Classification of singular points

If the Laurent series expansion of f (z) in the immediate neighborhood of z = a is expressed as

c−m c−m+1
f (z) = + + . . . + c0 + c1 (z − a) + . . . (1)
(z − a)m (z − a)m−1
we say z = a is an m-th order pole.
Example 1:
1
f (z) = .
z 2 (1 − z)
As
1
= 1 + z + z2 + z3 + . . . (2)
1−z

1
f (z) = (1 + z + z 2 + z 3 + . . .)
z2
1 1
= + + 1 + z + z2 + z3 + . . . (3)
z2 z
So z = 0 is a second order pole.
Example 2:
1 1 1 1
f (z) = + 3 + 4 + 5 + ... (4)
z2 z z z
It appears that the order of singularity at z = 0 is ∞. However,
( )
1 1 1 2 1 3
f (z) = 1 + ( ) + ( ) + ( ) + . . .
z2 z z z
1 1
=
z 2 1 − z1
1
=
z(z − 1)
1
= − − 1 − z − z2 − z3 − . . . , (5)
z
so z = 0 is actually a first order pole. To check the order of singularity, the function must be expanded in
the immediate neighborhood of the singular point.
If the order of singularity is ∞, it is called an essential singularity. Otherwise, singular points are called
removable singularities. At an essential singular point, the function behaves rather wildly as exemplified in
the following function:
Example
1
f (z) = exp(− 2 ).
z
1 1 1 1 1 1
exp(−2
)=1− 2 + 4
− + ..., (6)
z z 2! z 3! z 6
so z = 0 is an essential singular point. To see how this function behaves, let z approach to 0 along the real
axis. By setting z = ϵ,
1 1 1
f (z) → → → → 0. (7)
exp(1/ϵ2 ) e a very large number an even larger number
On the other hand, if z approaches to 0 along the imaginary axis, z can be set as z = ϵi. It follows
1
f (z) → → ea very large number → an even larger number → ∞. (8)
exp(1/(i2 ϵ2 ))

1
If z approaches to 0 along the line, y = x, z can be set as z = ϵ(1 + i) and

− 1
f (z) → e (1+i)2 ϵ2

e− 2ϵ2 i
1
=
i
= e 2ϵ2
1 1
= cos 2
+ i sin 2 (9)
2ϵ 2ϵ
which fluctuates perpetually.
Here is a sketch of |e−1/z |.
2

Residue
When f (z) is expanded by the Laurent series about an m-th order singular point of z = a as
c−m c−m+1 c−1
f (z) = + + ... + + c0 + c1 (z − a) + . . . , (10)
(z − a)m (z − a)m−1 (z − a)
the coefficient, c−1 , of 1/(z − a) is called the residue at that singular point and is denoted as

Res (f (z); a) . (11)

2
Calculation of residues
It is not very effective to compute the residue of f (z) by expanding f (z) by the Laurent series and by picking
up the coefficient of 1/(z − a) as we don’t need information of the rest of the coefficients. Computation of
the residues depends on the degree of singularities.

First order poles

If z = a is a first order pole, the Laurent series of f (z) is
c−1
f (z) = + c0 + c1 (z − a) + c2 (z − a)2 + . . . . (12)
(z − a)
By multiplying (z − a) on both sides of the expression above, one obtains

(z − a)f (z) = c−1 + c0 (z − a) + c1 (z − a)2 + c2 (z − a)3 + . . . . (13)

Therefore,

c−1 = lim (z − a)f (z). (14)

z→a

Second order poles

If z = a is a second order pole, the Laurent series of f (z) is
c−2 c−1
f (z) = + + c0 + c1 (z − a) + c2 (z − a)2 + . . . . (15)
(z − a)2 (z − a)
By multiplying (z − a)2 on both sides of the expression above, one obtains

(z − a)2 f (z) = c−2 + c−1 (z − a) + c0 (z − a)2 + c1 (z − a)3 + . . . . (16)

Direct substitution of z = a in eq.(16) picks up c−2 , not c−1 , not what we want. Noting that eq.(16) is an
identity, differentiating both sides yields another identity as
d ( )
(z − a)2 f (z) = c−1 + 2c0 (z − a) + 3c1 (z − a)2 + . . . . (17)
dz
Therefore,
d ( )
c−1 = lim { (z − a)2 f (z) }. (18)
z→a dz

Higher order poles

If z = a is an m-th order pole, the residue of f (z) at z = a can be expressed as

1 dm−1
c−1 = lim { m−1 ((z − a)m f (z))}. (19)
(m − 1)! z→a d z

Short cut (Residues for first order poles only)

If z = a is a first order pole and f (z) is expressed as

h(z)
f (z) = , (20)
g(z)

then
h(a)
c−1 = . (21)
g ′ (a)

3
(Proof)

c−1 = lim (z − a)f (z)

z→a
(z − a)h(z)
= lim
z→a g(z)
h(z) + (z − a)h′ (z)
= lim
z→a g ′ (z)
h(a)
= , (22)
g ′ (a)
where the l’Hospital’s theorem was used from eq.(22) to eq.(22).
Example 1
1
f (z) = 4 .
z +1
1
There are four zeros each of which is a first order pole. The singular points are

z = eπi/4 , e3πi/4 , e5πi/4 , e7πi/4 .

For example, the residue at z = eπ/4i is

( ) 1
Res f (z); eπi/4 =
(z 4 + 1)′ | z→exp(πi/4)
1
=
4z 3 |z→exp(πi/4)
1
= 3πi/4
4e
1 −3πi/4
= e
4 ( )
1 1 1
= − √ +√ i . (26)
4 2 2

Residue theorem
If f (z) has (isolated) singular points, z1 , z2 , z3 , . . . zn inside a closed loop, C, it follows
I ∑
n
f (z)dz = 2πi Res(f, zj ). (27)
C j=1

Application of Residue Theorem

There are three types of integrals where you can apply the residue theorem.

They are:
1 The singularities are the zeros to the equation,
z4 + 1 = 0 or z 4 = −1. (23)
Expressing both z and −1 in polar form, the equation above becomes

r4 e4iθ = 1e(π+2nπ)i (24)

from which it follows
π nπ
r = 1, θ= + , (25)
4 2
where n = 0, 1, 2, 3 . . ..

4
 1. ∫ 2π
f (sin θ, cos θ)dθ
0

2. ∫ ∞ ∫ ∞
f (x)dx or f (x)eix dx
−∞ −∞

3. ∫ ∞
f (xα , log x)dx α is non-integer.
0

First type
If one changes the variable from θ to z by setting

z = eiθ (28)
the integral from 0 to 2π with respect to θ is converted into a contour integral along the unit circle
I
(29)
|z|=1

By taking derivative of eq.(28),

dz = ieiθ dθ
(30)
= izdθ,
or
1
dθ = dz, (31)
iz
and
1 ( iθ )
sin θ = e − e−iθ
2i ( )
1 1
= z− , (32)
2i z

1 ( iθ )
cos θ = e + e−iθ
2( )
1 1
= z+ , (33)
2 z

so that

∫ 2π I
dz
f (sin θ, cos θ)dθ = f (z)
0 |z|=1 iz
∑ ( )
f (z)
= 2πi Residue ; aj , (34)
j
iz

where aj must be inside the unit circle.

5
Example
∫ 2π I dz
dθ iz
= 1
0 2 + cos θ |z|=1 (z + 1/z)
2+
I 2
2dz
= 2 + 4z + 1)
|z|=1 i (z
I
2 1
= dz
i |z|=1 z 2 + 4z + 1
∑ ( )
2 1
= 2πi Res ; ai . (35)
i z 2 + 4z + 1
√ √
The roots of the equation z 2 + 4z + 1 = 0 are −2 ± 3 of which only −2 + 3 is inside the unit circle so

( )
2 1 √
= 2πiRes ; −2 + 3
i z 2 + 4z + 1
1
= 4π | √
2z + 4 z=−2+ 3
2π
= √ . (36)
3
Note that we used the short-cut to find a residue at a first order pole.

Example ∫ 2π
cos6 θdθ.
0

(Solution)
∫ 2π I ( )6
1 1 dz
cos6 θdθ = (z + )
0 |z|=1 2 z iz
I
1 (1 + z 2 )6
= dz
64i |z|=1 z7
I
1 z 12 + 6z 10 + 15z 8 + 20z 6 + 15z 4 + 6z 2 + 1
= dz
64i |z|=1 z7
( I )
1 dz
= ... + 20 + . . .
64i |z|=1 z
I
1 dz
= 20
64i |z|=1 z
1
= × 20 × 2πi
64i
5
= π. (37)
8