14,34
28
15
5,9
11
8
8
deform
9
indented
28,31
keyhole
37
rectangular
39
sector of a circle
41
with branch cut
34
curve
closed
6
Jordan
6
simple
6
de Moivre's theorem
44
deform contour
9
entire function
41
essential singularity
15
formula
Cauchy's integral
11
function
analytic
5
Beta
37
entire
41
holomorphic
15
meromorphic
15
rational, of trigs
43
Gauss' theorem
8
Goursat, E.J.-B.
11
Green's theorem
8
holomorphic function
15
improper real integrals
19
indented contour
28,31
inequality
triangle
20
integral
on semicircle
19,20
integrals
improper, real
19
Jordan curve
6,8
Jordan's lemma
22
keyhole contour
37
Laurent series
15
line integral in complex plane
5
logarithmic term
14
meromorphic function
15
Morera's theorem
11
pole
15
principal value
28
PV
31

49
rational function of trig fns
real integrals

rectangular contour
residue
residue theorem
sector of a circle
semi-circular arc
series
simple, closed curve
singularity
theorem

triangle inequality

43
19
type 1
22
type 2
24
39
16
15
41
integral along
19,20
Laurent
15
6
essential
15
Cauchy's integral
8
de Moivre's
44
Gauss'
8
Green's
8
Morera's
11
residue
15
20

50