D.

T S A N G

P H 3 0 0 2 5 M AT H E M AT I C A L M E T H O D S :

C O M P L E X A N A LY S I S
SEMESTER 2, 2020/21

D E PA R T M E N T O F P H Y S I C S , U N I V E R S I T Y O F B AT H
2
1
Review of Complex Numbers and Functions

1.1 Complex Numbers

1.1.1 Mathematical Fields

Mathematical fields are sets of "numbers" that possess the usual

sense of addition, subtraction, multiplication and division. They
serve as a foundational concept for algebra, number theory, and
real and complex analysis. For example, any mathematical field
may act as the set of scalars for a vector space, due to the basic
properties that fields possess.

Mathematical Fields
Definition 1.1. In mathematics, a field is defined to be a set of mathematical objects that comes equipped
with an addition operation, and a multiplication operation, which both take two elements of the field and
produce another element of the field. For a field F , equipped with + and ·,

+ : F ×F → F
· : F ×F → F

F must be closed under + and ·, and must also obey the axioms ∀ a, b, c ∈ F :
(C+ ) a + b = b + a, (Commutativity of +)
+
(A ) a + (b + c) = ( a + b) + c, (Associativity of +)
(N+) ∃ an element 0 ∈ F such that a + 0 = a, (Neutral element under +)
( I+ ) ∃ an element − a ∈ F such that a + (− a) = 0, (Inverse under +)

( C· ) a · b = b · a, (Commutativity of ·)
( A· ) a · (b · c) = ( a · b) · c, (Associativity of ·)
( N· ) ∃ an element 1 ∈ F such that 1 · a = a, (Neutral element under ·)
( I· ) ∀ a 6= 0, ∃ an element a−1 ∈ F such that a−1 · a = 1, (Inverse under ·)

( D ) a · ( b + c ) = ( a · b ) + ( a · c ), (Distributivity of · over +)

Maths Shorthand
The above definition may seem very intimidating, but really all it ∃: "there exists".
says is that addition and multiplication (as well as subtraction and ∀: "for all".
division) behave as "usual" for a field. ∈: "in" or "is an element of".

The simplest example of a field is the set of rational numbers

with the usual addition and multiplciation (and therefore subtrac-
4 complex analysis

tion and division). However, the rational numbers don’t contain the
solution to some equations, such as

x · x = 2. (1.1)

To solve such equations, we must add the irrational numbers (such

√
as 2) to the set of rational numbers to get the set of real numbers,
R, that you are used to.
Similarly the real numbers cannot solve certain equations, like

x · x = −1. (1.2)

To “correct” this, we can add a number denoted by i to the reals, Note that engineers often use j instead
defined such that of i to avoid confusion with current.
We are not engineers.
i 2 = i · i = −1 (1.3)

1.1.2 Complex Numbers and their Representations

The Complex Numbers, C

Definition 1.2. We can then define the set of complex numbers C
as a field such that
We will adopt the normal convention
1. Every real number is also a complex number and write ab = a · b = a × b inter-
changeably for the multiplication
2. i is a complex number operation

3. Every complex number can be written in the form a + ib where

a, b ∈ R, and any such number is a member of C.

4. Addition is defined componentwise, as inherited from the

reals, such that for α = a1 + ia2 and β = b1 + ib2 , then

α + β = ( a1 + b1 ) + i ( a2 + b2 )

5. Multiplication is defined using multiplication inherited from the

reals and i2 = −1, such that

α · β = ( a1 + ia2 ) · (b1 + ib2 )

Figure 1.1: The point (cartesian/rect-
= a1 b1 − a2 b2 + i ( a1 b2 + a2 b1 ) angular) representation of a complex
number.

The Complex Conjugate

Definition 1.3. Let α = a + ib be a complex number. We define α
to be

α ≡ a − ib ∈ C (1.4)
We will always use the overbar to
The complex number α is called the complex conjugate of α. denote complex conjugation. Other
common notation includes e.g. α∗ .

From the definition of the complex conjugate we can immedi-

ately see that

αα = a2 + b2 ∈ R. (1.5)
 review of complex numbers and functions 5

Interpreting the real and imaginary “components” as vector compo-

nents in the 2-D plane, we see that αα corresponds to the square of
the distance from the origin, also known as the absolute value (or
modulus) squared.
Absolute Value
Definition 1.4. The absolute value or modulus of the complex
number α = a + ib is denoted |α| and is given by
p √
|α| ≡ a2 + b2 = αα

Does the set of complex numbers satisfy the properties of a

field? The set C is clearly closed since the real numbers we use to
build them are closed, and we see that by basing the addition and
multiplication operations on those inherited from the real numbers,
we obviously satisfy most of the C+ A+ N+ I+ C· A· N· I· D axioms.
The one tricky axiom we may wish to verify is I· , the existence
of the multiplicative inverse for any complex number 6= 0. For any
α = a + ib 6= 0 we can use the notion of a complex conjugate to
construct the inverse
α
α −1 ≡ 2 (1.6)
a + b2
We can immediately see that αα−1 = α−1 α = 1 since
αα αα
= =1 (1.7)
a2 + b2 αα
hence every non-zero α ∈ C has a multiplicative inverse also in C.
Exercise 1.1:
Verify the following:

Theorem 1.1. Let α and β be any complex numbers. Then

αβ = αβ, α + β = α + β, α = α.

Since we can associate each complex number with a vector in

the 2-D plane, we can also see that the component-wise addition of
complex numbers also corresponds to the component-wise addition
of vectors. Recalling the geometric fact that the length of any side
of triangle is less than or equal to the sum of the lengths of the
other two sides, we can, by analogy, deduce the triangle inequality.
The Triangle Inequality
Theorem 1.2 (The triangle inequality). For any two complex
numbers α and β, we have

| α + β | ≤ | α | + | β |.

Just as 2-D vectors can be expressed in either cartesian form

( x, y), or polar form (r, θ ), we can also express complex numbers in
a “polar” form.
6 complex analysis

We know that we can clearly express cartesian coordinates in

terms of polar coordinates

x = r cos θ, y = r sin θ. (1.8)

However, we must be careful in expressing the polar coordinates

in terms of the cartesian coordinates. While the expression for r is
clear,
q
r = x 2 + y2 , (1.9)

we have for tan θ = y/x, however if we specify

θ = tan−1 (y/x ), (1.10)

Figure 1.2: The polar representation of
complex number α.
this expression is invalid for points in the second and third quad-
rants (for the normal definition of the arctangent, tan−1 ). In practice
we must add or subtract π in order to correctly place points in
those quadrants. We can instead specify a related function


 tan−1 (y/x ) if x > 0,

− 1

tan (y/x ) + π if x < 0, y ≥ 0,





tan−1 (y/x ) − π if x < 0, y < 0,

θ = arctan( x, y) ≡ (1.11)

 +π/2 if x = 0, y > 0, Definition of arctan( x, y). In many


 programming languages, this 2


 −π/2 if x = 0, y < 0, variable function is often called
"atan2".


undefined if x = 0, y = 0.


However, even using this approach, there is ambiguity, as the

angle θ is only determined up to an integer multiple of 2π. To ac-
commodate this, we will refer to the value of any of these "angles"
for complex number z = x + iy, as the set of angles, called the
argument of z,

arg z ≡ {arctan( x, y) + 2kπ : k = 0, ±1, ±2, ...}. (1.12) Definition of arg z

For example, arg i = { π2 + 2kπ, for k = 0, ±1, ±2, . . .}.

It is convenient to have a notation for a specific angle from the
set of arg z. Any half open interval of length 2π will contain one
and only one value of the argument. Specify a particular choice of
this range, is called choosing a particular branch of arg z. (Note that
arg(0) cannot be reasonably defined for any branch.)
Normal convention defines the selection of the branch of arg z One handy notation is argτ (z) which
is used for the branch of arg(z) taking
from (−π, π ] to be the principal value of the argument, which is values from the interval (τ, τ + 2π ].
usually written as Arg z with a capital A. The principal value is Thus arg−π (z) is the principal value
particularly important in complex arithmetic in numerical codes, Arg (z).

and is inherently discontinuous, with value jumping by 2π as z

crosses the negative real axis. This line of discontinuity is called the
branch cut of Arg z.
With all these conventions in hand, we can write z = x + iy in the
polar form

z = x + iy = r (cos θ + i sin θ ) (1.13)

 review of complex numbers and functions 7

√
where r = |z| = zz and θ ∈ arg(z),
In many circumstances, the rectangular form (x + iy) or the polar
form (r [cos θ + i sin θ ]) may be more suitable than the other. The
rectangular form, for example is very convenient for addition or
subtraction, whereas in the polar form this can be very difficult.
However, multiplication and division of complex numbers in
polar form provides a very useful geometric interpretation of
multiplication and division.

Exercise 1.2: Polar Multiplication

Show the following using trigonometric identities:
Figure 1.3: Arg (z), where the principal
Theorem 1.3. The modulus of the product is the product of value of arg(z) has a branch cut
the moduli: (denoted by the jagged line) along
the negative real axis. Along this axis
|αβ| = |α|| β| Arg (z) = π, while slightly below the
negative real axis Arg (z) → −π.

and the argument of the product is the sum of the argu-

ments:

arg(αβ) = arg(α) + arg( β).

(Where the above is to be interpreted as saying that if par-

ticular values are assigned to the arguments on the left
hand side, one can find a value for the right hand side that
satisfies the relation.)

With the properties in Theorem 1.3, we may begin to see a con-

nection to the exponential function, where multiplication becomes
an addition of the exponents. Indeed we can formalise this by
providing a suitable definition for exp(z) where z = x + iy ∈ C
preserves the basic identities satisfied by the real exponential.
We want the complex exponential to share the product-to-sum-of-
the-exponents property of the real exponential such that

eα e β = eα+ β , ∀α, β ∈ C (1.14)

This allows us to simplify the problem of defining the complex

exponential considerably, since it allows the decomposition

ez = e x+iy = e x eiy (1.15)

such that all that remains is to define the exponential of a purely

imaginary number!
Another property that will help us proceed is the differentiation
property of the exponential function,

d z
e = ez (1.16)
dz
which we would like to be true, as in the real version.
8 complex analysis

Differentiation of a complex function is a bit tricky, (which we

will cover later), but we can just focus on the derivative of the
purely imaginary exponent which we want to be

d iy
e = eiy (1.17)
d(iy)

which by the chain rule gives

d iy
e = ieiy . (1.18)
dy

Taking the second derivative, using this rule, we see

d2 iy d
2
e = (ieiy ) = i2 eiy = −eiy (1.19)
dy dy

but we know that the differential equation f 00 (y) = − f (y) is solved

by functions of the form f (y) = A cos y + B sin y, where A and B are
constants.
We can determine these constants by noting that f (0) = ei0 =
e0 = 1 = A cos 0 + B sin 0, and f 0 (0) = iei0 = i = − A sin 0 + B cos 0.
Hence A = 1, B = i, and we must have
Euler’s Equation
Theorem 1.4 (Euler’s equation).

eiy = cos y + i sin y, ∀y ∈ R.

This naturally leads to the definition:

The Complex Exponential

Definition 1.5. If z = x + iy ∈ C, for x, y ∈ R, then ez is defined
to be the complex number

ez ≡ e x (cos y + i sin y).

which turns out to be the "right choice", as it obeys all the same
properties as the real exponential, and is holomorphic which we
will define and demonstrate later when we look in detail at the
differentiability of complex functions.

Exponential Polar Form

Euler’s equation allows us to write the polar form of a
complex number particularly compactly,

z = r (cos θ + i sin θ ) = r eiθ = |z|eiArg z (1.20)

The complex conjugate z can be written as

z = r (cos θ − i sin θ ) = re−iθ = |z|e−iArg z . (1.21)

 review of complex numbers and functions 9

Exercise 1.3: Taylor Expansion of Euler’s Equa-

tion
Show that Euler’s equation is formally consistent with the
usual Taylor series expansions

x2 x3 x4
ex = 1 + x + + + +...
2! 3! 4!
x2 x4
cos x = 1 − + −...
2! 4!
x3 x5
sin x = 1x − + −...
3! 5!

Exercise 1.4: De Moivre’s theorem

Prove the following,

Theorem 1.5 (De Moivre’s thoerem).

(cos θ + i sin θ )n = cos nθ + i sin nθ, n = 1, 2, 3, . . .

1.1.3 Powers and Roots of Complex Numbers

Let z = reiθ = r (cos θ + i sin θ ) be the polar form of the complex
number z. By Theorem 1.3 (or the properties of the exponential
function), we have

z2 = z · z = r2 ei2θ . (1.22)

Multiplying by z again, we have

z3 = r3 ei3θ . (1.23)

Continuing this procedure provides us with the generalisation of

De Moivre’s theorem for powers of z,

zn = r n eiθ = r n (cos nθ + i sin nθ ) (1.24)

Exercise 1.5: Generalised De Moivre’s Theorem

The expression above is a useful formula for raising a com-
plex number to a positive integer power n. Show that the
above equation also holds for negative n.

The question arises whether this generalised De Moivre’s theo-

rem will work for n = 1/m, so that that ζ = z1/m is the mth root of
z satisfying

ζ m = z. (1.25)

Certainly if we define
√
ζ= m
r eiθ/m , (1.26)
10 complex analysis

then ζ satisfies the equation above. However, the the presence of

additional roots due to multiple branches of the argument θ ∈
arg(z) complicates matters, as θ is then defined only up to a value
of 2π.
Example 1.1: Roots of Unity
There are exactly m distinct mth roots of unity, denoted by
11/m , and they are given by

2kπ 2kπ
ei2kπ/m = cos + i sin (k = 0, 1, 2, . . . , m − 1)
m m
(1.27)

Let ωm ≡ ei2π/m . The complete set of mth roots of unity is

2 m −1
1, ωm , ωm , . . . , ωm . (1.28)

We can also prove that

2 m −1
1 + ωm + ωm + . . . + ωm = 0. (1.29)

This result is obvious from a physical point of view, since

by symmetry the centre of "mass" (1 + ωm + ωm 2 + ... +
m − 1
ωm )/m of the system of m unit masses located at the mth
roots of unity must be at the origin.

1.1.4 The Complex Logarithm

We want to define the complex logarithm, log z, as the "inverse" of We will almost always take the loga-
the exponential function; i.e. rithm with base e, so log or Log will
always refer to this. In other texts the
notation ln or Ln is often used.
w = log z if z = ew . (1.30)

Since ew is never zero, we presume that z 6= 0. To find log z ex-

plicitly, let us write z in polar form as z = reiθ and write w in the
 review of complex numbers and functions 11

standard form w = u + iv. Then the equation z = ew becomes

reiθ = eu+iv = eu eiv . (1.31)

From this we can deduce that u = Log r, where this is the usual
(real) logarithm of a real number, and v = arg z = θ can take
multiple values.
The Complex Log
Definition 1.6 (The Complex Logarithm). If z 6= 0, then we de-
fine log z to be any of the infinitely many values

log z ≡ Log |z| + i arg z

= Log |z| + iArg z + i2kπ (k = 0, ±1, ±2, . . .)

The multiple-valued nature of log z simply reflects

that the imaginary part of the logarithm is the polar
angle θ; the real part is single-valued. We can choose
a single-value by using, once again, the notion of a
branch cut to resolve the ambiguity, as we did for Arg z.

Definition 1.7. We can define the principal value of of the

logarithm, Log z, to be the value inherited from the principal value
of Arg z, domain

Log z ≡ Log |z| + iArg z

1.2 Complex Functions

Let S be a set of complex numbers. A mapping which associates

each element of S to a complex number is called a complex val-
ued function, or function for short. We denote such a function by
range
symbols like

f :S→C (1.32)

where S is a subset of C. If z ∈ S, we write the association of the

Figure 1.4: A complex function f (z)
value f (z) to z by the special arrow with domain X, and range Y.

z 7→ f (z) (1.33)

If a function f is defined f : X → Y, (where X, Y ⊆ C) we call the

set X the domain of f , and the set Y the range of f .
Usually, when we use the word "function" to describe f we mean
that f assigns a single value w in the range to each permissible
value of z in the domain. Sometimes we explicitly state that " f is a
single-valued function". Of course there are equations that do not
define single-valued functions, for example w = arg z, w = z1/2 ,
and w = log z as we saw in the previous section. In general, if
for some values of z there corresponds more than one value of
12 complex analysis

f (z), then we say that f (z) is a multi-valued function (note that

technically these are not considered functions). We commonly
obtain multiple-valued functions by taking "inverses" of single
valued functions that are are not one-to-one (see discussion below),
such as in the case of the complex logarithm we discussed above.

1.2.1 Sets on the Complex Plane

Here we will briefly cover notions of point sets in the plane, in
order to be able to more carefully define the domains and ranges
which complex functions map between.

Open and Closed Discs

We will make frequent reference to the
Definition 1.8. Let α be a complex number. An open disc, de- neighbourhood D (0, 1), the unit open
noted by D (α, r ), of (real) radius r > 0 centred at α is the set of all disc.
complex numbers z, such that

D (α, r ) ≡ {z : |z − α| < r }.

Similarly the set D (α, r ) ≡ {z : |z − α| ≤ r } is called the closed

disc centred at α, of radius r.

Open and Closed Sets

Definition 1.9 (Open Sets). Let S be a subset of the complex
plane. S is called an open set if for every point α in S, there is a
Figure 1.5: The point α is an interior
disc D (α, r ) centred at α, and of some radius r > 0 such that this point of S if you can find some r > 0
disc D (α, r ) is completely contained in S. such that disc D (α, r ) ⊂ S. S is open if
it only contains interior points.
Each point α that obeys the condition above is called an inte-
rior point of set S. Open sets contain only interior points.
A point β is called a boundary point of set S if any ∀r > 0,
the disc D ( β, r ) contains at least one point in S, and at least
one point not in S. The set of all boundary points is called
the boundary or frontier of S.

Definition 1.10 (Closed Sets). A set S is called a closed set if it

contains all of its boundary points.

Note that the set {z : 0 < |z| ≤ 1} is not closed since it Figure 1.6: The point β is a boundary
does not contain the boundary point 0. Nor is it open since point of S̄ if any disc D ( β, r ) contains
points both inside and outside S̄. S̄ is
points in the set where |z| = 1 are boundary points. (Such
closed if it contains all its boundary
part open part closed sets are sometimes called clopen sets. points. The set of all boundary points
Unfortunately.) of a set S is called the boundary of S
and is sometimes denoted ∂S.
Also note that the same set can be considered an open subset
of one space, but not of another. For example, the interval
(0, 2π ) is an open subset of the real line R, but it is not an
open subset of C (all points in the interval are boundary
points in this case).
 review of complex numbers and functions 13

The closure of a set S is defined to be the union of S and all its

boundary points. We denote this closure by S.

Figure 1.7: Left: The set S is the open

first quadrant on the complex plane.
Right: The set S is the closed first
quadrant which includes the positive
real, and positive imaginary axes.
(Neither of these sets is compact.)

There are a few other definitions

A set S is said to be bounded if there exists a real number C > 0 of compactness, particularly that
such that every sequence of elements of S has
a convergent subsequence whose
limit is in S. This can be shown to be
|z| ≤ C, ∀z ∈ S (1.34) equivalent to our definition, at least on
the complex plane.
A set S is said to be compact if and only if it is both closed and
bounded.

Figure 1.8: Left: The open set S is

bounded, but is not compact. Right:
The closed set S is also bounded,
therefore it is compact.

A set S, is said to be connected if given any two points α and β

in S, there exists a path within S that joins α to β. (This relatively
strong definition of connectedness is sometimes called pathwise
connectedess; there are other notions of connectedness that have to
do with topology, which we will not get into in this module.)

Figure 1.9: Examples of connected sets.

A (path-wise) connected set is said to be simply connected if any

closed curve (i.e. a loop) inside the set can be continuously shrunk
down to a point.
14 complex analysis

Figure 1.10: Simply connected sets are

path-wise connected sets where any
closed loop can be shrunk to a point.

Simply Connected Not Simply Connected Not Simply Connected

Pathwise Connected Not Pathwise Connected Pathwise Connected

1.2.2 Limits, Sequences, and Continuity

The definition of the absolute value can be used to designate the
distance between two complex numbers (this is sometimes called a
metric). This allows us to easily define "open" sets (though it is not
the only way to do so, see e.g. any introductory text on topology,
where "openness" is defined by fiat). Having the concept of distance
we can also proceed to introduce notions of limits and convergence,
analogous to those you are used to with the real numbers.
Informally, when we have an infinite sequence of complex num- Figure 1.11: The convergence of a
bers, z1 , z2 , z3 , . . ., we say that the number w is the limit of the se- sequence {zn } to w. Each term zn for
n > N lies inside the open disc of
quence if the zn eventually (i.e. for large enough n) stay arbitrarily radius ε about w.
close to w. More precisely,

Limit of a Complex Sequence

Definition 1.11. A sequence of complex numbers {zn }∞
n=1 is said
to have the limit w ∈ C or to converge to w, i.e.

lim zn = w, or zn → w as n → ∞
n→∞

if ∀ε > 0, (ε ∈ R) ∃ N such that |zn − w| < ε, ∀n > N.

There is an additional notion of convergence that is

very useful, known as the Cauchy Criterion. In fact,
a necessary and sufficient condition for a complex
sequence to converge is that it obeys this condition

Theorem 1.6 (Cauchy Criterion). A sequence {zn }∞ n=1 will con-

verge if and only if ∀ε > 0, ∃ integer N such that
Definition of a Cauchy Sequence

|zn − zm | < ε, ∀m > N, ∀n > N.

(We will demur from rigorously proving this, for now, as

sufficiency is slightly tricky.) A sequence that obeys the
Cauchy Criterion is called a Cauchy Sequence. Note that
this provides an alternative and very useful definition of
convergence that doesn’t require reference to the notion of
limits.
 review of complex numbers and functions 15

A related concept is the limit of a complex-valued function f (z).

Roughly speaking, we say that the number wo is the limit of the
function f (z) as z approaches zo if f (z) stays close to wo whenever z
is sufficiently near zo . More precisely we have,

Limit of a Complex Function

Definition 1.12. Let f be a function defined on some (open set)
neighborhood of zo , with the possible exception of the point zo
itself. We say that the limit of f (z) as z approaches zo is the
complex number wo i.e.

lim f (z) = wo ,
z→zo

or f (z) → wo as z → zo

if ∀ε > 0, ∃δ > 0, (ε, δ ∈ R) such that

Figure 1.12: The limit of a function is
defined such that any open disc of wo
| f (z) − wo | < ε, if |z − zo | < δ. contains all the values assumed by f
in some open disc around zo (except
possibly the value f (zo )).

There is an obvious relation between the limit of a function

and the limit of a sequence; namely, if limz→zo f (z) = wo , then
for every sequence {zn } converging to zo , the sequence { f (zn )}
converges to wo . The converse of this statement is also true.
With these notions in hand, we can define the concept of continu-
ity of a function:

Continuous Functions
Definition 1.13 (Continuity at a point). Let f be a function Cultural Side Note: A more general
defined in an open disc centred at zo . Then f is called continuous (topological) notion of continuity
is that a mapping f : X → Y is
at zo if continuous if for any open subset
of the range, W ⊂ Y, the preimage
lim f (z) = f (zo ). of f (W ) (i.e. the domain set that
z→zo
corresponds to the range set W, also
known as f −1 (W ) if f is invertible), is
In other words, for f to be continuous at zo , it must have a an open subset of X. For X, Y ⊆ C, this
limiting value at zo and this limiting value must be f (zo ). definition is equivalent to ours.

Definition 1.14 (Continuity on a set). A function f is said to be

continuous on a set S if it is continuous at each point of S.

Clearly the definitions for continuity and limits are directly

analogous to those used for normal calculus and analysis on the
real numbers. Because of this analogy, many of the familiar theo-
rems and lemmas on real sequences, limits, and continuity remain
valid in the complex case, such as,
16 complex analysis

Limits of Complex Functions

Theorem 1.7. If f (z) and g(z) are complex functions for z ∈ C
and limz→zo f (z) = A and limz→zo g(z) = B, then

lim ( f (z) ± g(z)) = A ± B,

z→zo

lim f (z) g(z) = AB,

z→zo
f (z) A
lim = if B 6= 0.
z→zo g(z) B

Lemma 1.1. If f (z) and g(z) are continuous complex func-

tions at zo , then so are f (z) ± g(z) and f (z) g(z). The quotient
f (z)/g(z) is also continuous at zo provided g(zo ) 6= 0.

Exercise 1.6: Limits of Real and Complex Func-

tions
Let f (z) = u( x, y) + iv( x, y), for z = x + iy, and zo = xo + iyo ,
and wo = uo + ivo . Prove that

lim f (z) = wo
z→zo

if, and only if,

lim u( x, y) = uo , and lim v( x, y) = vo ,

x → xo x → xo
y→yo y→yo

i.e., a complex function has a limit if and only if its real and
imaginary components have the appropriate limits.
[Hint: Use the triangle inequality and the facts that |Re w| ≤
|w| and |Im w| ≤ |w|.]

1.2.3 Injective, Surjective and Bijective Functions

Finally, we review the notion of (single-valued) functions that
are one-to-one (also called injective), onto (also called surjective),
and invertible (also called bijective). Note that these notions are
applicable to more than just complex-valued functions.

One-to-One, Onto, and Invertible Functions

For single valued functions (mutli-valued functions, are,
technically, not considered functions):

Definition 1.15. An injective function is one for which every

element of the domain points to a unique element of the range.
Injective functions are also called one-to-one.
 review of complex numbers and functions 17

Definition 1.16. A surjective function is one for which every

element of the range has (at least) one element in the domain which
maps to it. Surjective functions are also called onto.

Definition 1.17. A bijective function is one that is both injective

and surjective. A bijective function perfectly pairs up members
of the domain and range with no overlaps or omissions. Bijective
functions are often called one-to-one and onto, or invertible,
since they admit single-valued inverse functions.

Multi-Valued Function Injective/One-to-one Surjective/Onto Bijective/One-to-one & Onto

(Not a function) (Not surjective) (Not injective) (Invertible)
2
Complex Differentiation and Holomorphic Functions

2.1 The Complex Derivative

Now that we have a reviewed the notions of complex numbers

and complex functions, we turn to the main topic of this part of
the module: the study of functions that are differentiable in (most
of) the complex plane. In studying differentiable functions of real
variables, we took such functions defined on intervals. For complex
variables, we have to select domains of definition in an analogous
manner.
Derivative of a Complex Function
Definition 2.1 (The Complex Derivative). Let U be an open
subset of C, and let z be a point of U. Let f be a function on U.
We say that f is complex differentiable at z if the limit
Figure 2.1: A function is only complex
f (z + h) − f (z)
f 0 (z) ≡ lim , (2.1) differentiable at a point z if f 0 (z) is the
h →0 h same well-defined value regardless
of which path the complex number
exists (where h ∈ C). This limit is the derivative of f at z, and h takes to 0. In other words the
can also be denoted by d f /dz. "direction" the derivative is taken in on
the complex plane does not matter.

Unless otherwise specified, for us, differentiable will always

mean complex differentiable. This is almost the same definition
of derivative as for real variables, with one crucial difference. On
the real line there are two "directions" from which h can approach
zero. But on the complex plane, there are an infinite number of
directions for which this can occur. Indeed, in order for a function
to be considered complex differentiable, this limit must approach
only a single value, regardless of which path h takes.
Example 2.1: The Complex Conjugate
Show that z is nowhere differentiable.
Solution: The finite difference approximation for f (z) = z
takes the form
f (z + h) − f (z) (z + h) − z h
= =
h h h
where h ∈ C. Now if h → 0 through real values, then
 complex differentiation and holomorphic functions 19

h = ∆x, and h = h, and we must have the above quotient

be 1. On the other hand if h → 0 from above, then h = i∆y
and h = −h, so the quotient is −1. Thus there is no simple
way of assigning a unique value to the derivative of z at any
point, and it is not differentiable.
Can you show Re(z), Im(z) and |z| are not differentiable?
1
The term holomorphic comes from
the Greek, "holos" for "entire", and
"morphe" for "form" or "shape". This is
sometimes used interchangeably with
Holomorphic (Analytic) Functions analytic, though these mean slightly
different things. We will discuss this
Definition 2.2 (Holomorphic Functions). Let U be an open sub- more later.
set of C. A function f : U → C is called holomorphic on U if
∀z ∈ U, f is complex differentiable at z.

Note that differentiability may be defined at a single point,

whereas a function can only be considered holomorphic1 (or ana-
lytic) on open sets. A function that is holomorphic on the whole Definition of an entire function.
complex plane is called an entire function.

Example 2.2: dzn /dz

Show that for any positive integer n,

d n
z = nzn−1
dz
Solution: Using the binomial theorem (Worksheet 1) we find
n ( n −1) n −2 2
(z + h)n − zn nzn−1 h + 2 z h + . . . + hn
= .
h h
Thus
d n (z + h)n − zn
z = lim = nzn−1
dz h →0 h

The usual proofs of the calculus of real variables concern-

ing the basic properties of differentiability are valid for complex
differentiability2 . 2
We will review them here, for your
We note that if f is differentiable at z then it is also continuous at reference, but will not go over them in
detail in lecture.
z because

f (z + h) − f (z)
lim ( f (z + h) − f (z)) = lim · lim h = 0
h →0 h →0 h h →0
=⇒ lim f (z + h) = f (z). Differentiable implies continuous.
h →0

since (from Thm. 1.7) the limit of the product is the product of the
limits.
20 complex analysis

Sum and Product Rules for Complex Derivatives

Let f , g be functions defined on the open set U ⊂ C, and dif-
ferentiable at some point z ∈ U.

Theorem 2.1 (Sum Rule). The sum f + g is differentiable at z and

( f + g)0 (z) = f 0 (z) + g0 (z)

Proof: This follows immediately from Theorem 1.7, that the

limit of the sums is the sum of the limits.

Theorem 2.2 (Product Rule). The product f g is differentiable at

z, and

( f g ) 0 ( z ) = f 0 ( z ) g ( z ) + f ( z ) g 0 ( z ).

Proof: We want to find the limit as h → 0 of

f (z + h) g(z + h) − f (z) g(z)

.
h
We can write the numerator in the form

f ( z + h ) g ( z + h ) − f ( z ) g ( z + h ) + f ( z ) g ( z + h ) − f ( z ) g ( z ),
| {z }
0

so that the derivative takes the form

 
f (z + h)− f (z) g(z + h)− g(z)
( f g)0(z) = lim g(z + h) + f (z)
h →0 h h
= f 0 ( z ) g ( z ) + f ( z ) g 0 ( z ).

The Quotient Rule

Theorem 2.3 (Quotient Rule). If g(z) 6= 0, then the quotient of
f /g is differentiable at z, and

g(z) f 0 (z) − f (z) g0 (z)

( f /g)0 (z) =
g ( z )2

Proof: First we have to show that 1/g is differentiable. As

for real calculus we have,
1 1
g(z+h)
− g(z) g(z + h) − g(z) 1
=− .
h h g(z + h) g(z)

Applying the limit gives,

d(1/g(z)) g0 (z)
=−
dz g ( z )2

which exists if g(z) 6= 0 and if g0 (z) exists. The quotient rule

is then proved using the above result and the product rule.
 complex differentiation and holomorphic functions 21

The Chain Rule

Theorem 2.4 (Chain Rule). Let w = f (z). Assume that f is dif-
ferentiable at z and function g(w) is differentiable at w. Then g ◦ f ,
The function ( g ◦ f )(z) is pronounced
is differentiable at z and "g circ f ", "g composed with f ", or
sometimes "g after f ". It means to
( g ◦ f )0 (z) = g0 ( f (z)) f 0 (z) apply function g to the result of f (z).
Obviously, the domain of g must
overlap with the range of f in order for
Proof: Since f (z) is differentiable at z, for h ∈ C, we define g ◦ f to have a non-null domain.
the function ϕ f (h) to be

f (z + h) − f (z)
ϕ f (h) ≡ − f 0 (z)
h
such that it is the "error" of the "finite difference" approxima-
tion to f 0 (z).
Then we have both

f (z + h) − f (z) = f 0 (z)h + hφ f (h) and lim ϕ f (h) = 0.

h →0

The existence of such a function ϕ f (h) obeying the above

equations can be shown to be equivalent to differentiability
of f at z.
We can then use this alternate definition of differentiability
and let w = f (z) and

k = f (z + h) − f (z)

so that

g( f (z + h)) − g( f (z)) = g(w + k ) − g(w)

Since g(w) is differentiable at w we have a function ϕ g (k )

such that

g(w + k) − g(w) = g0 (w)k + kϕ g (k), and lim ϕ g (k) = 0.

k →0

Substituting in for k, and dividing by h we find

g ◦ f (z + h) − g ◦ f (z) f (z + h) − f (z)
= g0 (w)
h h
f (z + h) − f (z)
+ ϕ g ( k ).
h
Taking the limit as h → 0, we note that k → 0 by continuity
of f at z. Then, since ϕ g (k) → 0 as k → 0, the last term van-
ishes, and we recover the chain rule.

It follows from the Sum and Product Rules and Example 2.2
that any polynomial in z,

P ( z ) = a n z n + a n −1 z n −1 + . . . + a 1 z + a 0 , (2.2)

is differentiable on the whole plane (i.e. P(z) is an entire function),

22 complex analysis

and has derivative given by

P0 (z) = nan zn−1 + (n − 1) an−1 zn−2 + . . . + a1 . (2.3)

Consequently, from the Quotient Rule, any rational function of z,

R(z) ≡ PN (z)/PD (z), is differentiable at every point of its domain
where the denominator PD (z) 6= 0. For the purposes of differenti-
ation, polynomial and rational functions in z can be treated as if z
was a real variable.

2.2 Holomorphic functions on C vs functions on R2

In the previous chapter we viewed complex functions of a

complex variable (e.g. f (z)), as somewhat arbitrary mappings from
the xy-plane to the uv-plane. We have individual names for the real
(x) and imaginary (y) parts of z, and for the real (u) and imaginary
(v) parts of f . Any pair of two-variable functions u( x, y) and v( x, y)
gives us a complex function (u + iv) in this sense. But notice that
there is something special about the pair

u1 ( x, y) = x2 − y2 , v1 ( x, y) = 2xy, (2.4) 3
In real calculus we don’t deal with
functions
√ that look at a number like
as opposed to (e.g.) 3 + 4 2 and square the 3 but cube the
4! The interesting calculus functions
treat the number as an indivisible
u2 ( x, y) = x2 − y2 v2 ( x, y) = 3xy. (2.5) module.

That is, the complex function u1 + iv1 treats z = x + iy as a single

"unit" because it can be written x2 − y2 + i2xy = ( x + iy)2 , and thus
respects the complex structure of z = x + iy. However, the formula-
tion u2 + iv2 requires us to break apart3 the real and imaginary parts
of z.
We want to consider functions that are functions of z, rather
than functions of x and y separately. Thus z2 = x2 − y2 + i2xy is
a function we consider "admissible", while functions like Re(z),
Im(z) and z are not, as they require explicit separation the real and
imaginary parts of z. Similarly |z| is not the kind of function we are
looking for since z = |z|/z, while the power series expansion of ez
shows us that it can be written in terms of only functions of z in its
entirety.
It will turn out that this vague criteria we are looking for, which
apparently respects the complex structure of z, can be expressed
simply in terms of differentiability. Thus the functions we want
to consider, that treat z as a complex number, rather than a pair of
coordinates, are functions that are holomorphic (thus the "holos"
etymology, since such a function respects the complex number in
its entirety). We will explore why these notions are equivalent more Figure 2.2: Close to a differentiable
in later chapters, but we can get a sense of this by considering how point zo , the holomorphic function
w = f (z) maps the circular domain to
functions approach a particular value f (zo ) = wo . a circular range, since it can only scale
In the domain D (zo , ε), the circular disc centred at zo , a holo- r by | a| and rotate by Arg ( a) about
point wo = f (zo ), where a = f 0 (zo ) and
morphic function w = f (z) at z ∈ D (zo , ε) has the well defined r = | z − z o |.
 complex differentiation and holomorphic functions 23

approximation as ε → 0,

f ( z ) ≈ wo + a ( z − zo ) =⇒ w − wo ≈ a(z − zo ) (2.6)

where wo = f (zo ) and a = f 0 (zo ) ∈ C is the well defined complex

derivative at zo . If we write a = | a|eiArg a , we see that if we zoom
in close enough to a differentiable point zo , a circular domain
centred at zo must map to a circular range centred at wo , since a
only provides an overall scaling and rotation about wo (see Figure
2.2). Hence holomorphic functions map locally circular domains to
locally circular ranges (with no reflections).
Let’s consider a general (non-holomorphic) function
on the complex plane, g(z) = u( x, y) + iv( x, y), where x = Re z,
and y = Im z, which is equivalent to a 2-dimensional function
g̃ : R2 → R2 , such that g̃({ x, y}) = {u( x, y), v( x, y)}. By analogy
with g̃, we can see that near a point zo = xo + iyo where g(zo ) = wo
and the partial derivatives are well defined, we can approximate
w = g(z) by
 
∂ ∂
g(z) ≈ wo + ∆x u( x, y) + ∆y u( x, y)
∂x ∂y Figure 2.3: For a general non-
  holomorphic function g(z) =
∂ ∂
+ i ∆x v( x, y) + ∆y v( x, y) . (2.7) u( x, y) + iv( x, y) (where x = Re z
∂x ∂y and y = Im z) and close to some
point zo = xo + iyo where the par-
where ∆x = Re(z − zo ) and ∆y = Im(z − zo ). tial derivatives of u( x, y) and v( x, y)
exist, circular domains are generally
Since for a non-holomorphic function there is no particular rela-
mapped to elliptical ranges, since
tionship between the partial derivatives of u( x, y) and v( x, y), we the partial derivatives are, in general,
see (from Figure 2.3) that "small" circular domains centred at zo unrelated.

must map to elliptical ranges centred at wo . That is, a function ap-

proaching along the y direction has a different derivative than when
approaching along the x direction, which leads to a "elliptical"
mapping, that treats x and y differently. Hence, non-holomorphic
functions locally map circular domains to elliptical ones.
The above discussion provides us with a few insights:

• To treat the complex input z as a unit, we want functions that

locally map circular domains to circular ranges with only scaling
and rotations allowed (no reflections – why?).

• The above condition is exactly satisfied by holomorphic func-

tions.

• Holomorphic functions are actually quite restricted compared to

general functions on R2 .

We can focus a bit more carefully on this last point, that holo-
morphic functions f (z) are special compared to general functions
u( x, y) + iv( x, y). Indeed it turns out that the existence of a complex
derivative is a very strong condition to place on a function, and that 4
We will not prove this yet, but can do
any holomorphic function must also be infinitely differentiable4 on so once we develop a few more tools,
the same domain, that is it has well defined (complex) derivatives at like Cauchy’s Integral Theorem.
all orders.
24 complex analysis

2.3 The Cauchy-Reimann Equations

Figures 2.2 and 2.3 gives us a hint that holomorphic functions must
have a special relationship between u( x, y) and v( x, y) that has to
be satisfied. Here we will find this relationship, by calculating the
derivative along two different paths.

Derivatives from different directions

If the function f (z) = u( x, y) + iv( x, y) is differentiable at zo = xo + iyo , the complex derivative
can be found using Definition 2.1. Approaching from the horizontal direction (see Figure 2.4), we
take h = ∆x → 0,

u( xo + ∆x, yo ) + iv( xo + ∆x, yo ) − u( xo , yo ) − iv( xo , yo )

f 0 (zo ) = lim
∆x →0 ∆x
u( xo + ∆x, yo ) − u( xo , yo ) v( xo + ∆x, yo ) − v( xo , yo )
   
= lim + lim i
∆x →0 ∆x ∆x →0 ∆x

We see that the terms in the square brackets are the definitions of the partial derivatives,

∂u ∂v
f 0 (zo ) = ( x o , y o ) + i ( x o , y o ). (2.8)
∂x ∂x
Similarly if we approach zo from the vertical direction we take h = ∆y → 0,

u( xo , yo + ∆y) − u( xo , yo ) v( xo , yo + ∆y) − v( xo , yo )
   
f 0 (zo ) = lim + lim i
∆y→0 i∆y ∆y→0 i∆y
∂u ∂v
= − i ( x o , y o ) + ( x o , y o ). (2.9)
∂y ∂y

In order for the value of the derivative to be the same no matter which direction the limit was
taken in, we require the real and imaginary components of Equations 2.8 and 2.9 to be the same,
resulting in the following theorem.

The Cauchy-Reimann Equations

Theorem 2.5 (The Cauchy-Reimann Equations). Let f (z) =
u( x, y) + iv( x, y) be defined on some open set S containing the
point zo . The first partial derivatives of u( x, y) and v( x, y) exist,
are continuous, and satisfy the Cauchy-Reimann equations:

∂u ∂v ∂u ∂v
= , =− , (2.10)
∂x ∂y ∂y ∂x

at zo = xo + iyo , if and only if f (z) is differentiable at zo . These

conditions hold at all points in S, if and only if f (z) is holomor-
phic on S.
Figure 2.4: A holomorphic function
has the same derivative no matter
(Why do you think we require the partial derivatives to be which direction the limit is taken
continuous in order for the Cauchy-Reimann (CR) equations in. Taking the limit vertically and
to be sufficient conditions for complex differentiability? horizontally allows us to derive the
Cauchy-Reimann equations.
I.e. what happens if the CR equations are satisfied but the
partial derivatives are not continuous?)
 complex differentiation and holomorphic functions 25

Exercise 2.1: CR equations and circular domains

By explicit calculation, show that functions (with con-
tinuous partial derivatives) that obey the Cauchy-Reimann
equations must map locally circular domains to locally circu-
lar ranges, and that such maps must consist of only a scaling
and rotation (with no reflections).

Example 2.3: The exponential function again

Prove that the function f (z) = ez = e x cos y + ie x sin y is
entire (i.e. is holomorphic on the entire complex plane) and
find its derivative.
Solution: Since

∂ x u = e x cos y, and ∂y v = e x cos y

∂y u = −e x sin y, and ∂ x v = e x sin y

the first partial derivatives are continuous and satisfy the

Cauchy-Reimann equations at every point in the plane.
Hence f (z) is entire. Also we notice that

∂u ∂v
f 0 (z) = +i = e x cos y + ie x sin y = f (z)
∂x ∂x
as we expect.

2.4 Holomorphic and Harmonic Functions

Laplace’s equation in 2-dimensions is an extremely useful equation

in mathematical physics. It can describe the behaviour of a variety
of physical systems, such as 2-D electrostatic potentials, or the
stream functions of irrotational incompressible fluid flow.

Laplace’s Equation
Definition 2.3 (Harmonic function). A real-valued function
φ( x, y) is said to be harmonic on a (simply-connected open)
set S if all its second-order partial derivatives exist and are
continuous on S and φ satisfies Laplace’s equation: Other common notations for the
Laplacian operator ∇2 are ∆ or ∇ · ∇.
∂2 φ ∂2 φ
∇2 φ ≡ + 2 =0 (2.11)
∂x2 ∂y

at each point in S.

Taking the partial derivatives of the Cauchy-Reimann equations

gives us

∂2 u ∂2 v ∂2 u ∂2 v
2
= , and 2
=− . (2.12)
∂x ∂x∂y ∂y ∂y∂x
26 complex analysis

Similarly we have
∂2 v ∂2 u ∂2 v ∂2 u
= , and = − . (2.13)
∂y2 ∂x∂y ∂x2 ∂y∂x
The equality of mixed partials (the Clairaut-Schwartz theorem),
then allows us to prove:
Harmonic Real and Imaginary Parts
Theorem 2.6. If f (z) = u( x, y) +iv( x, y) is holomorphic on a
(simply-connected open) set S, then the real functions u( x, y) and
v( x, y) are both harmonic on S.
Note will show later that if f (z)
is holomorphic then u and v have
Definition 2.4 (Harmonic Conjugate). Conversely, if we are
continuous partial derivatives of all
given a function u( x, y) that is harmonic on some open simply- orders.
connected set, we can find some v( x, y), also harmonic, such that
f (z) = u + iv is holomorphic on the same set. The function v( x, y)
is called the harmonic conjugate of u,

Example 2.4: Finding a harmonic conjugate

Construct an analytic function whose real part is

u( x, y) = x3 − 3xy2 + y

Solution: First we verify that u is harmonic on the whole plane:

∂2 u ∂2 u
+ 2 = 6x − 6x = 0.
∂x2 ∂y
Now we have to find the harmonic conjugate of u, v( x, y) such that the Cauchy-Reimann equa-
tions are satisfied. Thus we must have
∂v ∂u
= = 3x2 − 3y2 (2.14)
∂y ∂x
∂v ∂u
and =− = 6xy − 1. (2.15)
∂x ∂y
If we hold x constant and integrate Eq. 2.14 with respect to y we get

v( x, y) = 3x2 y − y3 + ψ( x )

where ψ( x ) is the "constant" of integration (since we only integrate with respect to y). Taking the
partial derivative of v with respect to x we find
∂v
= 6xy + ψ0 ( x ) = 6xy − 1
∂x
where the second equality comes from Eq. 2.15 above. This gives ψ0 ( x ) = −1, so ψ( x ) = − x + a
where a is some constant. Hence the harmonic conjugate of u is

v( x, y) = 3x2 y − y3 − x + a

and the analytic function is

f (z) = x3 − 3xy2 + y + i (3x2 y − y3 − x + a)

= z3 − i ( z − a ).
 complex differentiation and holomorphic functions 27

The Harmonic Conjugate

Theorem 2.7. If u( x, y) is harmonic on an open simply-connected
domain, the harmonic conjugate of u is given by the line integral
Z z̃
v(z̃) = (∂ x u)dy − (∂y u)dx + C (2.16)
z̃o

where C is a constant of integration, and the line integral is

performed over any path from some arbitrary fixed point z̃o =
{ xo , yo } to the point z̃ = { x, y} (i.e. the result is path indepen-
dent).
Figure 2.5: Above in Example 2.4 the
Note that if the domain is not simply-connected (e.g. a path z̃o → z̃ we chose to integrate
over was in two pieces, the vertical
punctured disc) the harmonic conjugate is not always guar- line segment { xo , yo } → { xo , y}
anteed to exist. (In this case the line integral may be path (along which dx = 0) followed by the
horizontal line segment { xo , y} →
dependent.)
{ x, y} (along which dy = 0). Since any
path will do (e.g. the dotted lines), this
one makes the integrals easy!
It may seem surprising that the condition that a complex function
being complex differentiable leads to such strong restrictions on
the functions u and v. This fact does have an extraordinary benefit
though, often problems involving Laplace’s equation on strange
domains can be more easily solved by considering composition of
holomorphic functions.
We know that if we have two functions

f : S1 → S2 , and g : S2 → S3 , (2.17)

that are both holomorphic with S1 , S2 , S3 ⊆ C then the composition

of the two functions,

g ◦ f : S1 → S3 (2.18)

is also holomorphic on S1 (by the chain rule).

Application: Solving Laplace’s Equation on

Weird Domains (by Conformal Mapping)
Suppose we are trying to find a real function u satisfying

∇2 u = 0 (2.19)

in S1 , with boundary condition u = h( x, y) on the boundary

∂S1 .
Figure 2.6: To solve Laplace’s equation
This is of course equivalent to finding a holomorphic func- on the half-annular domain S1 (here
tion, w(z), whose real part satisfies the boundary condition with a Dirichlet boundary condition),
we could use the holomorphic map
on ∂S1 . f (z) = Log z to map onto the rect-
If S1 is an awkward shape, but we can find a holomorphic angular domain S2 where it is easier
to find the harmonic function Re( g).
function f (z) that maps it to a more helpful domain S2 , then
The solution on S1 is then given by
Re[w(z)] = Re[( g ◦ f )(z)].
28 complex analysis

we can define

w = g◦ f
or w(z) = g( f (z)) (2.20)

We are now looking for a holomorphic function g on the nice

domain S2 such that the boundary condition

Re[( g ◦ f )(z)] = h( x, y) on ∂S1 (2.21)

is satisfied. As long as f (z) is invertible we can write the Cultural sidenote: a function that is
holomorphic and invertible is often
equivalent boundary condition on ∂S2 as called a holomorphism.

Re[ g(z)] = h( f −1 (z)) on ∂S2 (2.22)

2.5 Angles under holomorphic maps

Another property of holomorphic functions is that they are confor-

mal. Roughly, speaking this means that they preserve angles. In
order to understand this fully, we need to figure out what we mean
by angles on the complex plane.
Let’s consider an open set S ⊆ C and define γ : [ a, b] → S to be a
curve in S, so we can write

γ(t) = x (t) + iy(t) (2.23)

such that t ∈ [ a, b] is the parameter that traces the curve. We can

assume that γ is differentiable, so that it’s derivative is

γ0 (t) = x 0 (t) + iy0 (t) ∈ C. (2.24)

We can interpret the complex number γ0 (t) as a "vector" in the

direction of a tangent at the point γ(t). Thus the derivative, γ0 (t), if
it is not 0, defines the direction of the curve at the point.
Figure 2.7: The derivative γ0 (t) is in
If two curves γ(t) and η (t) both pass through some point zo , the direction of the tangent to the
such that zo = γ(to ) = η (t1 ), then the tangent vectors γ0 (to ) curve γ at point γ(t).
and η 0 (t1 ) determine an angle θ which is defined to be the angle
between the curves.
If we have two complex numbers z = a + ib and w = c + id, then

zw = ac + bd + i (bc − ad). (2.25)

We notice that ac + bd is what we would expect from the dot or

scalar product of the vectors { a, b} and {c, d}. We can thus define
an equivalent scalar product of the complex numbers to be

hz, wi = Re(zw). (2.26)

The cosine and sin of the angle θ between z and w on the complex
Figure 2.8: The angle between two
plane can be determined using this scalar product, curves is the angle between the
tangent vectors.
hz, wi hz, −iwi
cos θ = , and sin θ = . (2.27)
|z||w| |z||w|
 complex differentiation and holomorphic functions 29

Conformal Maps
Definition 2.5 (A Conformal Map). A mapping that preserves
angles and orientation (handedness) is called a conformal
map.

Theorem 2.8. If f : S → C is holomorphic and f 0 (zo ) 6= 0, at

some zo ∈ S, then the angle between the curves γ and η at zo is
the same as the angle between the curves f ◦ γ and f ◦ η at f (zo )
with the same orientation, and the mapping is conformal at zo .

Proof: We already have a sense, from discussing Figure 2.2,

of how holomorphic functions only allow local scaling and
rotations, and thus preserve angles and orientations, but
here we will give a more formal approach. Let the curves
γ and η be defined as above and γ0 (to ) = z, and η 0 (t1 )=
w. We then have the angle between γ and η at zo given by
Equation 2.27. We must also have ( f ◦ γ)0 (to ) = f 0 (zo )γ0 (to )
and ( f ◦ η )0 (t1 )= f 0 (zo )η 0 (t1 ).
Let f 0 (zo ) = α ∈ C. Then the inner product of the tangent
vectors of f ◦ γ and f ◦ η is given by,

hαz, αwi = Re(αzαw) = ααhz, wi, (2.28)

Figure 2.9: Level curves of u( x, y) =
x2 − y2 (solid) and its harmonic
since αα = | α |2
is real. The angle between f ◦ γ and f ◦ η at conjugate v( x, y) = 2xy (dotted). The
f (zo ) is then given by holomorphic map w(z) = z2 = u + iv
is conformal everywhere except z = 0
hαz, αwi hz, wi since w0 (0) = 0, and will of course
cos θ = = , (2.29) map these level curves to lines of
|αz||αw| |z||w| constant u and v. Note that the level
hαz, −iαwi hz, −iwi curves of u and v meet at right angles,
and sin θ = = , (2.30) since they map to the orthogonal grid,
|αz||αw| |z||w|
except at z = 0 (at this non-conformal
point, angles between curves are
This clearly fails when α = f 0 (zo ) = 0. actually doubled by w(z)).

Note that while maps like z 7→ z also preserve angle, they in-
volve a reflection such that the orientation of the angles is reversed.
Such maps are not considered conformal because of this orientation
change.
This conformal property of holomorphic maps, allows us to
determine another interesting relationship between a harmonic
function u( x, y) and its harmonic conjugate v( x, y). If we define the
level curves of u and v (see Figure 2.9) to be where

u( x, y) = constant (2.31)
v( x, y) = constant. (2.32)

then it must be that on the z-plane the level curves of u are orthog-
onal to the level curves of v everywhere the function w(z) = u + iv
is holomorphic with non-zero derivative, since, on the w-plane,
the lines of constant u are vertical lines, and lines of constant v are
horizontal lines.
30 complex analysis

Figure 2.10: Conformally mapping a

mosaic of Bath Gorgons using different
holomorphic maps (up to constant
scale factors).
 complex differentiation and holomorphic functions 31

2.6 Poles and Branch Cuts

As discussed above some complex functions are only holomorphic

on part of the complex plane. In some locations they may encounter
a pole, where the value of the function is infinite, or a branch cut,
where the function experiences a jump. In this section we will
examine a few examples of these in more detail to prepare us for
contour integration
Types of Singular Points
A singular point is a point at which a mathematical "misbe-
haves", i.e. it is not well defined, or in our case, blows up or
is no longer differentiable. There are few different classes
of singularities which we will encounter in this unit. Let U
be an open subset of C, containing point zo ∈ U, and let f
be a complex differentiable function in the domain U but isolated
excluding point zo (expressed as U \ {zo }). singularities
Definition 2.6 (Removable Singularities). A singularity is
called removable if there exists a holomorphic function g defined
on all of U, such that f (z) = g(z) for all z ∈ U \ {zo }. g(z) is a non-isolated
continuous replacement for the function f . singularity

A removable singularity is a point zo where the func-

tion f (zo ) appears to be undefined, but if we define
g(zo ) = limz→zo f (z) it stays well behaved. The standard
example of a removable singularity is the Sinc function
f (z) = sin(z)/z at point z = 0.

Definition 2.7 (Pole/Non-essential Singularity). A singularity

is called a pole or a non-essential singularity of f if there exists Figure 2.11: The singular points
discussed in Definitions 2.6 - 2.8 are
a holomorphic function g on U and natural number n ∈ N such isolated singularities. That is one
that can always draw a small enough
circle around each one such that no
g(z) other singularity is inside that circle.
f (z) = , ∀z ∈ U \ {zo } (2.33) A Non-isolated singularity, zo , is a
(z − zo )n
singularity that is also the limit point
The smallest number n for which this is true is called the order of of a sequence of other singularities,
such that ∀ real e > 0, there is always
the pole of f at zo . another singularity within the region
|z − zo | < e.
Poles play an important role in contour integration. A first
order pole, i.e. a pole where (z − zo ) f (z) is holomorphic on
U, is often called a simple pole.

Definition 2.8 (Essential Singularity). A singularity at zo is

called essential if ∀n ∈ N, the function g(z) = (z − zo )n f (z) is
not holomorphic at zo .

It can be shown that f has an essential singularity at a point

if and only if the (Laurent) power series expansion has
infinitely many negative powers (more on this later).
An example of an essential singularity is point zo with the
function f (z) = exp(1/(z − zo )).
32 complex analysis

Branch Points
Another type of special point for functions in the complex
plane are branch points.

Definition 2.9 (Branch Point). The point zo is called a branch

point for the complex (multiple) valued function f (z) if the value
of f (z) does not return to its initial value as a closed curve around
the point zo is traced, such that f varies continuously while the
path is traced.

What matters here is the local behaviour of the function f Figure 2.12: Surface plot of Im(log z).
near zo . What may happen on paths that are some distance The different branches of log z arise
because it is multi-valued, with value
away from zo is not relevant. More precisely, this behaviour
continuously spiralling upwards when
must occur for all curves that enclose the point zo and are going around the branch point. Such
sufficiently close to it. a surface indicating the branches is
called a Riemann surface. Different
branches are akin to different "floors"
in the spiral parking garage. (Source:
Wikimedia Commons user Leonid2.)

branch point
for log(z)

Figure 2.13: Two different regions

on which we can choose a branch of
log(z) to be holomorphic, since it does
Consider the function log(z). If we restrict ourselves to the re- not enclose the branch point.
gion U in the figure above to the left we may define log(z) uniquely,
since there is no way to draw a loop that encloses the branch point
z = 0 within the domain U. To define log(z) in U, we can, for exam-
ple, simply choose the angle θ to be between zero and π/2 for any
point z = reiθ ∈ [Link] we could define the θ in region U
to be between 2π and 3π/2.
We can enlarge the region we are looking at to the domain V,
and log(z) in V can still be defined uniquely, as long as we don’t
enclose a branch point to allow loops that will become discontinu-
ous.
 complex differentiation and holomorphic functions 33

Branch Points at Infinity

You may encounter some texts discussing branch points (or
singularities) at infinity, which may a bit difficult to under-
stand. However, it is easy to understand by considering the
inverse map:

w = 1/z (2.34)

such that w → 0 as z → ∞. Then we can examine what hap-

pens at the point w = 0 in the w-plane. For example:

log(z) = − log(w) (2.35)

and since w = 0 is a branch point of − log(w) we can con-

clude that z = ∞ is a branch point of log(z). Similar argu- with a wavy branch cut
ments can also be made about singularities at infinity.

Note that the choice of branch cut is completely arbitrary for a

particular function. Branch cuts are usually, but not always, taken
between pairs of branch points. For log(z) we have branch points
at z = 0 and z = ∞, which we can approach from any direction.
Figure 2.14: Possible different branch
The figure to the right shows multiple ways to choose the branch cuts for the same multi-valued func-
of log(z). All of them are equally valid, though only the first two tion log(z). Top: The principal branch
log(z) = Log(z), with a branch
are easy to write in terms of the principal value Log(z). Each of cut along the negative real axis,
these choices of branch cut joins the branch point at z = 0 with the so that z = reiθ has θ ∈ (−π, π ).
branch point at z = ∞. Middle: A different branch where
log(z) = Log(ze−iπ ) + iπ. Since Log
has the standard branch cut along the
2.6.1 Complex Powers and Inverse Trigonometric Functions negative real axis, Log(ze−iπ ) has a
branch cut along the positive real axis,
One important use of the logarithmic function is to define complex such that θ ∈ (0, 2π ). Bottom: A wavy
branch cut in the complex plane.
powers of z. The definition is motivated by the identity

zn = (elog z )n = en log z (2.36)

for any integer n.

Complex Powers
Definition 2.10 (Complex Powers). If α is a complex constant,
and z 6= 0, then we can define zα by

zα ≡ eα log(z) . (2.37)

Among other things, this means that each value of log(z) leads
to a particular value of zα .

Example 2.5: (−2)i

Find all the values of (−2)i :

Solution: Since log(−2) = Log2 + i (π + 2kπ ) we have

(−2)i = ei log(−2) = eiLog2 e−π −2kπ , k ∈ Z. (2.38)

34 complex analysis

It is clear that each branch of log z yields a branch of zα . For

example using the principal branch of log z we obtain the principal
branch of zα , namely eαLogz .
Since ez is entire, and Log(z) is analytic in the slit domain C \
(−∞, 0], the chain rule implies that the principal branch of zα is also
analytic in the slit domain.
Other branches can be constructed using other branches of
log(z), with different holomorphic domains.
"Branch chasing" — figuring out which branch to choose —
for complicated functions is often a tedious task; fortunately, for
simple applications this is not often necessary. We examine some of
the subtleties involved in the examples below.
Example 2.6: Clip the Branches!

Define a branch of (z2 − 1)1/2 that is holomorphic in the exte-

rior of the unit circle, |z| > 1.
Solutions: Our task, restated, is to find a function w = f (z)
that is holomorphic outside the unit circle and satisfies Figure 2.15: Bruce Banner consults
with the Ancient One about his choice
of branch cut.
w2 = z2 − 1. (2.39)

Note that the principal branch of (z2 − 1)1/2 , namely,

2 −1)
e(1/2)Log(z . (2.40)

will not work, since it has branch cuts wherever z2 − 1 is

negative real, and this constitutes the whole y-axis as well
as a portion of the x-axis. But if we experiment with some
alternative expressions for w we are led to consider the
solution

1 1/2
 
w = z 1− 2 , (2.41)
z
such that the principal branch of (1 − 1/z2 )1/2 , i.e.,
2
e(1/2)Log(1−1/z ) , has branch cuts where 1 − 1/z2 is nega-
tive real, and this only occurs when 1/z2 is real and greater
than one, i.e. the cut is between the segment on the real line
[−1, 1]. Thus,
2)
w = f (z) = ze(1/2)Log(1−1/z (2.42)

satisfies the requirement of being

holomorphic outside the unit circle.
branch cuts for branch cut for
 complex differentiation and holomorphic functions 35

Since we know that trigonometric functions can be expressed in

terms of exponentials, and the inverses of exponentials are multi-
valued logarithms, it should come as no surprise that the inverse of
trig functions are logarithms.
The Inverse Sine Function
The inverse sine function w = sin−1 z is defined by the equa-
tion

z = sin w, (2.43)

and is a multi-valued function given by

w = sin−1 z = −i log[iz + (1 − z2 )1/2 ]. (2.44)

Proof: From the equation

eiw − e−iw
z = sin w = (2.45)
2i
we find that

e2iw − 2izeiw − 1 = 0. (2.46)

Using the quadratic formula we can solve for eiw :

eiw = iz + (1 − z2 )1/2 , (2.47)

where of course the square root has two possible values. The
expression for w above thus follows.

Note that we can obtain a branch of the multi-valued function

sin−1 (z) by first choosing a branch of the square root, and then
selecting a suitable branch of the logarithm.

Example 2.7: Principal Branch of Sin−1

Suppose z is real and lies in the interval (−1, 1). If the princi-
pal values are used for the terms in (2.44), what is the range
of sin−1 z?
Solution: With principle values we have
2
Sin−1 z ≡ −iLog[iz + e(1/2)Log(1−z ) ]. (2.48)
For |z| = | x | < 1, clearly 1 − z2 lies in the interval (0, 1], and
its Log is real. Hence the exponential term is positive real.
Consequently the bracketed expression lies in the right half
plane. In fact, it also lies on the unit circle, since
q
|iz + (1 − z2 )1/2 | = x2 + (1 − x2 ) = 1. (2.49)
Taking the Log and multiplying by i, we get
π π
− < Sin−1 x < , (2.50)
2 2
which matches the normal convention for arcsin.
36 complex analysis

Exercise 2.2: Complex Inverse Trig

Find the inverse cosine and inverse tangent functions, and
show they are given by

cos−1 z = −i log[z + (z2 − 1)1/2 ] (2.51)

i i+z
tan−1 z = log , ( z 6 = ± i ). (2.52)
2 i−z

Exercise 2.3: Derivatives of Inverse Trig

Use the chain rule to determine the derivatives of sin−1 z,

cos−1 z and tan−1 z.
3
Complex Integration and Cauchy’s Integral Theorem

In the previous chapter, we discussed how the derivative on

the complex plane is modified by the extra degree of freedom with
which a point can be approached. This two-dimensional property
of the complex plane also modifies integration, since a general
(line) integral between two points will require us to choose one
of an infinite number of possible paths. In this chapter, we will
find that if a function f (z) is the complex derivative of an entire
function F (z), then we can use the standard fundamental theorem
of calculus approach and find that the integral only depends on
F (z) evaluated at the end points. We will also find that if a function
is holomorphic inside (and on) a closed loop, then its integral over
that loop must be zero. This result is known as Cauchy’s integral
theorem.

3.1 Contours

Contour integrals are carried out by integrating along a series of

curves in the complex plane. While we have an intuitive grasp for
what a curve is, let’s be a bit more explicit about the definitions of
smooth curves, contours, and the interior and exterior of a domain.
Smooth curves
Definition 3.1 (A smooth arc). A set of points γ in the complex
plane is said to be a smooth arc if it is the range of some conin-
uous complex-valued function z = z(t), where t ∈ [ a, b] is a
real-valued parameterisation, and satisfies the following conditions

(1) z(t) has continuous derivative on [ a, b],

(2) z0 (t) never vanishes on [ a, b],

(3a) z(t) is one-to-one on [ a, b].

smooth arc smooth closed not smooth not smooth

curve
38 complex analysis

Definition 3.2 (Smooth Closed Curves). The set of points γ is

called a smooth closed curve if it is the range of some continu-
ous function z = z(t) with t ∈ [ a, b], satisfying the conditions (1)
and (2) above, as well as

(3b) z(t) is one-to one on the half-open interval [ a, b) but z(b) =

z( a) and z0 (b) = z0 ( a).

Both smooth arcs and smooth closed curves are considered

to be smooth curves. Smooth arcs have distinct end points,
while smooth closed curves have their endpoints joined.

You can imagine that an artist with a pen on a paper, drawing

a smooth curve. She is not allowed to lift the pen from the paper
during the sketch; mathematically, we are requiring that z(t) is
continuous. Second we insist that the curves be drawn with an
even, steady stroke, specifically, the pen point must move with
a well-defined (finite) velocity that must also vary continuously.
Finally, we require that no point on the curve be drawn twice, the
Figure 3.1: The directed smooth arc
artist cannot cross over her own points. The only exception to described by z(t), taken from z( a) to
this is for smooth closed curves, where we allow the endpoints to z ( b ).
coincide.
initial/final
Suppose that the artist is able to draw a smooth arc between two point
points which obeys the rules we have laid out above. Then it should
be clear that there are exactly two natural orderings over which we
can traverse the smooth arc. We can start at one point and go to the choice of
direction
other, or vice versa.
A smooth arc, together with a choice of the order over which
to traverse it is called a directed smooth arc. The ordering can be
indicated by an arrow on the arc.
A directed smooth closed curve is slightly more complicated,
and we must decide which direction to traverse the curve along Figure 3.2: The directed smooth closed
curve is given by a smooth closed
from an initial point. We say say that a smooth closed curve is curve with an initial/end point and a
directed when we have (a) designated an initial point, and (b) choice of direction (counterclockwise).
chosen one of the two directions from this point.

Contours
Definition 3.3 (Contours). A contour Γ is either a single point
zo or a finite sequence of directed smooth curves (γ1 , γ2 , . . . , γn )
such that the terminal point of γk coincides with the initial point of
γk+1 , for each k = 1, 2, . . . , n − 1.

Definition 3.4 (Simple Closed Contours). A contour is said

to be closed or a loop if its initial and terminal points coincide.
A simple closed contour is a closed contour with no repeated
points other than the initial/final point.
 complex integration and cauchy’s integral theorem 39

contour simple closed contour closed

contour contour

With any simple closed contour, it turns (via a theorem due

to French mathematician Camille Jordan) out that we can always
define an interior domain, and an exterior domain.
The Jordan Curve Theorem
The formal proof of the Jordan Curve
Theorem is quite involved, so we will
Theorem 3.1 (The Jordan Curve Theorem). A simple closed not attempt it here.
contour separates the plane into two domains, each having the
curve as its voundary. One of these domains, called the interior, is
bounded; the other, called the exterior, is unbounded.

Interior

Interior

Exterior

Exterior

The Orientation of a Contour

Alternate names for counterclockwise
Imagine yourself walking along the complex plane, follow- include "anticlockwise", "lefthandwise",
ing the path of a contour. If the interior of the contour is and "widdershins" (meaning "counter
the sun’s direction"). The use of
on your left we would consider the contour to have a pos- widdershins is often contrasted
itive orientation. If the interior is on your right, we call it with "deisul" (clockwise - meaning
"following the sun") in folklore.
negative.

Definition 3.5 (Orientation of Contours). A contour is said to

be positively oriented if the interior domain is on the left as the
contour is traversed.
A contour is said to be negatively oriented if the interior domain
is on the right as the contour is traversed.

A circular contour is positively oriented if it is followed

counterclockwise, and negatively oriented if the it is fol-
lowed clockwise.
The contours in the figures above are both positive.

With these definitions of the contours along which we can inte-

40 complex analysis

grate, we are now ready to discuss contour integrals.

3.2 Contour Integrals

3.2.1 Complex-valued Integrals over the Real Line

Suppose we have a continuous function f (t), that takes a value
from the real interval [ a, b] and maps it to the complex plane.

f : [ a, b] → C. (3.1)

We can write f in terms of its real and imaginary parts

f (t) ≡ u(t) + iv(t), (3.2)

and the indefinite integral of f can just be written in terms of

normal integrals of real functions
Z Z Z
f (t)dt = u(t)dt + i v(t)dt (3.3)

Exercise 3.1: Integration By Parts

Verify that integration by parts is valid, i.e.,
Z Z
f (t) g0 (t)dt = f (t) g(t) − g(t) f 0 (t)dt, (3.4)

assuming that f 0 and g0 exist and are continuous. (Hint:

The proof is the same as in ordinary real calculus, from the
product rule)

Similarly, we can define the definite integral of F over [ a, b] to be

Z b Z b Z b
f (t)dt = u(t)dt + i v(t)dt. (3.5)
a a a

Thus the integral is defined in terms of the ordinary integrals

of the real functions u and v. Consequently, by the fundamental
theorem of calculus, the function
Z t
F (t) = f (s)ds (3.6)
a

is differentiable, and its derivative is f (t).

Exercise 3.2: Absolute Value of an Integral vs
Integral of an Absolute Value
Show that using simple properties of the integral of real-
valued functions, one has the inequality,
Z b Z b
f (t)dt ≤ | f (t)|dt (3.7)
a a

(Hint: Break it up into Riemann sums, and use the triangle

inequality.)
 complex integration and cauchy’s integral theorem 41

Integral Along A Curve

Definition 3.6 (Integral Along a Curve). Let smooth curve
γ : [ a, b] → U ⊆ C be defined on a closed interval of real numbers
[ a, b], and let f be a continuous function on the open set U. We
can define the integral of f along γ to be
Z b
dγ
Z
f (z)dz = f (γ(t)) dt. (3.8)
γ a dt

Note that it does not really matter what parameterisation we use,

as can be easily seen defining the new parameterisation as s(t), and
applying the chain rule to convert an integral over s to one over t.
The result is the same no matter what parameterisation is used.

Example 3.1: A Very Important Example

Compute the integral Cr (z − zo )n dz, where n is an inte-

H

ger, and Cr is the circle |z − zo | = r traversed once in the coun-

terclockwise direction.
Solution: A suitable parameterisation for Cr is given by z =
γ(t) = zo + reit , with t ∈ [0, 2π ]. Setting f (z) = (z − zo )n , we
have
Figure 3.3: The contour Cr for Example
f (γ(t)) = (zo + reit − zo )n = r n eint , and γ0 (t) = ireit 3.1.

Hence, we must have

I Z 2π
(z − zo )n dz = (r n eint )(ireit )dt
Cr 0
Z 2π
n +1
= ir ei(n+1)t dt. (3.9)
0

If n 6= −1 we have
I Z 2π
(z − zo )n dz = ir n+1 ei(n+1)t dt
Cr 0
2π
e i ( n +1) t
= ir n+1 = 0. (3.10)
i ( n + 1) 0

However, if n = −1 then we must have

I Z 2π Z 2π
n 0 0
(z − zo ) dz = ir e dt = i dt = 2πi (3.11)
Cr 0 0

Thus, regardless of the value of r,


I 0 for n 6= −1,
(z − zo )n dz = (3.12)
Cr 2πi for n = −1.

We will see that this calculation plays an important role

later.
42 complex analysis

3.2.2 Bounding Contour Integrals

Many times, in theory and in practice, it is not actually necessary to

evaluate a contour integral. What may be required is simply a good
upper bound on its magnitude. We therefore turn to the problem of
estimating contour integrals.
Suppose that function f is continuous on the directed smooth
curve γ and that f (z) is bounded by the constant M on γ : [ a, b] →
C, i.e.,

| f (z)| ≤ M, ∀z on γ. (3.13)

Let’s divide the parameterisation of γ up into n partitions each of

length ∆t = (b − a)/n, The integral can be approximated by the
Riemann sum
Z n

γ
f (z)dz ≈ ∑ f (tk )∆zk (3.14)
k =1

where tk ≡ a + k∆t, and ∆zk = γ(tk ) − γ(tk−1 ).

We have by the triangle identity

n n n
∑ f (tk )∆zk ≤ ∑ | f (tk )||∆zk | ≤ M ∑ ∆zk (3.15) Figure 3.4: An integral along a di-
k =1 k =1 k =1 rected smooth curve can be approx-
imated using a Riemann sum if you
Furthermore, we notice that the sum of the lengths ∆zk cannot be partition the parameterisation of curve
greater than the length `(γ) of γ (since chords are always shorter γ.
than the arc length), hence

n
∑ f (tk )∆zk ≤ M`(γ). (3.16)
k =1

Taking the limit as n → ∞ gives us

Z
f (z)dz ≤ M `(γ). (3.17)
γ

Applying this to a finite sum of different curves Γ = ∑m γm , we

have proved:

Bounding a Contour Integral

Theorem 3.2. If f is continuous on the contour Γ and if | f (z)| ≤
M for all z on Γ, then
Z
f (z)dz ≤ M `(Γ). (3.18) This theorem is sometimes called
Γ Darboux’s inequality.

In particular, we have,
Z  
f (z)dz ≤ max | f (z)| × `(Γ). (3.19)
Γ z on Γ
 complex integration and cauchy’s integral theorem 43

Integrands with Antiderivatives

† Such a function F, where F 0 (z) =
Theorem 3.3. Suppose that there exists† a holomorphic function f (z), is sometimes called a primative
F (z) on U such that F 0 (z) = f (z), and that curve γ has value of f , or an antiderivative of f . Note
α = γ( a), and β = γ(b) at the end points. Then, that this is not guaranteed to exist for
arbitrary f .
Z
f (z)dz = F ( β) − F (α) (3.20)
γ

Proof: We know from the definition above

Z Z b
f (z)dz = F 0 (γ(t))γ0 (t)dt.
γ a

By the chain rule, the expression under the integral sign is

d
the derivative dt F (γ(t)). Then by ordinary calculus on the
real and imaginary parts of F, the integral is equal to
Z b
f (z)dz = F (γ(t)) = F (γ(b)) − F (γ( a)). (3.21)
γ a

In general for such an f , if a path Γ consists of n connected

curves γ1 , . . . , γn , with z j being the end point of the jth curve, then
we can chain together the above to find
Z
f (z)dz = F (z1 ) − F (z0 ) + . . . + F (zn ) − F (zn−1 )
Γ
= F ( z n ) − F ( z0 ). (3.22)

Thus, if the antiderivative of the integrand exists, the integral over a

path Γ acts exactly as we expect from regular calculus.

Lemma 3.1. If f is a continuous function on U that has a holomorphic

antiderivative F, and Γ is any closed path, then
Z
f (z)dz = 0. (3.23)
Γ

This follows from the theorem above. It is also straightforward to

prove the following: Theorems 3.3 and 3.4 are the complex
analogues of the path independence
Theorem 3.4. Let U be a connected open set, and let f be a continuous of a potential function that is the
antiderivative of a vector field.
function on U. If the integral of f along any closed path in U is equal to
0, then f has an antiderivative F on U, that is, there exists a function F
which is holomorphic such that F 0 (z) = f (z).

It may seem that Theorem 3.4 is a little bit useless, since it seems
that we’d have to test all possible closed paths and show the in-
tegrals are all zero. In the next section we will see that Cauchy’s
integral theorem will provide a simple condition for this property
to hold.
For now we can simply summarise the two theorems above by
saying that a given continuous function has an antiderivative in
domain U if and only if its integral around every loop in U is zero.
44 complex analysis

3.3 Cauchy’s Integral Theorem

In the last section we saw that if a continuous function f possesses

a holomorphic anti-derivative in a domain U, its integral around
any loop in U is zero and vice versa. Now we are going to show
how this property ties in with whether f is itself holomorphic or
not. Our first task will be to develop the necessary geometry.

3.3.1 Deformation of Contours

The critical notion in this regard is the continuous deformation
of one loop into another within the domain U. Deformations are
quite easily visualised but somewhat harder to express in precise
mathematical language. (We already discussed these briefly when Figure 3.5: Deforming contours can be
reviewing the notion of simply connected domains.) In informal thought of as stretching and shrinking
an infinitely elastic rubber band.
terms, we say that a loop Γ0 can be continuously deformed into
a loop Γ1 in the domain U if Γ0 (considered as an elastic string
with indicated orientation) can be continuously moved, stretched,
and shrunk about the complex plane without leaving U, in such a
manner that it ultimately coincides with Γ1 .

Figure 3.6: Examples of continuous

deformations of contours within the
domain U.

Reversing orientation of a contour by

Expanding circular contours Shrinking a contour to a point shrinking to a point

Continuous Deformation
Definition 3.7 (Continuously Deformable Loops). The loop
Γ0 is said to be continuously deformable to the loop Γ1 in the
Loops that can be continuously
domain U, if there exists a function z(s, t) continuous on the unit deformed into one another are some-
square 0 ≤ s ≤ 1, 0 ≤ t ≤ 1, that satisfies the following conditions times called homotopic. The deforma-
tion between the two contours is called
(i) For each fixed s in [0, 1], the function z(s, t) parameterises a a homotopy.
closed contour in U.

(ii) The function z(0, t) parameterises the closed contour Γ0 .

(iii) The function z(1, t) parameterises the closed contour Γ1 .

 complex integration and cauchy’s integral theorem 45

Example 3.2: Deformation in an Annulus

By finding a deformation function z(s, t), prove that the loop
Γ0 : z = e2πit , for 0 ≤ t ≤ 1, can be continuously deformed
to the loop Γ1 : z = 2e2πit , for 0 ≤ t ≤ 1 in the domain U
consisting of an annulus with 21 < |z| < 3.
Solution: The intermediate loops Γs , for 0 ≤ s ≤ 1 are con-
cetric circles with radii varying from 1 to 2. The function

z(s, t) = (1 + s)e2πit , for 0 ≤ s ≤ 1, 0 ≤ t ≤ 1 (3.24)

is a deformation function which takes Γ0 to Γ1 .

A few elementary observations about continuous deformations

are in order. First, notice that if z(s, t) generates a deformation of
loop Γ0 into loop Γ1 , then z(1 − s, t) deforms Γ1 into Γ0 . Furthermore,
if in a given domain Γ0 can be deformed to a single point and Γ1
can be deformed to a point, then Γ0 can be deformed into Γ1 .
It should be clear that simply connected domains, i.e. connected
domains without holes, are of particular interest to the study of
deformed loops since we have,
Definition 3.8 (Simply Connected Domains). Any domain U possess-
ing the property that every loop in U can be deformed (through U) to a
point is called a simply connected domain.

Deformation Invariance Theorem

Theorem 3.5 (Deformation Invariance Theorem). Let f be a
function that is holomorphic on a domain U, containing closed
contours Γ0 and Γ1 . If these loops can be continuously deformed
into on another in U then
Z Z
f (z)dz = f (z)dz (3.25)
Γ0 Γ1

A rigorous proof of the Deformation Invariance Theorem is

beyond the scope of this course, however, we will instead
attempt to prove a slightly weaker version of this theorem
for the special case when the two contours are linked by
a deformation function z(s, t) whose second-order partial
derivatives are continuous, and that f 0 (z) is continuous.

(Weaker) Proof: Assume that the deformation function z(s, t)

has continuous partial derivatives up to second order for
0 ≤ s ≤ 1, and 0 ≤ t ≤ 1, and f 0 (z) is continuous. Now, for
each fixed s the equation z = z(s, t), 0 ≤ t ≤ 1, defines the
loop Γs in U. Let I (s) be the integral of f along this loop,
Z 1
∂z(s, t)
Z
I (s) ≡ f (z)dz = f (z(s, t)) dt. (3.26)
Γs 0 ∂t
46 complex analysis

We wish to take the derivative of I (s) with respect to s. The

assumptions guarantee that the integrand above is continu-
ously differentiable, so with constant limits of integration,
we are allowed to bring the derivative inside the integral
(this is called the Liebniz Integral Rule; see margin),
Theorem 3.6 (The Liebniz Integral
∂2 f
Z 1 
dI (s) 0 ∂z ∂z
= f (z(s, t)) + f (z(s, t)) dt. (3.27) Rule). Let f ( x, t) be a function such
ds 0 ∂s ∂t ∂t∂s that both f ( x, t) and its partial derivative
∂ x f ( x, t) are continuous in t and x.
On the other hand observe that Suppose that the boundary functions a( x )
and b( x ) are both continuous and have
∂2 z
 
∂ ∂z ∂z ∂z continuous derivatives. Then,
f (z(s, t)) = f 0 (z(s, t)) + f (z(s, t)) . (3.28)
∂t ∂s ∂t ∂s ∂t∂s d
Z b( x )
f ( x, t)dt
dx a( x )
Because of the equality of mixed partials, the expression
= f ( x, b( x ))b0 ( x )
above is the same as the the integrand of dI/ds above. Thus, − f ( x, a( x )) a0 ( x )
we have Z b( x )
Z 1   + ∂ x f ( x, t)dt.
dI (s) ∂ ∂z a( x )
= f (z(s, t)) dt (3.29)
ds 0 ∂t ∂s This is straightforward to prove
using the fundamental theorem of
∂z ∂z
= f (z(s, 1)) (s, 1) − f (z(s, 0)) (s, 0). (3.30) calculus.
∂s ∂s
But since each Γs is closed we have z(s, 1) = z(s, 0) and
∂s z(s, 1) = ∂s z(s, 0), so that dI/ds is zero, and consequently
I (s) must be a constant. In particular I (0) = I (t) so that
Z Z
f (z)dz = f (z)dz. (3.31)
Γ0 Γ1

A rather straightforward, but important, consequence of the

deformation invariance theorem is known as Cauchy’s integral
theorem:
Cauchy’s Integral Theorem
Theorem 3.7 (Cauchy’s Integral Theorem). If f is holomorphic
in a simply connected domain U, and Γ is any closed contour in U,
then
Z
f (z)dz = 0 (3.32)
Γ

Proof: In a simply connected domain, any loop can be

shrunk to a point. The integral of a continuous function
over a single point is zero, therefore by the deformation
invariance theorem, the integral over any loop that can be
continuously deformed to a point must be zero.
This leads us naturally to apply the results above to Theo-
rems 3.3 and 3.4 from the previous section:

Theorem 3.8. If a function is holomorphic on a simply con-

nected domain, then in that domain it has an anti-derivative, its
contour integrals are independent of path, and its loop integrals
vanish.
 complex integration and cauchy’s integral theorem 47

Example 3.3: Very Important Example Revisited

In the previous section we showed that if Cr is any positively
oriented circle centred at zo with any radius r, and n is an
integer then

I 0 for n 6= −1,
(z − zo )n dz = (3.33)
Cr 2πi for n = −1.

If n is a positive integer or zero, then Cauchy’s integral

theorem applies to the integral, since the integrand is holo-
morphic in the entire complex plane, which is obviously
simply connected.
(z − zo )n has an anti-derivative (z − zo )n+1 /(n + 1), and the
loop integral must be zero.
If n is negative, then (z − zo )n is only analytic in the punc-
tured plane, with the point zo deleted. This domain is not
simply connected, so the theorems of this section do not
apply.
In fact for n = −1 the function (z − zo )n does not even have
an antiderivative in the punctured plane (since any branch
of log(z) will have a discontinuity at a branch cut), and the
loop integral fails to vanish.
For n ≤ −2 (z − zo )n still has an antiderivative,
(z − zo )n+1 /(n + 1), away from point zo , and the loop in-
tegral is zero.
Thus we see that both cases can occur when the domain is
not simply connected.

The main value of the deformation invariance theorem is that it

allows us to replace complicated contours with more familiar ones,
for the purpose of integration.

Example 3.4: Elliptical contour

Evaluate Γ 1z dz where Γ is the ellipse defined by x2 + 4y2 = 1

R

traversed once in the positive (counter-clockwise) sense.

Solution: The integrand 1/z is holomorphic everywhere on
the complex plane except for the origin. We can continu-
ously deform the this contour to the unit circle Γ0 oriented
positively. Thus we must have Figure 3.7: Deforming an elliptical
contour to a circular one.
1 1
Z Z
dz = dz = 2πi (3.34)
Γ z Γ0 z
48 complex analysis

Example 3.5: Missing the poles

Evaluate
ez
I
dz (3.35)
|z|=2 z2 −9

where the integral is around the circle with |z| = 2 once in

the positive (ccw) direction.
Solution: The integrand ez /(z2 − 9) is holomorphic every-
where except at z = ±3, where the denominator vanishes.
We can see that since the contour does not contain either
of these poles, it can be shrunk to a point in the domain of
analyticity, and thus the integral is zero. We can obtain the Figure 3.8: Contour and poles for
same result by applying Cauchy’s theorem directly. Example 3.5.

Example 3.6: In or Out?

Determine the possible values for

1
I
dz, (3.36)
Γ z−a
where Γ is any circle not passing through z = a, traversed
once in the counterclockwise direction.
Solution: The integrand is holomorhic in the domain
U = C \ { a} consisting of the complex plane with the point
z = a removed. If this point lies exterior to Γ, then Γ can be
continuously deformed to a point in U, and so the integral
vanishes. If a lies in the interior of Γ, the contour can be
continuously deformed in U to a positively oriented circle
centred at z = a, and thus by Equation (3.12) in Example 3.1,
we have

I
dz 0 if a lies outside Γ,
= (3.37)
Γ z−a 2πi if a lies inside Γ.

Example 3.7: Two Poles One Contour

Find
3z − 2
Z
dz (3.38)
Γ z2 − z
where Γ is the simple closed contour shown in Figure 3.9.

Solution: We don’t need an exact description of the contour

Γ. Since the integrand f (z) = (3z − 2)/(z2 − z) is holomor-
phic in U = C \ {0, 1}, i.e., the entire complex plane with the
points z = 0 and z = 1 removed, we can deform Γ into the
 complex integration and cauchy’s integral theorem 49

barbell shaped contour shown in Figure 3.9 without chang-

ing the value of the integral (by the Deformation Invariance
Theorem). This integral can be further simplified by noting
that the integration along the line segment is to the left for
one part of the contour, and to the right later in the contour,
at the same place in the complex plane and therefore must
cancel out.
Thus we must have
Z Z Z
f (z)dz = f (z)dz + f (z)dz (3.39)
Γ C0 C1

Figure 3.9: Contours and poles for

where C0 is the counterclockwise circular contour around
Example 3.7.
the pole at z = 0 and C1 is the counterclockwise circular
contour around the pole at z = 1.
We will derive a much more powerful and flexible method
to evaluate the integrals around these poles when we cover
Cauchy’s Residue Theorem, however, for now we can use a
partial fraction expansion to rewrite the integrand as

3z − 2 A B
= + ,
z2 − z z z−1
Az − A + Bz
= ,
z ( z − 1)
⇒ A = 2, B = 1.

which gives us

3z − 2
  I  
2 1 2 1
I I
dz = + dz + + dz,
Γ z ( z − 1) C0 z z−1 C1 z z−1
2dz dz 2dz dz
I I I I
= + + + ,
C0 z C0 z − 1 C1 z C1 z − 1
| {z } | {z } | {z } | {z }
2×2πi 0 0 2πi

= 6πi.

3.4 Cauchy’s Integral Formula

We now turn to the second important result of this chapter due

to Cauchy, Cauchy’s Integral Formula. Given a function f that is
holomorphic inside and on the simple closed contour Γ, we know
H
from Cauchy’s Integral Theorem that Γ f (z)dz = 0. However, if we
consider the integral

f (z)
I
dz,
Γ z − zo

where zo is an interior point to the contour Γ, we should not ex-

pect that this integral vanishes, since the integrand is no longer
holomorphic inside the contour, and Cauchy’s theorem no longer
applies.
50 complex analysis

In fact, rather remarkably, we will show that for all zo inside Γ,

the value of this integral is proportional to f (zo ).

Cauchy’s Integral Formula

Theorem 3.9 (Cauchy’s Integral Formula). Let Γ be a simple
closed positively oriented contour. If f is holomorphic in some
simply connected domain U containing Γ and zo is any point in
the interior of Γ, then

1 f (z)
I
f (zo ) = dz. (3.40)
2πi Γ z − zo
Proof: The function f (z)/(z − zo ) is holomorphic every-
where in U except for point zo . Hence by the Deformation
Invariance Theorem, we can deform the contour freely to
the circle Cr centred at zo (see Figure 3.10) which is also
positively oriented.
We can then write, Figure 3.10: Contours and pole for the
proof of Cauchy’s Integral Formula.
f (z) f (z)
I I
dz = dz, (3.41)
Γ z − zo Cr z − zo
f (zo ) f (z) − f (zo )
I I
= dz + dz. (3.42)
Cr z − zo Cr z − zo
| {z }
f (zo )×2πi

where we’ve just split the denominator into two parts since
f (z) = f (zo ) + ( f (z) − f (zo )). For the second integral we
can shrink the circle Cr to arbitrarily small radius r. Since f
is holomorphic and thus does not blow up anywhere on Cr ,
we can define the maximum magnitude

Mr ≡ max [| f (z) − f (zo )|]. (3.43)

z on Cr

Then for z on Cr we have, by Theorem 3.2,

f (z) − f (zo ) | f (z) − f (zo )|

I Z
dz ≤ dz (3.44)
Cr z − zo Cr r
Mr Mr
≤ × `(Cr ) = × 2πr, (3.45)
r r
so that we find
f (z) − f (zo )
I
dz ≤ 2πMr . (3.46)
Cr z − zo

But since f (z) is holomorphic, and thus a continuous func-

tion, we must have Mr → 0 as r → 0, so this term must
vanish as we shrink the contour. Therefore we are left with
f (z)
I
dz = 2πi f (zo ), (3.47)
Γ z − zo
which proves the theorem.
 complex integration and cauchy’s integral theorem 51

One remarkable consequence of Cauchy’s formula is that by

merely knowing the values of the holomorphic function f on Γ we
H
can compute the integral Γ [ f (z)/(z − zo )]dz for any zo , and hence
all values of f inside Γ. In other words, the behaviour of a function
holomorphic in a region is completely determined by its behaviour
on the boundary!
Example 3.8: Using Cauchy’s Integral Formula
Compute the integral

ez + sin z
I
dz (3.48)
Γ z
where Γ is the circle |z − 2| = 3 traversed once in the counter-
clockwise direction.
Solution: The function f (z) = ez + sin z is holomorphic inside
and on Γ, and the point zo = 0 lies inside this circle. Hence
we can use Cauchy’s Integral Formula to get

ez + sin z
I
dz = 2πi f (0) = 2πi (e0 + sin 0) = 2πi.
Γ z
(3.49)

Example 3.9: Using Cauchy’s Int. Formula Again

Evaluate the integral
cos z
Z
dz (3.50)
Γ z2 − 4
along the contour shown in Figure 3.11.
Figure 3.11: Contour and poles for
Solution: The integrand fails to be holomorphic at points Example 3.9.
z = ±2. However, only one of these (z = 2) is inside the
contour Γ. Thus we can recast the integrand as

cos z (cos z)/(z + 2)

Z Z
dz = dz
Γ z2 − 4 z−2
Γ
cos z
= 2πi
z + 2 z =2
cos 2
= 2πi
4
iπ cos 2
= .
2

We will return to Cauchy’s Integral Formula shortly, but first we

need to consider the general properties of complex power series
and introduce the concept of analyticity.
4
Series Representations of Complex Functions

4.1 Convergence of Series

In Chapter 1, we briefly reviewed the concept of sequences. Here

we will develop a few more tools to deal with series, which, as you
should already know from 1st year Maths, can form a sequence
made of partial sums.
Formal Definition of a Series
Definition 4.1 (Series). A series is an expression of the form
c0 + c1 + c2 + . . ., or, ∑∞ k =0 ck where the terms ck are complex
numbers. The nth partial sum of a series, usually denoted Sn , is
defined to be the sum of the first n + 1 terms, i.e. Sn ≡ ∑nk=0 ck . If
the sequence of partial sums {Sn } has a limit S, the series is said
to converge to S, and we write S = ∑∞ k =0 ck . A series that does
not converge is said to diverge.

Notice that the notion of a convergence of a series has been

defined in terms of convergence for a sequence.
Clearly one way to demonstrate that a series converges to S is to
show that the remainder after summing the first n + 1 terms, i.e.,
n
S− ∑ ck → 0 (4.1)
k =0
as n → ∞. Here we will use this to prove that the Geometric Series
converges only inside the unit circle.
The Geometric Series
Theorem 4.1 (Geometric Series). The geometric series 1 + z2 +
z3 + z4 + . . . converges to the value 1/(1 − z) for any complex We will use this extensively in the
number z with |z| < 1. proofs to come!

Proof : Let Sn (z) = 1 + z + . . . + zn . Multiplying by z we find

zSn = Sn + zn+1 − 1

Solving for Sn we get

1 − z n +1
Sn = .
1−z
 series representations of complex functions 53

The series converges to S = 1/(1 − z) iff S − Sn → 0 as n → 0,

1 1 − z n +1
S − Sn = − (4.2)
1−z 1−z
z n + 1
= (4.3)
1−z
Taking the limit n → ∞ we must then have
∞
1
∑ zn = 1 − z , iff |z| < 1, (4.4)
n =0

since zn+1 → 0 only for z in the interior of the unit circle.

Another important way to establish the convergence of a series

involves comparing it with another series whose convergence is
known.
The Comparison Test
Theorem 4.2 (The Comparison Test). Suppose that the terms ck
satisfy the inequality
| c k | ≤ Mk (4.5)
where 0 ≤ Mk ∈ R, for all integers k larger than some number K.
Then, if the series ∑∞ ∞
k =0 Mk converges, so does ∑k =0 ck .

Proof: This theorem seems very obvious, however, we will

prove it just to be pedantic. This is most easily done using
Theorem 1.6, the Cauchy Criterion for the convergence of
sequences.
For any real ε > 0, since the series ∑k Mk converges, ∃ N ∈ N
such that the partial sums must obey,
! !
n+m n
∑ Mk − ∑ Mk < ε, ∀n ≥ N, ∀m > 0.
k =0 k =0

And hence we must have that

| Mn + 1 + Mn + 2 + . . . + Mn + m | < ε
⇒ Mn + 1 + Mn + 2 . . . + Mn + m < ε
(since each Mk is real and positive). Choosing
n > max( N, K ), we therefore must have
| c n +1 | + | c n +2 | + . . . + | c n + m | < ε
Applying the triangle identity, |cn + . . . + cn+m | ≤ |cn | + . . . +
|cn+m |, we find that
|cn+1 + cn+2 + . . . + cn+m | < ε,
⇒ |Sn+m − Sn | < ε, ∀n > max( N, K ), ∀m > 0,
where {Sn } = {∑nk ck } is the sequence of partial sums, which
we have just shown satisfies the Cauchy Criterion, and thus
must also converge.
54 complex analysis

Example 4.1: Comparing To the Geom. Series

Show that the series
∞
3 + 2i
∑ ( k + 1) k
(4.6)
k =0

converges.
Solution: We can compare the series above with the conver-
gent geometric series, with z = 1/2,
∞
1 1
∑ k
= 1
= 2. (4.7)
k =0 2 1− 2
√
Since |3 + 2i | = 13 < 4, we see that for k ≥ 3,

3 + 2i 4 1
k
< k
≤ k, (4.8)
( k + 1) ( k + 1) 2

hence, the series must converge.

Example 4.2: Absolute Convergence

Definition 4.2 (Absolutely Convergent). A series ∑∞ k =0 c k
is said to be absolutely convergent if the series ∑∞ |
k =0 kc | is
convergent.

Prove the following:

Theorem 4.3. Any absolutely convergent series must also be con-

vergent.

Proof: Let Mk = |ck |, and apply the comparison test.

The Ratio Test

Theorem 4.4 (Ratio Test). Suppose that the terms of the series
∑∞ k =0 ck have the property that the ratios | ck +1 /ck | approach a
limit L as k → ∞. Then the series converges if L < 1 and diverges
if L > 1.

You will prove this on a worksheet.

Example 4.3: Applying the Ratio Test

Show that the series ∑∞ k

k =0 4 /k! converges.
Solution:

c k +1 4k+1 k! 4
= k
= . (4.9)
ck ( k + 1) ! 4 k+1

This ratio approaches zero as k → ∞, therefore by the Ratio

Test, the series converges.
 series representations of complex functions 55

The kinds of sequences and series that often arise in complex

analysis are those where the terms are functions of a complex z.
Thus if we have a sequence of functions f 1 (z), f 2 (z), . . ., we must
consider the possibility that for some values of z the sequence
converges, while for others it diverges, as we saw for the geometric
series.
In applying this theory to holomorphic functions, we need a
somewhat stronger notion of convergence. Consider the sequence
of real functions

{ f n ( x )} = { x n } (4.10)

on the half open interval x ∈ [0, 1). Clearly for any given x in
the interval x n → 0 for sufficiently large n. We call this kind of
convergence pointwise convergence.
However, the curve y = x n for large n is a poor approximation to
the curve y = 0 since as x → 1 the difference between the functions
remains large. What we want instead is the property of uniform
convergence. Figure 4.1: The functional sequence
{ f n ( x )} = { x n } is pointwise conver-
Uniform Convergence gent, but not uniformly convergent on
the interval 0 ≤ x < 1.

Definition 4.3. The functional sequence { f n (z)}n is said to con-

verge uniformly to f (z) on the set U if for any real ε > 0, ∃ N ∈
N such that ∀n > N,

| f (z) − f n (z)| < ε, ∀z ∈ U. (4.11)

We say that the series ∑∞k =0 f k ( z ) converges uniformly to S ( z )

on U if the sequence of partial sums {Sn (z)} converges uniformly
to S(z) on U.

The essential feature of uniform convergence is that for a given

ε > 0, one must be able to find an integer N that is independent of
z ∈ U such that the error | f (z) − f n (z)| is less than ε for all possible
n > N. In contrast, for pointwise convergence, N can depend upon
z. Note, uniform convergence on U implies pointwise convergence
on U, i.e. uniform convergence is a stronger property.

4.2 Power Series

Power Series
Definition 4.4 (Power Series). A power series is defined as a se-
ries of the form
∞
∑ a n ( z − z o ) n = a0 + a1 ( z − z o ) + a2 ( z − z o )2 + . . . (4.12)
n =0

where an is the complex coefficient of the nth term, and zo is some

fixed point on the complex plane.
56 complex analysis

By using the comparison test, it is easy to prove the following:

Theorem 4.5. Let { an } be a sequence of complex numbers, and let r > 0
be a real number such that the series ∑∞ n
n=0 | an |r converges. Then the
series
∞
∑ an zn (4.13)
n =0

must converge absolutely and uniformly for |z| ≤ r.

Example 4.4: ∑ zn /n!

For any r > 0 show that the series

∑ zn /n! (4.14)

converges absolutely and uniformly for |z| ≤ r.

Solution: Let us define the series ∑n cn where cn = r n /n!,
then we have

c n +1 r n+1 n! r
= n
= . (4.15)
cn ( n + 1) ! r n+1

Take n ≥ 2r. Then the right hand side is ≤ 1/2 Hence for all
n sufficiently large, we have

cn+1 /cn ≤ 1/2 (4.16)

Therefore there exists some positive integer N such that

cn ≤ c N /2n− N = (c N 2 N )/2n (4.17)

for all n > N. We may therefore apply the comparison test

with the geoemetric series, with z = 1/2, to get absolute and
uniform convergence.

A given power series about a particular point can be shown to

converge within a particular radius r, called the radius of conver-
gence. Outside this radius, the series diverges.
Radius of Convergence
Theorem 4.6 (Radius of Convergence). Let ∑ an zn be a power
series. If it does not converge absolutely for all z, then there exists
a number r (that may be zero) called the radius of convergence,
such that the series converges absolutely for |z| < r, and does not
converge absolutely for |z| > r

Proof: Suppose that the series does not converge absolutely

for all z. Let r be the least upper bound of those numbers
s ≥ 0 such that ∑ | an |sn converges. Then ∑ | an ||z|n diverges
if |z| > r, and converges if |z| < r by the comparison test.

If a power series has non-zero radius of convergence, then it

is called a convergent power series. If D is a disc centred at the
 series representations of complex functions 57

origin and contained in the disc D (0, r ), where r is the radius of

convergence, then we say that the power series converges on D.

Example 4.5: Radius of Convergence of ∑ nzn

Find the radius of convergence of the series
∞
∑ nzn . (4.18)
n =0

Solution: We can apply the ratio test and see that this series
converges if limn→∞ |cn+1 /cn | < 1 and diverges if this limit
is larger than 1.

( n + 1 ) z n +1 n+1
lim n
= lim |z| = |z| (4.19)
n→∞ nz n → ∞ n

Thus ∑ nzn converges if |z| < 1 and diverges if |z| > 1.

4.3 Analytic Functions

It is convenient for us to have a name for the type of functions

that can be expressed as a convergent power series. We call such
functions analytic and with some clever use of Cauchy’s Integral
Formula, we will see that a function is analytic if and only if it is
holomorphic!

Analytic Functions
Definition 4.5 (Analytic Function). A function f (z) is called an-
alytic about a point zo if there exists a power series
∞
∑ an (z − zo )n (4.20)
n =0

and some r > 0 such that the series converges absolutely for |z −
zo | < r, and that for such z we have
∞
f (z) = ∑ an (z − zo )n . (4.21)
n =0

It should be obvious that polynomial functions can be expressed

as power series (just set an = 0 for n greater than the degree of the
polynomial), but a much wider class of functions turn out to be
expressible as convergent power series as well.
Suppose f is a function on the open set U. We say that f is
analytic on U if f is analytic at every point of U.
The next theorem, is easy to prove, but it gives us practice in a
way of finding power series expansions for a function at a point.
58 complex analysis

Power Series at zo = 0
Theorem 4.7. Let f (z) = ∑ an zn be a power series (centred at
z = 0) whose radius of convergence is r. Then f is analytic in the
open disc D (0, r ) ≡ {z : |z| < r }.

Proof: We have to show that f has a power series expansion

at an arbitrary point zo of the disc so that |zo | < r. Let s > 0 be
such that |zo | + s < r. We shall see that f can be represented by
a convergent power series at zo converging on a disc D (zo , s) with
radius s centred at zo . Figure 4.2: If f is a power series
centred at z = 0, with a radius of
We can write z as convergence r, we can choose a point
zo within D (0, r ) and a disc D (zo , s)
z = zo + (z − zo ) (4.22) where |z| + s < r such that f is analytic
anywhere in D (zo , s).
so that zn = (zo + (z − zo ))n , so that by the binomial theorem we
have
∞
!
n  
n n−k
f (z) = ∑ an ∑ z k
(z − zo ) . (4.23)
n =0 k =0
k o

If |z − zo | < s, then |zo | + |z − zo | < r, and hence the series

∞ ∞
" #
n  
n
∑ |an |(|zo | + |z − zo |) = ∑ |an | ∑ k |zo | |z − zo | (4.24)
n n−k k
n =0 n =0 k =0 † We are allowed to exchange the

order of summation for any finite

converges. Then we can interchange the order of the summations†
sum, by the normal commutativity
to get of addition. If an infinite sum is
absolutely convergent, then for any
∞ ∞
"   #
n n−k e > 0 we can always find an N large
f (z) = ∑ ∑ an zo (z − zo )k (4.25) enough so that the sum of of the the
k =0 n = k
k terms for n > N are smaller than e.
Thus we can exchange the sums for
| {z }
bn
the first (finite) N terms in the series
which converges absolutely also. Thus, the function f is analytic at to get the same value, whereas the
difference the infinite tails can be made
zo . arbitrarily small.
Generalising slightly, we can say that the above theorem shows
that if f has a power series expansion on a disc D (zo , r ), that is

f (z) = ∑ an (z − zo )n (4.26)

for |z − zo | < r, then f is analytic on this disc.

4.4 Differentiation of Power Series

Let D (0, r ) be a disc of radius r > 0. A function f on the disc

for which there exists a power series ∑ an zn having a radius of
convergence ≥ r such that
∞
f (z) = ∑ an zn (4.27)
n =0

for all z in the disc is said to admit a power series expansion on

this disc. We shall now see that such a function is complex differen-
tiable and thus is holomorphic on this domain, and the derivative is
 series representations of complex functions 59

indeed given by the "obvious" power series,

∞
∑ nan zn−1 = a1 + 2a2 z2 + 3a3 z3 + . . . . (4.28)
n =0

The Derivative of a Power Series

Theorem 4.8. If f (z) = ∑ an zn has radius of convergence r > 0,
then:

(i) The series ∑ nan zn−1 has the same radius of convergence,

(ii) The function f is holomorphic on D (0, r ) and its derivative is

equal to ∑ nan zn−1 .

Proof:

(i) For any w ∈ D (0, r ), except w = 0, choose z ∈ D (0, r ) \ {0}

such that |z| < |w|. Then by Example 4.5 the series
∞ n
z
∑ n
w
(4.29)
n =0

converges (since z/w < 1) and hence each term n|z/w|n must
be bounded by some M ∈ R, so that

n|z|n < M |w|n , ∀n ∈ N. (4.30)

Thus we see that

M
n | a n | z n −1 < | an ||w|n , ∀n ∈ N (4.31)
|z|
and the series must be convergent by the comparison test
since |M a wn is convergent for |w| < r. Therefore the radii
z| ∑ n
of convergence must be the same.

(ii) Let |z| < r, and choose some small enough real δ > 0 such
that |z| + δ < r. We consider complex numbers h such that
|h| < δ. We have

f (z + h) = ∑ an (z + h)n
= ∑ an (zn + nzn−1 h + h2 Pn (z, h)) (4.32)

where Pn (z, h) is a polynomial in z and h given by

n  
n k −2 n − k
Pn (z, h) = ∑ h z . (4.33)
k =2
k

We can then use the estimate

n  
n k −2 n − k
| Pn (z, h)| ≤ ∑ δ |z| = Pn (|z|, δ). (4.34)
k =2
k

Rearranging we see

f (z + h) − ∑ an zn − ∑ nan zn−1 h = ∑ h2 Pn (z, h) (4.35)

| {z }
f (z)
60 complex analysis

Since we showed above that both series on the left are abso-
lutely convergent for any z within the radius of convergence
of f , the series on the right hand side must also be absolutely
convergent.

We divide by h to obtain the finite difference

f (z + h) − f (z)
− ∑ nan zn−1 = h ∑ an Pn (z, h) (4.36)
h
≤ |h| Pn (|z|, δ) (4.37)

where the right hand side goes to zero as h → 0 (since δ is

independent of h). Therefore we must have that

f (z + h) − f (z)
f 0 (z) ≡ lim
h →0 h
= ∑ nan zn−1 . (4.38)

Thus we see that the following lemma must be true:

Analytic Functions are Holomorphic

Lemma 4.1. Any analytic function is also holomorphic within its
radius of convergence.

The kth Derivative of a Power Series

From Theorem 4.8 we see also see that the kth derivative of
f = ∑ an zn is given by the series

f (k) (z) = k!ak + hk (z) (4.39)

where hk (z) is a power series that has no constant term, so

that hk (z) → 0 as z → 0. Therefore we obtain the standard
expression for the coefficients of the power series in terms of
the derivatives

f ( n ) (0)
an = (4.40)
n!
If instead we are expanding about a point zo , then we find

f (n) ( zo )
an = . (4.41)
n!

To show that the definitions of analytic and holomorphic are

interchangeable, it only remains to be shown that the implication
also goes the other way, by circling back to Cauchy’s Integral
Formula.

4.5 Holomorphic Functions are Analytic

Let’s consider a function f (z) that is holomorphic on and inside a

circle CR : |z − zo | = R, centred on point zo .
 series representations of complex functions 61

Using Cauchy’s Integral Formula on CR we can swap the names

of symbols around replace z by w and zo by z, we find

1 f (w)
I
f (z) = dw (4.42)
2πi CR w−z

for any z inside CR .

Note that the denominator of the integrand above can be re-
placed by a geometric series in the interior of the CR
Theorem 4.9 (Integrals of Convergent
Sequences of Functions). Let { gn } be
1 1
= a sequence of continuous functions on
w−z w − zo − (z − zo ) a domain U, converging uniformly to a
function g. Then
 
1 1
=
w − zo 1 − ζ
Z Z
lim gn (z)dz = g(z)dz.
n→∞ γ γ
1  
= 1 + ζ + ζ2 + . . . (4.43)
w − zo
Proof: Uniform convergence on a
z−zo domain U means that for any real
where ζ ≡ w−zo , has |ζ | < 1 allowing the geometric series to
number e > 0 one can find an N ∈ N
converge since |w − zo | = R and z is in the (open) interior of CR , so such that | gn (z) − g(z)| < e for any
we must have |z − zo | < R. n ≥ N and z ∈ U, therefore,
Since f (z) is holomorphic in and on the circle, it is also bounded, Z Z Z
gn dz − gdz ≤ | gn − g|dz
and the series is uniformly convergent in this domain, thus we are γ γ γ

allowed† to integrate term by term, swapping the integral and the ≤ e`(γ),

infinite sum, where `(γ) is the length of the contour.

Hence, the integral of the sequence
∞ also uniformly converges to the
1 f (w) n
I
f (z) = ∑ 2πi CR ( w − zo )
ζ dw (4.44) integral of the limit.
n =0
∞ —
z − zo n
 
1 f (w)
I
= ∑ 2πi CR ( w − z o ) w − zo
dw (4.45)
Theorem 4.10 († Integrals of Conver-
n =0
∞ I  gent Series). If ∑ f n is a series of continu-
1 f (w)
= ∑ 2πi CR ( w − z o ) n +1
dw (z − zo )n (4.46) ous functions converging uniformly on U,
then
n =0 |
∞ ∞ Z
{z } Z
an
∑
γ n =0
f n (z)dz = ∑ f n (z)dz.
n =0 γ
This proves the following theorem:
Proof: Let { gk } = {∑kn=0 f n }. Then
Holomorphic Functions are Analytic by the Theorem 4.9 above, the integral
of this sequence of partial sums also
converges to the infinite sum of the
Theorem 4.11. In the circle CR where f (z) is holomorphic, f (z) integrals of each term.
can be expressed in terms of a convergent power series:
∞
f (z) = ∑ an (z − zo )n (4.47)
n =0

where
1 f (w)
I
an = dw. (4.48)
2πi CR ( w − z o ) n +1

Therefore combining the above with Theorem 4.8 we now

see that a function is analytic if and only if it is holomor-
phic. We may now use the terms interchangeably, as we
previously hinted.
62 complex analysis

Thus we see that any differentiable function can be expanded

as a power series — a very remarkable fact that is characteristic of
complex differentiability.

4.5.1 Holomorphic/Analytic Functions are Infinitely Differentiable

Using very similar arguments we can prove another somewhat
weird looking, but surprisingly useful theorem:

Theorem 4.12. Let γ be a directed curve in an open set U and let g(z)
be a continuous function for all z on γ. For any z ∈ U that is not on the
curve γ we can define the function

g(w)
Z
G (z) ≡ dw. (4.49)
γ w−z
Note that in contrast to Eq. (4.42), here
This function is analytic (and holomorphic) on U \ γ, and its nth derivative γ is not necessarily a closed loop and
g(z) is a completely arbitrary function
is given by
(not necessarily holomorphic) that is
only required to be continuous along
g(w)
Z
G (n) (z) = n! dw. (4.50) γ.
γ ( w − z ) n +1

Proof: Let zo ∈ U, and zo not on γ. Since U is an open set, we

can choose a circle CR of some radius R > 0 centred at zo such that
it does not overlap the curve γ (see Figure 4.3).
For z in the interior of CR we can expand the denominator of the
integrand using a convergent geometric series as before, and we
find
∞ Z
g(w)
G (z) = ∑ ( w − z o ) n +1
dw(z − zo )n (4.51)
n =0 γ

since we know |w − zo | > |z − zo | for all w on γ, and z in the

Figure 4.3: For a given point zo , we can
interior of CR . Thus the function G (z) is analytic (and holomorphic choose a radius R < min(|γ(t) − zo |)
by Theorem 4.8) since we can write it as a power series about any so that CR does not overlap with the
curve γ(t).
point zo in U \ γ,
∞
G (z) = ∑ an (z − zo )n (4.52)
n =0
g(w)
Z
an ≡ dw,
γ ( w − z o ) n +1

which converges within the radius R.

From Equation (4.41) and Theorem 4.8 we see that we can write
an as,

G (n) ( zo )
an = (4.53)
n!
Therefore we must have that
g(w)
Z
G (n) (zo ) = n!an = n! dw. (4.54)
γ ( w − z o ) n +1
 series representations of complex functions 63

Why do we care about functions and derivatives of the strange

integral form given by Equation (4.49)? Well, consider Cauchy’s
Integral Formula written in the form of Equation (4.42), for any
function f (z) holomorphic in the interior of CR .
1 f (w)
I
f (z) = dw.
2πi CR w−z
for any z interior to CR . It follows from Theorem 4.12 above, that
such a function must be have a derivative given by
1 f (w)
I
f 0 (z) = dw. (4.55)
2πi CR ( w − z )2
But it if we set G (z) = f 0 (z) and g(w) = f (w)/(w − z) (which
is continuous and well defined for w along CR , since z is not on
CR ), then we have that f 0 (z) is also expressible as a power series by
Equations (4.52), and hence is also analytic.
This can be carried on to show that any analytic (holomorphic)
function can be differentiated any number of times, and the nth
derivative will also be analytic (holomorphic),
Holomorphic Functions are Infinitely Differen-
tiable
Theorem 4.13. If f is holomorphic (analytic) in a domain U, then
all its derivatives f 0 , f 00 , . . . f (n) , . . . exist and are also holomor-
phic (analytic) in U. Hence, holomorphic functions are infinitely
differentiable.

We also get a general expression for the derivatives of f , in

terms of an integral!
Generalised Cauchy Integral Formula
Theorem 4.14 (The Generalised Cauchy Integral Formula). If
f is holomorphic inside and on the simple closed positively oriented
contour Γ and if z is any point inside Γ, then its nth derivative is
given by

n! f (w)
I
f (n) ( z ) = dw. (4.56)
2πi Γ ( w − z ) n +1

Example 4.6: Derivatives of e5z

Compute Γ e5z /z3 dz where Γ is the circle |z| = 1 traversed

R

once counterclockwise.
Solution: Observe that f (z) = e5z is holomorphic inside an
on Γ. Therefore, from Theorem 4.14 above, with z = 0 and
n = 2, we have

e5z 2πi 00
Z
dz = f (0) = 25πi. (4.57)
Γ z3 2!
64 complex analysis

4.6 Taylor and Maclaurin Series

Taylor and Maclaurin Series

Definition 4.6 (Taylor Series). If f is analytic at zo , then the se-
ries
f 00 (zo )
f (zo ) + f 0 (zo )(z − zo ) + ( z − z o )2 + . . .
2!
∞
f (k) ( zo )
= ∑ k!
(z − zo )k , (4.58)
k =0

is called the Taylor Series for f around zo . When zo = 0, it is

also known as the Maclaurin Series for f .

Theorem 4.15. If f is analytic in the open disc |z − zo | < R, then

the Taylor series converges to f (z) for all z in this disc. Futhermore, the
convergence of the series is uniform for any closed subdisc |z − zo | ≤ S <
R (which implies pointwise convergence in the open disc of radius R).

Proof: Let CT be a (positive) circular contour with S < T < R.

Then by Theorems 4.11 and 4.14, and we have for |z| ≤ S,
∞
f (z) = ∑ an (z − zo )n (4.59) Figure 4.4: Given S < R, to prove that
n =0 the Taylor Series converges uniformly
1 f (w) f (n) ( zo ) for any |z − zo | ≤ S, we make use of the
I
an = n + 1
dw = . (4.60) circular contour CT , where S < T < R.
2πi CT (w − zo ) n!
Thus the Taylor Series converges to f (z) converges for |z| < S. We
just need to show that this converges uniformly in this domain as
well.
Let us define the partial sum of the Taylor series
N
f N (z) ≡ ∑ an (z − zo )n (4.61)
n =0

From Equation (4.3) in our derivation of the Geometric Series, we

know that for |ζ | < 1, we have

1 ζ n +1
− (1 + ζ + ζ 2 + . . . + ζ n ) = (4.62)
1−ζ 1−ζ
Taking ζ = (z − zo )/(w − zo ) the error in the partial sum is
N
f (w)
I
f (z) − f N (z) = dw − ∑ an (z − zo )n (4.63)
CT w − z n =0

f ( w ) z − z o n +1
 
1
I
= dw (4.64)
2πi CT w − z w − zo
since |ζ | < 1 for |z − zo | < S < |w − zo | = T.
Since f is holomorphic and continuous on CT , there must exist a
maximum for f (w) in the integrand. Setting

M ≡ max | f (w)|,
w on CT
 series representations of complex functions 65

the error is then bounded by

  N +1
1 M2πT S
| f (z) − f N (z)| ≤ (4.65)
2πi (S − T ) T
The right hand side is independent of z, and since S/T < 1, given
any ε > 0, we can always choose an N large enough such that the
error is less than ε for any n > N. Therefore the convergence is
uniform for |z| ≤ S.
The above theorem immediately implies the following lemma:
Lemma 4.2. The Taylor Series will converge to f (z) everywhere inside
the largest open disc, centred at zo , over which f is analytic.

Example 4.7: Tinker, Something, Soldier, Sailor

Compute and state the convergence properties of the Taylor
series for (a) Log (z) around zo = 1, (b) 1/(1 − z) around
zo = 0 and (c) ez around zo = 0.
Solution:

(a) The derivatives of Log z in general are

dk Log z (a)
= (−1)k+1 (k − 1)!z−k for k = 1, 2, . . . ,
dzk
Branch cut for
so that evaulating at z = 1 gives us
∞
(−1)k+1 (z − 1)k
Log z = ∑ k
. (4.66)
k =1
(b)
This converges for |z − 1| < 1, since the largest open pole for
disk centred at +1 we can have before we run into the
branch point at z = 0 has a radius of 1.
1
(b) The derivatives of 1− z are given by

dk
(1 − z)−1 = k!(1 − z)− j−1 .
dzk
Evaluating these at z = 0 gives us the Taylor series,
∞
1 2!z2
= 1+z+ + . . . = ∑ zk (4.67)
1−z 2! k =0

(which of course is just the geometric series), which

converges for |z| < 1, since there is a pole where the
function is no longer holomorphic at z = 1.

(c) Since the derivative of ez is just ez , we can evaluate

this at z = 0 to get

z2 z3
ez = 1 + z + + +... (4.68)
2! 3!
as expected. This converges for all z, since ez is ana-
lytic on the entire complex plane.
66 complex analysis

4.7 Laurent Series

We are now ready to introduce a very important (for Complex

Analysis anyways) generalisation of the power series, the Laurent
Series.
Laurent Series
Obviously, it is easily to generalise the
Definition 4.7 (Laurent Series). A Laurent Series is a series following for a Laurent Series centred
that can be written as around a point zo , where
∞
∞
f (z) = ∑ an (z − zo )n .
f (z) = ∑ an z n
. (4.69) n=−∞
−∞

Let U be a set of complex numbers. We say that the Laurent

series converges absolutely on U if the two series,

f + (z) ≡ ∑ an zn and f − (z) ≡ ∑ an zn

n ≥0 n <0

converge absolutely on U. If this is the case then f (z) can be re- Similarly, f (z) is said to converge
garded as the sum, uniformly on U if f + (z) and f − (z)
both converge uniformly on U.
f ( z ) = f + ( z ) + f − ( z ). (4.70)

Laurent Series of a Function

Let r, R ∈ R be positive numbers with 0 ≤ r < R. Consider
the closed † annular domain A consisting of all complex † Itis straightforward to instead prove
numbers z such that r ≤ |z| ≤ R, which is bounded by the a related theorem on the open annulus
positively oriented contours CR and Cr . following a similar procedure to the
proof of Theorem 4.15 (on the conver-
Theorem 4.16. Let A be the closed annulus above, and let f be a gence properties of Taylor expansions)
by introducing contours that sit in
holomorphic function everywhere in A. Let some s and S define the between the boundary and an interior
radii of the closed sub-annulus such that r < s ≤ |z| ≤ S < R. closed annulus†† . This would show
uniform convergence of the Laurent
Then f has a Laurent expansion, Series for any closed annular domain
∞ interior to the open annulus, but only
f (z) = ∑ an zn , (4.71) pointwise convergence in the whole
open annulus.
n=−∞

which converges absolutely and uniformly on s ≤ |z| ≤ S. The co-

efficients an are obtained by the formula: †† Under no circumstances should

 you refer to a closed annulus as a

 1 H f (w) dw, if n ≥ 0, "clenched annulus".
n +1
an = 2πi HCR fw(w) (4.72)
 1 dw, if n < 0.
2πi Cr wn+1

Proof: For z such that s ≤ |z| ≤ S, we can use Cauchy’s Integral

Formula over the closed contour Γ shown in Figure 4.5,
1 f (w)
I
f (z) = dw
2πi Γ w − z
1 f (w) 1 f (w)
Z Z
= dw − dw (4.73)
2πi CR w − z 2πi Cr w − z
where we note that the two radial paths (chosen such that they do
not intersect z) cancel out, and Cr is traversed clockwise, acquiring
a minus sign.
 series representations of complex functions 67

The first integral is handled in the same way as the proof of

Theorem 4.11, by expanding using the geometric series, and inter-
changing the integral and the sum,
∞  
1 f (w) f (w)
Z
dw = ∑ n +1
dw zn . (4.74)
2πi CR w − z n =0 w

The second integral can be expanded in a similar fashion, by

writing
 w
w − z = −z 1 − . (4.75)
z
Then, since |w/z| < 1 for w on Cr , the geometric series converges,
allowing us to expand,
∞
wk
 
1 1 1 w  w 2
= 1+ + + . . . = ∑ k +1 . (4.76)
z 1 − w/z z z z k =0 z

So that the second integral becomes

∞
!
1 f (w) 1 wk
Z Z

2πi Cr k∑
− dw = − f (w) dw, (4.77)
2πi Cr w−z =0 z k +1 Figure 4.5: Above: The closed annulus
A centred at z = 0 and bounded by
−∞  
1 f (w)
Z
the circular contours Cr and CR . For
= ∑ n +1
dw zn , (4.78) any z such that r < s ≤ |z| ≤< R,
n=−1
2πi Cr w
the Laurent Series for a function
holomorphic on A converges. Below:
where k = −(n + 1) and we can interchange the sum and the in- For f holomorphic in the open interior
tegral because the series converges uniformly. This completes the of annulus A, i.e. {z : r < |z| < R},
proof of Theorem 4.16. we can use the contour Γ to prove
Theorem 4.16.
Example 4.8: Laurent Series of 1/[z(z − 1)]
Find the the Laurent Series (centred at z = 0) for the func-
tion
1
f (z) = , for 0 < |z| < 1. (4.79)
z ( z − 1)
Solution: We can write f using partial fractions
1 1
f (z) = − . (4.80)
z−1 z
Then for one term we get the geometric series,
1 1
=− = −1 − z − z2 − . . . (4.81)
z−1 1−z
1
f ( z ) = − − 1 − z − z2 + . . . . (4.82)
z
What happens if we want the Laurent series for |z| > 1?
Then we can write
   
1 1 1 1 1 1
= = 1+ + 2 +... , (4.83)
z−1 z 1 − 1/z z z z
so that we find the Laurent Series,
1 1 1
f (z) = 2
+ 3 + 4 +.... (4.84)
z z z
68 complex analysis

Example 4.9: Different Regions, Different Series

For the function
1
(4.85)
(z − 1)(z − 2)
find the Laurent series (centred at zo = 0) in
(c)
(a) the region |z| < 1,
(b)
(b) the region 1 < |z| < 2,

(c) the region |z| > 2. (a)

Solution: Using partial fractions we can write

1 1 1
= − (4.86)
(z − 1)(z − 2) z−2 z−1
Now in each region we have to proceed slightly differently
to keep the series convergent:
Figure 4.6: The different domains for
(a) For |z| < 1 we have Example 4.9. The function 1/[(z −
1 1 1 1 ∞  z k 1)(z − 2)] has poles at z = 1 and z = 2.

2 k∑
=− z =− (4.87)
z−2 21− 2 =0
2
∞
1
= − ∑ zk (4.88)
z−1 k =0

This gives us the series expansion

∞  
1 1
(z − 1)(z − 2) k∑
= 1 − k +1 z k (4.89)
=0 2
which converges in the region |z| < 1.

(b) For 1 < |z| < 2, the first expansion in part (a) is valid,
but we need to replace the second.
1 1 1 1 ∞ 1
z k∑
= 1
= k
(4.90)
z z1− z =0 z
So that the series can be written as
∞ ∞
1 zk 1
= − ∑ k +1 − ∑ k +1 (4.91)
(z − 1)(z − 2) k =0 2 k =0 z
1 1 1 z
= ...− − − − −... (4.92)
z2 z 2 4
for 1 < |z| < 2.

(c) For |z| > 2 we can use Equation (4.90) as well as

∞
1 1 1 2k
z−2
=
z1− 2
= ∑ z k +1
, (4.93)
z k =0

which gives us,

∞
1 2k − 1 1 3 7
= ∑ k +1 = 2 + 3 + 4 + . . . .
(z − 1)(z − 2) k=0 z z z z
(4.94)
 series representations of complex functions 69

4.8 Singularities and Zeros Revisited

With Laurent Series properly defined and well understood, we are

now ready to revisit our discussion of singularities that were given
by Definitions 2.6 - 2.8. In terms of Laurent Series, these definitions
now become:
Isolated Singularities
Definition 4.8 (Singularities in terms of Laurent Series coef-
ficients). Let f have an isolated singularity at zo , and let us have
a Laurent Series expansion of f convergent in 0 < |z − zo | < R
given by
∞
f (z) = ∑ an (z − zo )n . (4.95)
n=−∞

Then, Cultural Sidenote: If f is holomorphic

on an open set U except at a discrete
(i) If an = 0 for all n < 0 we say that zo is a removable sin- set of points which are poles, then f is
sometimes called meromorphic on U.
gularity of f ; This derives from the Greek "meros"
meaning part, and "morphe" meaning
(ii) If a−m = 0 for some positive integer m, but an = 0 for all shape or form and should not to be
n < −m, we say that zo is a pole of order m for f ; confused with the term Mer-Morphic.

(iii) If an 6= 0 for an infinite number of negative values of n we

say that zo is an essential singularity of f .

We can also use the Laurent Series expansion to classify the

zeros of an analytic function f , i.e., point zo where f is analytic and
f (zo ) = 0.
Zeros of order m
Definition 4.9. A point zo is called a zero of order m for the
function f if f is analytic at zo , and f and its first m − 1 deriva-
tives vanish at zo but f (m) (zo ) 6= 0.

Using the Laurent series expansion, it is then easy to show

Theorem 4.17. Let f be analytic on zo . Then f has a zero of order

m at zo if and only if f can be written as,

f (z) = (z − zo )m g(z) (4.96)

where g is analytic at zo and g(zo ) 6= 0.

Just as a pole of order 1 is called a simple pole, a zero of order 1

is often called a simple zero. For instance the zeros of the function
sin z, which occur at integer multiples of π are all simple zeros,
since the first derivative at these points is nonzero.
5
The Calculus of Residues

We have already seen how the theory of contour integration pro-

vides insight into the properties of holomorphic functions. We have
developed all the tools we need to compute integrals of analytic
functions in terms of their power series expansions, and use these
to evaluate certain real integrals.

5.1 Cauchy’s Residue Theorem

Let us consider the problem of evaluating the integral

I
f (z)dz, (5.1)
Γ

where Γ is a simple closed positively oriented contour and f (z) is

holomorphic on and inside Γ except for a single isolated singularity
at zo lying interior to Γ.
As we showed in the previous chapter, the function f (z) has a
Laurent Series expansion,
Figure 5.1: If f (z) is holomorphic
∞ inside and on Γ except at a simple pole
f (z) = ∑ an (z − zo )n (5.2) at zo , we can deform the contour Γ to
n=−∞ the circle C, where the Laurent Series
for f (z) centred at zo converges in the
punctured disc bounded by C.
converging in some punctured neighbourhood of zo , such as the
circle C (see Figure 5.1). Via the Deformation Invariance Theorem
(Theorem 3.5), the integral over Γ can be deformed to integration
over C without modifying the result of the integral,
I I
f (z)dz = f (z)dz. (5.3)
Γ C

Since the Laurent Series (uniformly) converges everywhere on the

contour C, the integral on the right hand side can be computed by
term-wise integration of the Laurent Series along C. For all n 6= −1
the integral is zero [see Example 3.1], and for n = −1 we obtain the
value of 2πia−1 . The coefficient a−1 of this Laurent expansion about
the pole zo is the key to the calculus of residues.
 the calculus of residues 71

Cauchy’s Residue Theorem

Definition 5.1 (Residues of a Pole). If f has an isolated singu-
larity at the point zo , then the coefficient a−1 of 1/(z − zo ) of the
Laurent series expansion for f around zo is called the residue of
f at zo and is denoted by

Res( f ; zo ), Reszo f , or Res(zo ),

where the latter can be used when there is no ambiguity as to

which function we mean.

Theorem 5.1 (Cauchy’s Residue Theorem). Let U be an open

set and Γ a closed positively oriented contour in U that is tra-
versed once. Let f be analytic on U except at a finite number of
points z1 , . . . , zm . Then,
I m

Γ
f (z)dz = 2πi ∑ Res( f ; zk ). (5.4)
k =1

Proof: We can deform the contour Γ to the contour consisting

of circles Ck surrounding each of the m poles, which are well sepa-
rated, since the singularities are isolated, and connected by linear
paths that are traversed once in each direction. Since the integrals
along the straight paths cancel out, the contour integral becomes,

I m I

Γ
f (z)dz = ∑ f (z)dz (5.5)
k =1 Ck
m
= 2πi ∑ Res(zk ), (5.6)
k =1

by applying the results of Example 3.1 for each circular contour.

Figure 5.2: Contours for the proof of

Cauchy’s Residue Theorem. Such a
finite sum of closed contours Cn is
often called a chain (sometimes shown
with the connecting cancelling paths
omitted).
72 complex analysis

Example 5.1: Residue of ze3/z

Find the residue at z = 0 of the function f (z) = ze3 /z and
compute
I
ze3/z dz, (5.7)
|z|=4

where the contour is traversed once in the counterclockwise

direction.
Solution: Since ew has a Taylor expansion,
∞
wk
ew = ∑ k!
, (5.8)
k =0

for all w ∈ C. The Laurent expansion for f (z) around z = 0

is then given by,
Figure 5.3: Contour and pole for the
∞  k function f (z) = ze3/z for Example
1 3 32 33
ze3/z =z∑ = z+3+ + +... (5.9) 5.1. Note that the pole at z = 0 is an
k =0
k! z 2!z 3!z2 essential singularity.

Hence we find,

32 9
Res(0) = a−1 = = . (5.10)
2! 2
Since z = 0 is the only singularity inside |z| = 4, we have

9
I
ze3/z dz = 2πi · = 9πi. (5.11)
|z|=4 2

5.2 Calculating Residues

Cauchy’s Residue theorem allows us to calculate the contour inte-

grals around isolated singularities in terms of their residues, which
are given by the value of the Laurent series coefficient a−1 . In par-
ticular, for the three different types of isolated singularities, we
see:

• If f has a removable singularity at zo , then all the coefficients of

the negative powers of (z − zo ) in the Laurent expansion are zero,
and so, the residue of a function at a removable singularity is
zero.

• If f has an essential singularity at zo (e.g. in Example 5.1) to

determine the residue we must determine the Laurent series
expansion for f about zo , and the coefficient a−1 .

• If f has a pole at zo , it turns out that we can find a simple for-

mula to evaluate the residue.

If the singularity is a simple pole (i.e. a pole of order 1) then for z

 the calculus of residues 73

near zo where the Laurent Series is convergent, we have

a −1
f (z) = + a0 + a1 (z − zo ) + a2 (z − zo )2 + . . . (5.12)
z − zo
⇒ ( z − z o ) f ( z ) = a −1 + ( z − z o ) [ a 0 + a 1 ( z − z o ) + . . . ] , (5.13)

giving us the following theorem:

Residue of a Simple Pole

Theorem 5.2 (Residue at a Simple Pole). Let f be a function
that has an isolated simple pole at zo . Then the residue of f at zo
can be given by,

Res( f ; zo ) = lim (z − zo ) f (z). (5.14)

z→zo

Example 5.2: Simple Poles

Find the residues at the simple poles of

ez
f (z) = . (5.15)
z ( z + 1)

Solution: f (z) has simple poles at z = 0 and z = −1. There-

fore we have
ez
Res( f ; 0) = lim z f (z) = lim = 1, (5.16)
z →0 z →0 z + 1
ez
Res( f ; −1) = lim (z + 1) f (z) = lim = − e −1 . (5.17)
z→−1 z→−1 z

Example 5.3: Denominators with Simple Zeros

Prove the following theorem:

Theorem 5.3. Let f (z) = P(z)/Q(z), where the functions P(z)

and Q(z) are both analytic at zo , and Q(z) has a simple zero at
zo , while P(zo ) 6= 0. Then, the residue of f at zo is given by,

P(zo )
Res( f ; zo ) = . (5.18)
Q0 (zo )

Proof: Obviously, f has a simple pole at zo , so we can apply

Theorem 5.2. Using the fact that Q(zo ) = 0, we see that

P(z)
Res( f ; zo ) = lim (z − zo ) ,
z→zo Q(z)
P(z)
= lim h i,
z→zo Q(z)− Q(zo )
z−zo
P(z)
= .
Q0 (z)
74 complex analysis

To obtain the general formula for the residue at a pole of order

m we need some method of picking out the coefficient a−1 from the
Laurent expansion.
Residue of an mth Order Pole
Theorem 5.4 (Residue of an mth Order Pole). If f has a pole of
order m at zo , then

1 d m −1
Res( f ; zo ) = lim [(z − zo )m f (z)] . (5.19)
z→zo (m − 1)! dzm−1

Proof: The Laurent expansion for f around zo is

a−m a
f (z) = + . . . + −1 + a 0 + a 1 ( z − z o ) + . . . ,
(z − zo )m z − zo
( z − z o ) m f ( z ) = a − m + . . . + a −1 ( z − z o ) m −1
+ a 0 ( z − z o ) m + a 1 ( z − z o ) m +1 + . . . .
We can differentiate the above m − 1 times to find
d m −1
[(z − zo )m f (z)] = (m − 1)! a−1 + m!a0 (z − zo )
dzm−1
( m + 1) !
+ a1 ( z − z o )2 + . . . ,
2
d m −1
⇒ lim m−1 [(z − zo )m f (z)] = (m − 1)!a−1 .
z→zo dz

Example 5.4: Higher Order Poles

Compute the residues at the singularities of
cos z
f (z) = . (5.20)
z2 ( z − π )3

Solution: f (z) has a pole of order 2 at z = 0 and a pole of

order 3 at z = π. Applying Theorem 5.4 we find

1 d 2
Res( f ; 0) = lim [z f (z)]
1! dz
z →0
 
d cos z
= lim
z→0 dz ( z − π )3
−(z − π ) sin z − 3 cos z
 
= lim
z →0 ( z − π )4
3
= − 4.
π
1 d2
Res( f ; π ) = lim [(z − π )3 f (z)]
z→π 2! dz2
1 d2 h cos z i
= lim
z→π 2 dz2 z2
1 (6 − z2 ) cos z + 4z sin z
 
= lim
z→π 2 z4
6 − π2
=− .
2π 4
 the calculus of residues 75

Exercise 5.1: Residue Practice

Compute the residue at each singularity of the following
functions

cos z
(a) cot z; (d) z2
;
e3z e1/z
(b) z −2 ; (e) z −2 ;
z +1 z −1
(c) z2 −3z+2
; (f) sin z .

Example 5.5: Missing a Pole

Evaluate
1 − 2z
I
dz. (5.21)
|z|=2 z(z − 1)(z − 3)

where the contour is traversed once in the counterclockwise

direction.
Solution: The integrand f (z) = (1 − 2z)/[z(z − 1)(z − 3)] has
simple poles at z = 0, z = 1, and z = 3. However, only the
first two of these points lie inside the contour. Thus by the
residue theorem
I
f (z)dz = 2πi [Res(0) + Res(1)]. (5.22)
|z|=2

Evaluating the residues we find,

1 − 2z 1
Res(0) = lim z f (z) = lim = , (5.23) Figure 5.4: Contour and poles for
z →0 z→0 ( z − 1)( z − 3) 3
f (z) = (1 − 2z)/[z(z − 1)(z − 3)] in
1 − 2z 1 Example 5.5. Note that the contour
Res(1) = lim (z − 1) f (z) = lim = . (5.24) does not enclose the pole at z = 3.
z →1 z →1 z ( z − 3 ) 2

This gives us
 
1 1 5πi
I
f (z)dz = 2πi + = . (5.25)
|z|=2 3 2 3

Exercise 5.2: Contour Integral Practice

Evaluate each of the following integrals by means of
Cauchy’s Residue Theorem. You may assume that all con-
tours are positively oriented.

sin z 1
H H
(a) |z|=5 z2 −4 dz (e) |z|=1 z2 sin z dz
ez 3z−2
H H
(b) |z|=3 z(z−2) dz (f) |z|=3 z4 +1 dz
H
(c) tan z dz 1
H
|z|=2π (g) |z|=8 z2 +z+1 dz
eiz
H
1/z sin(1/z ) dz
H
(d) |z|=3 z2 (z−2)(z+5i ) dz (h) |z|=1 e
6
Applications of Cauchy’s Residue Theorem

Having established Cauchy’s residue theorem and the basics of the

calculus of residues, we now turn to their applications in evaluating
certain useful real integrals.

6.1 Trigonometric Integrals Over [0, 2π ]

We would like to use Cauchy’s residue theorem to evaluate real

integrals of the form Figure 6.1: Q: Why did the Mathemati-
cian name his dog Cauchy?
Z 2π A: Because his favourite long-dead
Q(cos θ, sin θ )dθ (6.1) uncle was named Cauchy and the dog
0 was the only family he had left.
where Q(cos θ, sin θ ) is a rational function (with real coefficients) of
cos θ and sin θ which is finite over [0, 2π ].
We want to recast this integral into a parameterised form of a
contour integral
Z
F (z)dz (6.2)
C1

where C1 is the positively oriented unit circle |z| = 1. We start by

parameterising C1 by

z = eiθ , (0 ≤ θ ≤ 2π ). (6.3)

For such z we have 1/z = e−θ . We can use these substitutions in the
expressions for the trigonometric functions

eiθ + e−iθ
 
1 1
cos θ = = z+ ; (6.4)
2 2 z
eiθ − e−iθ
 
1 1
sin θ = = z− . (6.5)
2i 2i z
When integrating along C1 , we also have

dz
dz = ieiθ dθ ⇒ dθ = . (6.6)
iz
With these substitutions we can rewrite the trigonometric integral
as
Z 2π I
Q(cos θ, sin θ )dθ = F (z)dz (6.7)
0 C1
 applications of cauchy’s residue theorem 77

where the new integrand F is

    
1 1 1 1 1
F (z) ≡ Q z+ , z− · . (6.8)
2 z 2i z iz

By Cauchy’s residue theorem our trigonometric integral must equal

2πi times the sum of the residues of those poles of F which lie
inside C1 .

Example 6.1: Trig Integral

Evaluate

sin2 θ
Z 2π
I= dθ. (6.9)
0 5 + 4 cos θ
Solution: We note that the denominator, 5 + 4 cos θ
is never zero, so the integrand is finite over
[0, 2π ]. We can then perform the substitutions
above for cos θ, sin θ, and dθ to find,

h i2
1
I z − 1z
2i dz
I= h  i . (6.10)
C1 1
5+4 2 z+ z 1 iz

Some algebra reduces this to

1 ( z2 − 1)2
I
I=− dz. (6.11)
4i C1 z2 (2z2 + 5z + 2)

The integrand,

1 ( z2 − 1)2
F (z) = − (6.12)
4i z2 (2z2 + 5z + 2)

has simple poles at z = −1/2 and z = −2, and a pole of order 2 at z = 0. However, only −1/2 and
0 lie inside the unit circle, so we have
   
1
I = 2πi Res F; − + Res( F; 0) . (6.13)
2

Using the techniques of the preceding section, we have

 
1
Res F, − = lim (z + 1/2) F (z)
2 z→− 12

1 ( z2 − 1)2
=− lim
4i z→− 12 2z2 (z + 2)
3
=− ;
16i
78 complex analysis

1 d 2
Res ( F, 0) = lim [z F (z)]
z→0 1! dz
( z2 − 1)2
 
1 d
= − lim
4i z→0 dz 2z2 + 5z + 2
1 (2z2 + 5z + 2) · 2(z2 − 1)2z − (z2 − 1)2 (4z + 5)
=−
4i (2z2 + 5z + 2)2 z =0
5
= .
16i
Thus, the contour integral is given by

−3
 
5 π
I = 2πi + = . (6.14)
16i 16 4

Example 6.2: Trig Integrals Cont’d

Evaluate
dθ
Z π
I= . (6.15)
0 2 − cos θ
Solution: The trick here is that the integral is taken over
[0, π ] instead of [0, 2π ]. However, it is easy to see that since
cos θ = cos(2π − θ ),
Z 2π
dθ dθ
Z π
= ,
0 2 − cos θ π 2 − cos θ
and, therefore,
Z 2π
dθ
= 2I.
0 2 − cos θ Figure 6.2: Contour and poles for
Example 6.2.
Substituting for cos θ and dθ, we have
1 dz
I
2I =  ·
C1 2 − 12 z + 1z iz

2 dz
I
=− .
i C1 z2 − 4z + 1
By the quadratic formula the zeros of the denominator are
√ √
z− ≡ 2 − 3, z+ ≡ 2 + 3,

so that the integrand

2i 2i
F (z) ≡ =
z2 − 4z + 1 (z − z− )(z − z+ )
has simple poles at these points. But only z− lies inside C1 ,
which has residue given by

Res( F; z− ) = lim (z − z− ) F (z)

z→z−
2i
= lim
z→z− z − z+
2i i
= = −√ .
z− − z+ 3
 applications of cauchy’s residue theorem 79

This gives us

i 2π
2I = −2πi √ = √
3 3
π
I= √ .
3

Exercise 6.1: Practice on Trig Integrals on the Unit Circle

Using Cauchy’s Residue Theorem, verify each of the following:
R 2π Rπ dθ
√
dθ 2π (d) = π 2;
(a) 0 2+sin θ = √ ; −π 1+sin2 θ
3
R 2π dθ
(e) = √ 2π , for a2 < 1;
Rπ 8dθ 8π 0 1+ a cos θ 1− a2
(b) 0 5+2 cos θ = √ ;
21
R 2π sin2 θ 2π
√
(f) 0 a+b cos θ dθ = b2
(a − a2 − b2 ),
√
dθ 3π 5
Rπ
(c) 0 (3+2 cos θ )2 = 25 ; for a > |b| > 0.

6.2 Improper Integrals and the Cauchy Principal Value

If f ( x ) is a function continuous on the nonnegative real axis a ≤

x < ∞, then the improper integral of f over [ a, ∞) is defined by
Z ∞ Z b
f ( x )dx ≡ lim f ( x )dx, (6.16)
a b→∞ a

provided this limit exists.

Example 6.3: An Improper Integral

Find
Z ∞
e−2x dx. (6.17)
0

Solution:
Z ∞ Z b
e−2x dx = lim e−2x dx
0 b→∞ 0
b
−e−2x
= lim
b→∞ 2 0
" #
−e − 2b 1 1
= lim + = .
b→∞ 2 2 2

Similarly when f ( x ) is continuous on (−∞, a], we set

Z a Z a
f ( x )dx ≡ lim f ( x )dx. (6.18)
−∞ c→−∞ c

If it turns out that both of the above limits exist for a function f
continuous on the whole real line, then f is said to be integrable
80 complex analysis

over (−∞, ∞), and we can define

Z ∞ Z 0 Z b
f ( x )dx ≡ lim f ( x )dx + lim f ( x )dx (6.19)
−∞ c→−∞ c b→∞ 0
Z 0 Z ∞
= f ( x )dx + f ( x )dx. (6.20)
−∞ 0

Thus the improper integral over (−∞, ∞) can be computed by

taking a single limit,
Z ∞ Z R
f ( x )dx = lim f ( x )dx. (6.21)
−∞ R→∞ − R

However, this last limit may exist even for certain non-integrable
functions f . Consider f ( x ) = x. This function is non integrable over
(−∞, ∞) since the limit

b2
Z b
lim xdx = lim (6.22)
b→∞ 0 b→∞ 2
is not finite. However, we if we instead take the dual limit
R
x2
Z R
lim xdx = lim = lim 0 = 0. (6.23)
R→∞ − R R→∞ 2 −R R→∞

For this reason we introduce the following definition:

Cauchy Principal Value of an Improper Integral
This is not to be confused with the
Definition 6.1 (The Cauchy Principal Value). Given any func- principal value we discussed in
tion f continuous on (−∞, ∞), the limit Chapter 1, that chooses a branch of a
multi-valued complex function.
Z R
lim f ( x )dx (6.24)
R→∞ − R Several different notations are used
for the Cauchy Principal Value of the
(if it exists) is called the Cauchy Principal Value of the integral integral of a function f including
of f over (−∞, ∞). We typically write
R
• p.v. f ( x )dx;
R
• PV f ( x )dx;
Z ∞ Z R R
P f ( x )dx ≡ lim f ( x )dx. (6.25) • V.P. f ( x )dx;
−∞ R→∞ − R R∗
• L f (z)dz.

R∞
In our previous example, P −∞ f ( x )dx = 0. Whenever the
improper integral exists, it must equal its principal value. Contour
integration and the theory of residues can be used to compute
principal value integrals for certain functions.
Cauchy Principal Value and Semicircular Con-
tours
Let f (z) be a holomorphic except at a finite number of poles.
Suppose that f ( x ) is the restriction to the real line of the
function f , and is a continuous function of real variable x
that is well defined (finite) for the entire real line. Let γR
be the closed positively oriented path, denoted in Figure
6.3, consisting of the real interval between [− R, R] and the
semi-circle of radius R, SR , in the upper half plane.

Figure 6.3: The contour consisting of

the real line between [− R, R], and the
semi-circle SR .
 applications of cauchy’s residue theorem 81

Theorem 6.1 (Evaluating Improper Integrals). If there exists

real numbers B > 0 and α > 0 such that for all |z| sufficiently
large, we have

| f (z)| ≤ B/|z|1+α (6.26)

Then
Z
lim f (z)dz = 0, (6.27)
R→∞ SR

and
Z ∞
P f ( x )dx = 2πi ∑ residues of f in the upper half plane.
−∞
(6.28)

Proof: The integral over SR has f bounded by B/R1+α by as-

sumption, and thus is itself bounded by ( B/R1+α ) × πR. Since
πB/Rα tends to 0 as R → ∞ for α > 0, the integral over SR must be
zero. Then we must have
Z ∞ Z
P f ( x )dx = lim f (z)dz (6.29)
−∞ R→∞ γR

so that application of Cauchy’s residue theorem gives us the de-

sired result.
Example 6.4: Cauchy Principal Value
Evaluate
Z ∞
dx
I=P . (6.30)
−∞ x4 + 4
Solution: We can define the integral
Z R
dx
IR ≡ (6.31)
−R x4 + 4
which can be interpreted as a contour integral of an analytic
function,
dz
Z
IR = (6.32)
γR z4 +4
where γR . This complex function is holomorphic except at 4
different poles given by Figure 6.4: Contour and Poles for
Example 6.4.
z4 = −4 ⇒ zk = 21/2 eiπk/4 , k = 1, −1, 3, −3 (6.33)
The poles within the contour γR as R → ∞ are those in the
√
upper half plane, which are the poles z1 = 2eiπ/4 = 1 + i
√
and z3 = 2ei3π/4 = −1 + i. Meanwhile the poles z−1 = 1 − i
and z−3 = −1 − i are in the exterior of the contour γR .
We can ignore the integral along the semicircular path SR as
R → ∞ as long as the integrand goes to zero faster than 1/R.
The inequality
1
≤ B/R4 (6.34)
z4 +4
82 complex analysis

is indeed satisfied for some constant B when |z| = R, hence

the theorem above applies, and we have
Z ∞
dx
P = 2πi [Res(z1 ) + Res(z3 )]
−∞ x4 +4
z − z1 z − z1
 
= 2πi lim 4 + lim 4
z → z1 z + 4 z → z3 z + 4

1
= 2πi
(z1 − z−1 )(z1 − z3 )(z1 − z−3 )

1
+
(z3 − z1 )(z3 − z−1 )(z3 − z−3 )
 
1 1
= 2πi +
4i (2 + 2i ) −4i (−2 + 2i )
−1 − i 1 − i
 
= 2πi +
16 16
π
= .
4

Exercise 6.2: Practice on Integrals over (−∞, ∞)

Using the calculus of residues verify each of the following,
R∞ dx
R∞ dx
(a) P −∞ x2 +2x +2 = π; (d) P −∞ ( x2 +1)( x2 +4) = π
6;
R∞ x2
R∞ x
(b) P −∞ ( x2 +9)2 dx = π
6; (e) P −∞ ( x2 +4x +13)2 dx = − 27
π
;
R∞ x 2 +1 π
R∞ x2 π
(c) 0 x 4 +1
dx = √ ; (f) 0 ( x2 +1)( x2 +4)
dx = 6.
2

6.3 Improper Integrals Involving Trig Functions

In this section we develop methods to evaluate integrals of the

general forms
Z ∞ Z ∞
P( x ) P( x )
P cos mxdx, or P sin mxdx, (6.35)
−∞ Q( x ) −∞ Q( x )

where m is real and P( x )/Q( x ) is a rational function continuous on

the real line. We can often use the semicircular contour approach
from above, but some modifications are necessary.
Example 6.5: Real Trig, Complex Problems
Compute
Z ∞
cos 3x
I=P dx (6.36)
−∞ x2 + 4
Solution: In utilising semicircular contours our first inclina-
tion is to deal with the complex function

cos 3z
f (z) = . (6.37)
z2 + 4
 applications of cauchy’s residue theorem 83

However, with this choice for f (z) the modulus does not go
to zero in either the upper or lower half-plane. We see that
when z = ±ρi we have

cos 3z e−3ρ + e3ρ

lim = lim → ∞. (6.38)
ρ→∞ z2 + 4 ρ → ∞ 2| − ρ2 + 4|

To get around this problem, we notice that since cos 3x is the

real part of e3ix , we have
Z ∞
e3ix
I = Re( Io ), where Io ≡ dx. (6.39)
−∞ x2 +4
To evaluate this new integral we can deal with the function

e3iz
f (z) ≡ (6.40)
z2 +4
This function has singularities at z = ±2i. We can evaluate
the absolute value of f (z)

|e3ix e−3y | e−3y

| f (z)| = | f ( x + iy)| = = , (6.41)
| z2 + 4| | z2 + 4|
and see that in the upper half plane where y ≥ 0,

1
| f (z)| ≤ , (6.42)
| z2 + 4|
so that for any R > 2, the integral over the upper semi-circle
SR is bounded by
Z
πR
f (z)dz ≤ (6.43)
SR R2 −4
which goes to zero as R → ∞.
We then have
Z R
Io = lim f ( x )dx = 2πi Res( f ; 2i ) (6.44)
R→∞ − R
e3iz e −6
Res( f ; 2i ) = lim (z − 2i ) f (z) = lim = (6.45)
z→2i z→2i z + 2i 4i
2πi π Figure 6.5: Contour and poles for
⇒ Io = = 6. (6.46) Example 6.5.
4ie6 2e
Thus we have
π
I = Re( Io ) = . (6.47)
2e6

A useful result for evaluating indefinite integrals of trigonomet-

ric functions is known as Jordan’s Lemma.
84 complex analysis

Jordan’s Lemma† † Jordan’s Lemma is not to be con-

fused with Jordan’s Lambda.

Lemma 6.1 (Jordan’s Lemma). Let R be a positive real number,
and let f be a continuous complex valued function defined every-
where on the (positively oriented) semicircle SR in the upper half
plane. Suppose that for a given R there exists a non-negative real
number N ( R) such that | f (z)| ≤ N ( R), ∀z ∈ SR . Then for all
real s > 0,

πN ( R)
Z
f (z)eisz dz ≤ . (6.48)
SR s

Proof: The contour SR can be parameterised by z = Reiθ where

θ ∈ [0, π ]
Note that an equivalent lemma holds
Z Z π
f (z)eisz dz = f ( Reiθ )eiRs cos θ − Rs sin θ (iReiθ dθ ). for the semicircular contour in the
SR 0 lower half plane if s < 0.
Then we have
Z Z π
f (z)eisz dz ≤ R | f ( Reiθ )||eiRs cos θ − Rs sin θ |dθ
SR 0
Z π
≤ RN ( R) e− Rs sin θ dθ.
0

The integrand in the last line above is symmetric about θ = π/2, so

that we see
Z π Z π/2
− Rs sin θ
e dθ = 2 e− Rs sin θ dθ
0 0

From Figure 6.6, we see that sin θ ≥ 2θ/π for θ ∈ [0, π/2], so that
integral is bounded by
Z π/2 Z π/2 π Figure 6.6: From the plot of sin θ we
π
2 e− Rs sin θ dθ ≤ 2 e−2Rsθ/π dθ = [1 − e− Rs ] ≤ . can clearly see that sin θ ≥ 2θ/π for
0 0 Rs Rs θ ∈ [0, π/2].

Thus we must have

πN ( R)
Z
f (z)eisz dz ≤ ,
SR s
which proves Jordan’s Lemma.
Let’s take a look at an example where Jordan’s Lemma proves
useful.
Example 6.6: Using Jordan’s Lemma
Compute
Z ∞
x sin x
I=P dx (6.49)
−∞ 1 + x2
We can evaluate the integral
Z ∞
xeix
Io ≡ P dx (6.50)
−∞ 1 + x2
and then take the imaginary part to evaluate I. To apply
 applications of cauchy’s residue theorem 85

Jordan’s lemma, we see

z
≤ M/R (6.51)
1 + z2

for z ∈ SR for some constant M given large enough R. Thus

Jordan’s lemma (with s = 1) tells us

zeiz
Z
πM
lim dz ≤ lim =0 (6.52)
R→∞ SR 1 + z2 R→∞ 1 × R

so that by Cauchy’s residue theorem we have

zeiz
 
2πi
Io = 2πi Res ; +i = . (6.53)
(z + i )(z − i ) 2e

Taking the imaginary part gives us

π
I = Im( Io ) = . (6.54)
e

Converting the trigonometric function into the real or imaginary

part of an integral doesn’t always work, consider the following
example.

Example 6.7: Using Jordan’s Lemma Redux

Evaluate the integral
Z ∞
sin x
I=P dx. (6.55)
−∞ x+i
Solution: We note that we cannot say that this integral is
given by the imaginary part of the integral
Z ∞
eix
P dx (6.56)
−∞ x+i
since the complex valued denominator spoils this approach.
Moreover we can’t use sin z/(z + i ) either since it is un-
bounded in both upper and lower half-planes. Instead we
Figure 6.7: Pole and Contours for
will try the substitution sin x = (eix − e−ix )/(2i ), which Example 6.7.
splits the integral into two parts
Z ∞ Z ∞ −ix
eix
 
1 e
I= P dx − P dx . (6.57)
2i −∞ x+i −∞ x+i

For the first integral we have

Z ∞
eix
I1 ≡ P dx = 0 (6.58)
−∞ x+i
since by Jordan’s lemma we must have

eiz
Z
lim dz = 0 (6.59)
R→∞ S+
R
z+i
86 complex analysis

where S+R is the semicircular contour in the upper half plane,

since 1/(z + i ) → 0 along S+ R as R → ∞, and there are no
singularities in the upper half plane.
For the second integral we have
Z ∞ −ix
e
I2 ≡ P dx (6.60)
−∞ x+i

involves the function e−iz which is unbounded in the upper

half-plane, so we can close the contour in the lower half-
plane with the semicircle S−R . Applying Jordan’s Lemma for
s < 0 in the lower half plane we have

e−iz
Z
lim dz = 0. (6.61)
R→∞ S−
R
z+i

Noting that the closed contour for the lower half plane is
negatively oriented, we can obtain
 − 
e iz 2πi
I2 = −2πi Res ; −i = −2πi lim e−iz = − , (6.62)
z+i z→−i e

giving the solution

1 π
I= ( I1 − I2 ) = . (6.63)
2i e

Exercise 6.3: Practice on Improper Trig Integrals

Using the calculus of residues verify each of the following,
R ∞ cos(2x)
(a) P −∞ x2 +1 dx = eπ2 ;
R ∞ x sin x
(b) P −∞ x2 − 2x +10
dx = 3eπ3 (3 cos 1 + sin 1);
R∞ x
(c) 0 ( xcos π
2 +1)2 = 2e .

Exercise 6.4: More Improper Trig Integrals

Compute the following integrals,
R ∞ 3ix
(a) P −∞ xe−2i dx;
R∞ x sin(3x )
(b) P −∞ x 4 +4
dx;
R∞ e−2ix
(c) P −∞ x2 +4 dx.

6.4 Indented Contours: Singularities on the Real Line

In the previous sections the integrands f ( x ) were assumed to be

defined and continuous over the whole interval of integration. In
this section, we will explore definite real integrals where | f ( x )| →
 applications of cauchy’s residue theorem 87

∞ at particular (finite) points in that interval. We will see that these

improper integrals can be evaluated using the Cauchy’s principal
value approach, and by choosing the appropriate contours to be
able to apply the residue theorem.
Let f ( x ) be continuous and finite on [ a, b] except at the point
c ∈ ( a, b). Then the (improper) integral of f over [ a, b] is given by

Z b Z c −r Z b
f ( x )dx ≡ lim f ( x )dx + lim f ( x )dx, (6.64)
a r →0 a s →0 c + s

for r > 0 and s > 0, provided these limits exist.

Example 6.8: An Improper Integral on One Side

Determine the value of the improper integral,
Z 1
1
√ dx. (6.65)
0 x

Solution: The integrand goes to infinity at x = 0, so our im-

proper integral is one sided:
1
Z 1
1
Z 1
1 √
√ dx = lim √ dx = lim 2 x = 2. (6.66)
0 x s →0 s x s →0 s

Let’s consider an example where the integral of a singular inte-

grand does not converge when taken from each side.

Example 6.9: An Undefined Indefinite Integral

Determine the limits of both,
Z 2−r Z 4
dx dx
lim and lim . (6.67)
r →0 1 x−2 s →0 2+ s x−2
Solution: Both limits are infinite, since
Z 2−r 2−r
dx
lim = lim Log | x − 2| = lim Log r − Log 1 → −∞
r →0 1 x−2 r →0 1 r →0
Z 4 4
dx
lim = lim Log | x − 2| = lim Log 2 − Log s → ∞.
r →0 2+ s x−2 s →0 2+ s s →0

The value of the improper integral

Z 4 Z 2−r Z 4
dx dx dx
= lim + lim (6.68)
1 x−2 r →0 1 x − 2 s →0 2+ s x−2

does not exist because it depends on how each of r and s approach

0. We can define the Cauchy Principal Value of this integral by
approaching the singularity symmetrically.
88 complex analysis

The Cauchy Principal Value

of an Integral Over a Singularity
Definition 6.2 (The Cauchy Principal Value of an Integral
Over a Singularity). Let f ( x ) be continuous and finite on [ a, b]
except at the point c ∈ ( a, b). Then the symmetric limit
Z b  Z c −r Z b 
P f ( x )dx ≡ lim f ( x )dx + f ( x )dx (6.69)
a r →0+ a c +r

Rb
is called the Cauchy Principal Value of the integral a f ( x )dx.
When the f ( x ) is continuous on the whole real line except at c,
then the integral over (−∞, ∞) has Cauchy Principal Value
Z ∞  Z c −r Z R 
P f ( x )dx ≡ lim f ( x )dx + f ( x )dx . (6.70)
−∞ R→∞ −R c +r
r →0+

If the integral is over an interval with a finite number of isolated

singularities we extend the definition of the principal value in the
natural way.

R4 1
Thus we saw that while the improper integral 1 x −2 dx does not
exist, its Cauchy Principal Value is
Z 4
dx
P = lim Log 2 − Log r + Log r − Log 1
1 x−2 r →0+
= Log 2.

Figure 6.8: To apply Cauchy’s Residue

theorem to find the integral of a
function which is singular at z = c,
we use the semi-circular contour Sr to
close the contour as we approach the
singularity at c by taking z → 0.

consider the principal value of an integral from (−∞, ∞),

given by Equation (6.70). In order to use a closed contour and
Cauchy’s Residue Theorem to evaluate the principal values of such
 applications of cauchy’s residue theorem 89

integrals we need to connect the two sides of the principal value

at the singularity x = c. If we continued along the x-axis, we’d
pass through c and the integrand would become undefined. We
instead want to detour around c in a way that depends on r, so that
we can take the limit as r → 0. It is often convenient to choose the
semi-circle Sr , as shown in Figure 6.10, so that we need to find the
value
Z
lim f (z)dz. (6.71)
r → 0 + Sr

If c is a first order pole we can determine the value of this inte-

gral with the following lemma.
Integral of an Arc Around a Simple Pole
Lemma 6.2 (An Arc Around a Simple Pole). If function f has
a simple pole at z = c and Tr is the counterclockwise directed cir-
cular arc defined by

Tr = {z = c + reiθ |θ1 ≤ θ ≤ θ2 } (6.72)

then
Z
lim f (z)dz = i (θ2 − θ1 )Res( f ; c). (6.73)
r →0+ Tr Figure 6.9: Contour of the circular arc
for Lemma 6.2.

The value of (6.71) for the clockwise directed half-circle Sr in

Figure 6.10 is given by
Z
lim f (z)dz = −iπ Res( f ; c). (6.74)
r → 0 + Sr

Proof: Since f has a simple pole at c, its Laurent expansion has

the form
∞
a −1
f (z) = + ∑ ak (z − c)k (6.75)
z − c k =0

valid in some punctured neighbourhood of c, 0 < |z − c| < ρ1 where

the Laurent expansion converges. For radius r such that 0 < r < ρ1 ,
we can then write
dz
Z Z Z
f (z)dz = a−1 + g(z)dz (6.76)
Tr Tr z−c Tr

where g(z) ≡ ∑∞ k
k =0 ak ( z − c ) is analytic at c and hence is bounded
in some neighbourhood of this point,

| g(z)| ≤ M, for |z − c| < ρ2 . (6.77)

Therefore, for 0 < r < ρ2 , we must have

Z
g(z)dz ≤ M `( Tr ) = M(θ2 − θ1 )r (6.78)
Tr
90 complex analysis

which goes to zero as r goes to zero. This gives us

Z
lim g(z)dz = 0. (6.79)
r →0+ Tr

For the remaining term we can parameterise the arc as we did in

Example 3.1 to obtain
dz 1
Z Z θ2 Z θ2
= iθ
rieiθ dθ = i dθ = i (θ2 − θ1 ). (6.80)
Tr z−c θ1 re θ1

which is independent of r. Thus the limit of the integral around the

circular arc Tr is given by
Z
lim f (z)dz = a−1 i (θ2 − θ1 ) = i (θ2 − θ1 ) Res( f ; c). (6.81)
r →0+ Tr

Example 6.10: An Indented Contour

Evaluate
Z ∞ ix
e
I=P dx. (6.82)
−∞ x
Solution: First we note that the integrand is continuous
except at x = 0. Hence
Z −r ix Z R ix 
e e
I = lim + . (6.83)
R→∞
+
− R x r x
r →0

We can introduce the complex function f (z) ≡ eiz /z which

has a simple pole at the origin, but is elsewhere analytic. To
apply Cauchy’s Residue Theorem we must form a closed
contour containing the segments [− R, −r ] and [r, R]. Observ-
ing that Jordan’s Lemma applies to f (z) we can close the
contour on the "outside" using the semi-circular contour of
radius R, S+R , in the upper half plane joining z = − R and
z = R. To join z = −r to z = r we indent around the origin
using the half circle Sr going clockwise around the origin.
Since there are no singularities inside this closed contour, the
residue theorem gives us
Z −r ix
eiz dz
Z R ix
e dx e dx eiz dz
Z Z
+ + + =0 (6.84)
−R x Sr z r x S+
R
z Figure 6.10: Contour for Example 6.10.
Applying Jordan’s Lemma, we have

eiz dz
Z
lim = 0. (6.85)
R→∞ S+
R
z
While Lemma 6.2 provides

eiz dz
Z
lim = −iπ Res(0) = −iπ. (6.86)
r → 0 + Sr z
This gives us the solution
Z ∞ ix
e dx
I=P = iπ. (6.87)
−∞ x
 applications of cauchy’s residue theorem 91

Extra: What About Higher Order Poles on the Real Line?

You are probably wondering what happens if instead of a simple pole of the integrand along the
real line, there is a higher order pole or essential singularity. This case is more complicated than
that of the simple pole since Lemma 6.2 no longer applies. One may be tempted to conclude that
such an integral is always undefined, however, this is not always true.
If a function has a Laurent expansion with a term given by

a−n (z − c )−n , (6.88)

with integer n ≥ 2 and a−n 6= 0, then this term has an anti-derivative

a−n
( z − c )1− n . (6.89)
1−n
The integral of this term along Tr of Figure 6.9 is then given by
θ
a − n 1− n i (1− n ) θ 2
Z
In (r ) ≡ a−n (z − c)−n dz = r e (6.90)
Tr 1−n θ1
a−n −(n−1) −i(n−1)θ2
=− r [e − e − i ( n −1) θ1 ] (6.91)
n−1
a−n −(n−1) −i(n−1)θ1 −i(n−1)(θ2 −θ1 )
=− r e [e − 1] (6.92)
n−1
Here clearly Ir → ±∞ as r → ∞ as long as the exponential terms do not cancel out. That is,



 0 if n ≤ 0 (since this term is analytic and goes to zero by continuity);

i (θ − θ ) a if n = 1 (by Lemma 6.9);

2 1 1
lim In (r ) =
r →0 + 

 0 if n ≥ 2 and ∃k ∈ Z such that θ2 − θ1 = n2kπ−1 ;

±∞

otherwise.
(6.93)

So we see that for higher order poles we usually get undefined integrals over these shrinking arcs
unless the angle traversed by the arc happens to be just the right value for the order of the pole.
For the clockwise semi-circular arc Sr like we saw in Figure 6.10, where θ2 = 0 and θ1 = π, we
have



 0 if n ≤ 0;

−iπa

if n = 1;
−1
lim In (r ) = (6.94)
r →0 + 

 0 if n is odd and n > 2;

±∞

otherwise.

Exercise 6.5: Practice on Indented Contours

Using the calculus of residues, verify each of the following,
R∞ R∞
e2ix cos x −1
(a) P −∞ x +1 dx = πie−2i ; (d) 0 x2
dx = − π2 ;
R∞ eix
R∞ sin x −2
(b) P −∞ ( x −1)( x −2) dx= πi (e2i − ei ); (e) P −∞ ( x2 +4)( x −1) dx = 5 [cos(1) − e ];
π

R∞ sin(2x )

1 1
 R∞
(c) dx = − ; x cos x
0 2
x ( x +1) 2 π 2 e 2 (f) P −∞ x2 −3x +2 dx = π [sin(1) − 2 sin(2)].
92 complex analysis

6.5 Integrals Involving Branch Cuts

While trying to apply residue theory to compute an integral of

f ( x ) it may turn out that the complex function f (z) is multi-valued.
If this happens we need to modify our approach by taking into
account not only the isolated singularities of f (z), but also the
branch points and branch cuts. Since the functions are not analytic
(holomorphic) at these branch points and branch cuts, we must
carefully select the contours we use to apply Cauchy’s Integral
and Residue Theorems. Indeed, we will often find it necessary
to integrate along the branch cuts in the following instructive
examples.
√
Example 6.11: Integrating z on the Unit Circle
Consider the square root function
√ √
f (z) = z = reiθ/2 , 0 ≤ θ < 2π. (6.95)

such that it has a branch cut along the positive real axis.
Determine the integral of f (z) around the unit circle.
Solution: While there are no poles inside the unit circle, we
can’t just apply the Cauchy Integral Theorem directly to Figure
√ 6.11: To calculate the integral
the unit circle |z| = 1 since there is a discontinuity on this of z around the unit circle, using the
Cauchy Integral Theorem, we need
contour as we cross the positive real line at z = 1. Instead we to bypass the branch cut by taking
will construct the contour Γ, such that the closed contour Γ. The straight
I Z Z Z Z lines γ1 and γ2 are drawn separated
f (z)dz = f (z)dz + f (z)dz + f (z)dz + f (z)dz. from the branch cut for clarity, but
Γ C1 γ1 Cε γ2 are to represent evaluations using the
lower and upper values of the branch
Note that here the integrals over the straight lines γ1 and γ2 respectively along the real axis.
do not cancel out, since along γ1 the integrand has the value
√ iπ √
xe = − x, while the integrand along γ2 has the value
√
x.
We see that the integral around the branch point at the
origin must be bound by
Z √ √
zdz ≤ 2πε ε (6.96)
Cε

so that if we shrink the radius ε → 0+ we find

Z √
lim zdz → 0. (6.97)
r →0+ Cε

Since there are no poles inside the contour Γ, Cauchy’s

Integral Theorem says
I
f (z)dz = 0 (6.98)
Γ

So that we find
Z √ Z 0 √ Z 1√
zdz = − (− x )dx − xdx
C1 1 0
Z 1√
= −2 xdx = −4/3 (6.99)
0
 applications of cauchy’s residue theorem 93

Example 6.12: Integrating Along a Branch Cut

Calculate the integral
Z ∞
dx
I≡ √ (6.100)
0 x ( x + 4)
√
where x denotes the principal (positive) value for x > 0.
Solution: Including the explicit limits, what we are looking
for is
Z R
dx
I = lim √ .
R→∞ ε x ( x + 4)
ε →0+
√
We start by considering the branch of z defined by
√ √
z = e(Log r+iθ )/2 = reiθ/2 , 0 ≤ 0 < 2π.

which has a branch cut along the positive real axis. With this
√
choice of z we consider the complex function
1
f (z) ≡ √ .
z ( z + 4)
Then according to our choice of branch, on the upper side of
√
the branch cut, we have f ( x ) = 1/[ x ( x + r )] as desired. Figure 6.12: Contour and poles for
Example 6.12.
Now, we would like to be able to use Cauchy’s Residue
Theorem to evaluate this, so we need to find a closed con-
tour containing the positive real line, taking into account
the branch point (and pole) at the origin as well as the pole
at z = −4. We consider the closed contour of Figure 6.12,
where ε is small enough and R large enough so that the pole
at −4 lies inside the contour. Then we have
Z Z Z Z Z 
= + + + f (z)dz = 2πiRes( f ; −4)
Γ CR γ1 Cε γ2

As we let the radius R → ∞, we see

dz
Z Z
lim f (z)dz ≤ lim √
R→∞ CR R→∞ CR | z || z + 4|
dz
Z
≤ lim √
R → ∞ CR R ( R − 4)
2πR
≤ lim √ = 0.
R→∞ R ( R − 4)
As we let the inner radius ε → 0+ we find,
2πε
Z
lim f (z)dz ≤ lim √ = 0.
ε →0+ Cε ε →0+ ε (4 − ε )
Thus all that remains are the integrals along the top and
bottom of the branch cut. On the top of the branch cut the
integrand takes the value
1
f (x) = √
x ( x + 4)
94 complex analysis

while on the bottom of the branch cut, the integrand takes

the value
1
f (x) = − √
x ( x + 4)

Thus we find
Z R
−dx
Z ε 
dx
lim √ + √ = 2πi Res( f ; −4)
R→∞ R x ( x + 4) ε x ( x + 4)
+
ε →0
Z R
dx
lim 2 √ = 2πi Res( f ; −4)
R→∞ ε x ( x + 4)
ε →0+
π
I = πi Res( f ; −4) = .
2

Example 6.13: A Pole Along the Branch Cut

Determine the Cauchy Principal Value
Z ∞
dx
I≡P , where 0 < λ < 1. (6.101)
0 xλ (x − 4)

Solution: Here we have a singularity at x = +4, which lies

on the interval of integration, and we should also take the
positive real branch of x λ in the denominator. The integral
can then be recast as the limit
 Z 4− δ Z R 
dx
I = lim + . (6.102)
R→∞
+
ε 4+ δ x ( x − 4 )
λ
ε,δ→0

To evaluate this we must draw the contour as we did in the

previous example around the branch, but also indent around
the singularity at z = 4. We choose the branch Figure 6.13: Poles and Contour for
Example 6.13.
1
f (z) = , for z = reiθ , 0 ≤ θ < 2π.
eλ(Log r+iθ ) (reiθ − 4)

and form the contour Γ shown in Figure 6.13. From

Cauchy’s Integral Theorem we then have,
I
0= f (z)dz
ΓZ Z Z Z
= + + −
+
CR γ1 Sδ γ2
Z Z Z Z 
+ + + + f (z)dz (6.103)
Cε γ3 Sδ+ γ4

since f (z) has no singularities in the interior of the closed

contour. We can gather the contour integrals ofγ1 with γ4
and γ2 with γ3 , by noting that the integrands differ by a
factor of e−2πiλ between the top and bottom of the branch
 applications of cauchy’s residue theorem 95

cut yielding,
4− δ Z R 
Z
dx
0 = (1 − e−2πiλ ) +
ε 4+ δ x λ ( x − 4)
Z Z Z Z 
+ + −
+ + f (z)dz
CR Sδ Cε Sδ+
(6.104)

So that the solution is given by

Z ∞
dx
I=P
x λ ( x − 4)
0
−1
Z Z Z Z 
= lim + + + f (z)dz (6.105)
R→∞ 1 − e−2πiλ CR Sδ− Cε Sδ+
+
ε,δ→0

For 0 < λ < 1 it is straightforward to repeat the estimates

from Example 6.12, to find
Z Z
lim f (z)dz = 0, and lim f (z)dz = 0.
ε→0+ Cε R → ∞ CR
(6.106)

Along the upper half circle Sδ+ , function f agrees with the
principal value

1
f + (z) ≡ (6.107)
eλLog z (z − 4)

while along the lower half circle Sδ− , the function f takes the
value

f − (z) = e−2πiλ f + (z) (6.108)

The residues of interest are

Res( f + ; 4) = lim e−λLog z = 4−λ (6.109)

z →4
−λ −2πiλ
Res( f − ; 4) = 4 e (6.110)

Since z = 4 is a simple pole of f + and f − , Lemma 6.2 tells us

Z
f (z)dz = −πi Res( f + ; 4) = −iπ4−λ (6.111)
Sδ+
Z
f (z)dz = −πi Res( f − ; 4) = −iπ4−λ e−2πiλ (6.112)
Sδ−

which are independent of δ.

Thus we have

1 + e−2πiλ
I = iπ4−λ
1 − e−2πiλ
eiπλ + e−iπλ
= π4−λ i iπλ
e − e−iπλ
−λ
= π4 cot(πλ). (6.113)
96 complex analysis

Exercise 6.6: Practice on Integrals with Branch Cuts

Using the calculus of residues, verify each of the following,
R∞ √ R∞
x π xα π (1− α )
(a) 0 x 2 +1
dx = √ ; (d) 0 ( x +1)2
dx = 4 cos(απ/2)
,
2

R∞ for −1 < α < 3, α 6= 1;

x α −1 π
(b) 0 x +1 dx = sin(πα)
, R ∞ α −1
(e) 0 x2x+ x+1 dx = √ 2π
cos 2απ6+π csc(απ ),


3
for 0 < α < 1; for 0 < α < 2, α 6= 1;
R∞ xα 9α−1 πα R∞ α
(c) 0 ( x +9)2
dx = sin(πα)
, (f) P 0 x2x−1 dx = 2 sinπ(πα) [1 − cos(πα)],
for −1 < α < 1, α 6= 0; for −1 < α < 1, α 6= 0.