11

Examples of Contour Integrals

dx
1 + x2
0
First, recognise that the integrand is even so that we can express this as 1/2 the integral along the
whole real line:
Z
1 dx
2 1 + x2
To apply the residue theorem, we must deal with closed contours. So, let us construct a closed
contour composed of the real line extending from x = R to x = +R, C0 , and add the semi-circle
C1 in the upper-half complex plane, of radius R, centre the origin. We will eventually let R .
Call this closed contour C = C0 + C1 .

Now consider the contour integral, to which we wish to apply the residue theorem,
Z

I=
C

dz
1 + z2

First, consider the effect of adding the semi-circular part. On C1 , we can set z = R exp(); then
dz = R exp()d. The integral becomes
Z
0

Re d

2
2
1+R e
R

Z
e d
0

e2

as R . In other words, the semi-circular contour contributes zero to the integral. Returning
to the contour integral
Z
Z
dz
dz
I=
=
2
1
+
z
(z

)(z
+ )
C
C
we recognise that the integrand possesses two simple poles at z = and z = . The latter pole
is outside of C so does not contribute. The residue of the enclosed pole at z = is 1/2. Hence,
applying the residue theorem, we have that I = 2/2 = . Finally,
Z
0

dx

=
2
1+x
2

1
dx
1
+
x4

Create a closed contour C as in the previous example and evaluate

I=

Z
C

1
dz
1 + z4

using the residue theorem. First, we check that the contribution from the added semi-circular part
of C contributes zero to the integral: following the procedure of the previous example, we can see
1

that this now varies as 1/R3 and so tends to zero as R . The integrand clearly has poles where
the denominator is zero: we need the find the zeros of z 4 = 1. Let
z 4 = 1 = e = e+2n
where n = 0, 1, 2, 3. Then
zn = e/4+2n/4
i.e.
1
z0 = e/4 = (1 + )
2
1
z1 = e3/4 = (1 + )
2
1
z2 = e5/4 = (1 )
2
1
z3 = e7/4 = (1 )
2
Note that only z0 and z1 are enclosed within the contour. Now, we can factorise
z 4 + 1 = (z z0 )(z z1 )(z z2 )(z z3 )
so that

I=
C

1
dz
(z z0 )(z z1 )(z z2 )(z z3 )

The residue of the pole at z = z0 is

1
2
1
=
=
(z0 z1 )(z0 z2 )(z0 z3 )
4( 1)
2 2(1 + ) 2
Similarly, residue at z = z1 is

1
1
2
=
=
(z1 z0 )(z1 z2 )(z1 z3 )
4(1 + )
2 2 2(1 + )
Then, applying the residue theorem,


2
1
1

I = 2
+
=
4 1 1+
2

Z 2

I=
0

d
1 2p cos + p2

where 0 < p < 1. Note that this is a real-valued integral (and hence should have a real-valued
outcome). Whenever we see an integral over the angles 0 2, we should think immediately as one possible option - of converting this to a contour integral around the unit circle C in the
complex plane by setting z = exp(). Then, dz = exp()d and cos = (z + 1/z)/2. Therefore,
dz/z
1

p(z
+ 1/z) + p2
C
Z
dz
=
2
2
C pz (1 + p )z + p
Z
dz

=
p C (z p)(z 1/p)
Z

I =

Since p lies between zero and one, there is an enclosed simple pole at z = p whose residue is
1
p 1/p
Hence, invoking the residue theorem,
I = 2

1
2
=
p p 1/p
1 p2

Interesting example: The Laplace Transform.

This is rather akin to the Fourier Transform but involves an interesting contour.
We define the Laplace Transform of the function f (t) to be
Z

F (p) =

f (t)ept dt

The inverse tramsform is

1
F (p)ept dp
2 C
where the contour C, called the Bromwich contour, in the complex p-plane runs vertically at
Re(p) = from to + and where is any value such that the contour is to the right of any
poles of F (p). It is often written
Z

f (t) =

1
f (t) =
2

12

Z +

F (p)ept dp

Principal Value (PV)

There are instances, in physics and elsewhere, where a contour passes through a pole. Consider
here, only a simple pole (first order) on the real line at x = p.
The PV is defined as that integral in which a small section, symmetric with respect to the pole, is
chopped out:
Z B

PV = P

Z p

f (x) dx = lim

f (x) dx

f (x) dx +

0

Z B

p+

where the point p lies between A and B (all real).

To evaluate such an integral using contour integration, close the contour in some suitable way (as
before, by adding in a part which contributes zero to the overall integral). For the purposes of
illustration, suppose this is a semi-circle C1 in the upper-half complex plane. In addition, let us
by-pass the pole on the real line at p by following a semi-circle, C0 say, centred on the pole at p and
with infinitesimally small radius , in the upper-half complex plane. This is now a closed contour,
C say, and we can apply the residue theorem in the usual way. However, we need to check the
value of this little added semi-circle around the pole. Let us suppose we have the Laurent series for
f (z) for the annular region centred on the pole. Since the pole is simple, the only negative power
term will be a1 /(z p) and, near the pole, this will be the dominant one (other terms will tend
to zero as the radius tends to zero). Hence,
Z

f (z) dz =
C0

C0

a1
dz = a1
zp
3

Z 0
1

e

e d = a1

Note that this is taken in a counterclockwise sense; we have added this to close the contour so we
need to subtract this amount from the contour integral around the closed contour C:
Z

PV =

f (z) dz + a1
C

To put this into words: the principal value PV is equal to 2 times the residues of all
fully enclosed poles plus times the residues of the poles lying on the original path.
Example of PV
Z

I=P

(x2

1
dx
+ 1)(x 1)

Applying closure as described above (large semi-circle in upper-half plane whose contribution tends
to zero as radius tends to infinity - check this) where we imagine a small semi-circle by-passing the
simple pole at z = 1. The integrand has poles also at z = and z = ; only the former is enclosed.
Its residue is
1
2( 1)
The residue of the pole on the contour at z = 1 is 1/2. Applying the residue theorem (as modified
to obtain the principal value) we get that
I = 2

13

1
+ =
2( 1)
2
2

Cauchy Integral Formula

Theorem

1
f (z)
f (z0 ) =
dz
2 C z z0
where f (z) is analytic in and on the contour C and the point z0 is within C.
I

Proof. (intuitive) Since f (z) is analytic in and on C, then f (z)/(z z0 ) has a simple pole at z0
and has residue f (z0 ). The residue theorem then implies that
I

f (z) dz = 2f (z0 )
C

Hence proved.
Extension If we differentiate Cauchys integral formula w.r.t z0 we obtain contour integral expressions for derivatives to all orders:

f (z0 ) =
f 00 (z0 ) =
f (n) (z0 ) =

1
f (z)
dz
2 C (z z0 )2
I
2!
f (z)
dz
2 C (z z0 )3
I
f (z)
n!
dz
2 C (z z0 )n+1
I

Therefore, analytic functions have derivatives to all orders (not true of real-valued functions). Also,
this implies that all analytic functions have a Taylor series (and vice-versa).

14

Taylor Series derived from the Cauchy Integral Formula

First, recall the infinite geometric progression

1
= 1 + z + z2 + z3 +
1z
provided that |z| < 1.
Let f (z) be analytic in and on a disc centred on z0 . Let the contour C lie wholly within this disc.
Then, Cauchys integral formula can be written as
1
f (z) =
2

I
C

f (u) du
uz

where z is within C and u represents values on C. Let us re-express the integrand as follows:

1
uz

=
=
=
=

1
(u z0 ) (z z0 )


1
z z0 1
1
u z0
u z0
#
"


1
z z0 2
z z0
+
+
1+
u z0
u z0
u z0

(z z0 )n

n=0

1
(u z0 )n+1

on expanding as a geometric progression whose convergence requires that



 z z0 


u z  < 1
0

or, in words, the point z must lie within C (as assumed at the outset).
Using
f (n) (z0 ) =

n!
2

I
C

f (z)
dz
(z z0 )n+1

then Cauchys integral formula may be recast:

1
f (u) du
2 C u z
I

1 X
f (u)
(z z0 )n
du
2 n=0
(u

z0 )n+1
C
I

f (z) =
=
=

an (z z0 )n

n=0

where
an =

1
2

I
C

f (u) du
1
= f (n) (z0 )
n+1
(u z0 )
n!

This is Taylors theorem. It is valid in the largest open disc, centre z0 , within which f (z) is analytic
(edge of disc can be taken out until it meets first singularity).

15

Laurent Series derived from the Cuachy Integral Formula

Theorem Let f (z) be analytic on and between two concentric circles C1 and C2 , centre z0 and
radii R1 and R2 (R2 < R1 ). Then the Laurent series is

f (z) =

an (z z0 )n

n=

where

1
an =
2

I
C

f (u)
du
(u z0 )n+1

and C is any closed contour, within the annulus and enclosing z0 . (Note that, in practice, we can
increase R1 and decrease R2 until each of the circles reaches points where f (z) becomes singular.
Proof Start with Cauchys integral formula applied to a contour C 0 lying on the outer and inner
circles of the annulus but created to exclude - to cut out - the point z0 as follows: let C 0 =
C1 + C 00 + C20 + C 000 where C1 is the outer anti-clockwise circle; C20 is the inner clockwise circle;
C 00 and C 000 are equal, but oppositely directed, straight line sections leading from C1 and C20 (and,
hence, whose contributions to the integral will cancel).

Now, apply Cauchys integral formula to the region inside C 0 where f (z) is analytic:
1
f (u) du
f (z) =
0
2 C u z
I
I
1
f (u) du
1
f (u) du
=

2 C1 u z
2 C2 u z
= AB
I

where C2 is just C20 but taken in the anti-clockwise direction (hence the minus sign). The integral
A, taken round C1 , since z lies within C1 , is just a repeat of the Taylor series argument from the
previous section so that we may write
A=

an (z z0 )n

n=0

where

1
an =
2

I
C1

f (u) du
(u z0 )n+1

By contrast, the argument for the integral B is different since now z lies outside of C2 . Let us use
the fact that, since u lies on C2 ,


 u z0 


z z  < 1
0
6

This time, let us express

1
uz

1
1
=
zu
(z z0 ) (u z0 )


1
u z0 1
=
1
z z0
z z0
"
#


1
u z0
u z0 2
=
1+
+
+
z z0
z z0
z z0
=

Then we can write B as


I
1
1 X
1
n
B =
f (u)(u z0 )
2 z z0 n=0 C2
(z z0 )n

1
1
=
n+1 2
(z

z
)
0
n=0
=
=

(z z0 )m

m=1

1
2

(u z0 )n f (u) du

C2

I
C2

f (u) du
(u z0 )m+1

an (z z0 )n

n=1

where we have changed the summation index, letting m = n 1, then in the final line renaming
the index by letting n = m (remember these are just dummy suffices). The coefficient an is here
1
an =
2

I
C2

f (u) du
(u z0 )n+1

Finally, in the integral A, note that the contour C1 can be moved in towards any contour C that
lies within the annulus without changing the value of A (no singularities are crossed); likewise, in
the integral B, the contour C2 can be pushed outwards to any contour C. Hence, we arrive at the
form expressed in the theorem above.
Note that, if the pole at z0 is of order m, then we can quickly show that the above contour integral
expressions for an will reduce to zero for all n < m. Remember that, for a pole of order m, m
will be the highest negative power of z z0 and that the summation in the Laurent series will start
at the term n = m.
End of section on Complex Analysis