MATH2045

Vector Calculus
and
Complex Variable Theory

Dr Giampaolo D’Alessandro
September 2019
 MATH2045: Information about the module

1. Organisation of module:
This module is divided into two parts. Part I is on vector calculus and Part II is on complex variable
theory.

2. Arrangement of the lectures:

Part I will occupy weeks 1-5, with revision sessions in week 12. Part II will occupy weeks 6-11.

3. Blackboard:
Information on the module will be given on the Blackboard site. Students are expected to follow the
Blackboard site, and will be deemed to have received information posted on the Blackboard site for the
module.

4. Extra information:
Chapter 1, on vector algebra, is for revision. We will skip through this very quickly.

5. Problems class:
There will be a weekly problems sheet and a weekly problems class. Solutions to the problems will be
posted on Blackboard and we will go through them in the problems class.

6. Assessment:
Continuous assessment is worth 20% of the final mark. See detailed information on Blackboard.

7. Lectures and tutorial sessions:

There are three lectures and one tutorial session per week.

8. Acknowledgements:
Earlier versions of these notes are due to Prof Jim Anderson, Dr Bob Craine, Dr Ian Hawke, Prof Gareth
Jones, Prof Tim Sluckin, Prof Marika Taylor and Prof James Vickers.

9. Errors in these notes:

Considerable proofreading, but also editing, has occurred in the last two years. If you spot something
that you think is wrong, please contact the lecturer.

i
Contents

I Vector Calculus 1

1 Vectors - Revision 5
1.1 Vectors and Scalars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Symbols for vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Operations on vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.1 Equality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.2 Vector Addition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.3 Scalar Multiplication (multiplication by a number) . . . . . . . . . . . . . . . . . . . 6
1.4 Standard Basis Vectors and vector components . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Components and Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5.1 Length in Component Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5.2 Sum in Component Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5.3 Scalar multiplication in component form . . . . . . . . . . . . . . . . . . . . . . . . 8
1.6 Dot Product of Two Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.6.1 Definition of Dot Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.6.2 Properties of Dot Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.6.3 Projections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.7 Vectors in more than 2 dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.7.1 Three dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.7.2 Higher dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.8 Cross and Triple Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.8.1 Cross Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.8.2 Triple Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.8.3 Cross Products and Mechanics (optional) . . . . . . . . . . . . . . . . . . . . . . . . 12
1.9 Planes and lines in 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.9.1 Lines in 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.9.2 Planes in 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2 Vector and scalar valued functions and their derivatives 17

2.1 Curves and vector valued functions of a single variable . . . . . . . . . . . . . . . . . . . . . 17
2.2 Derivative of a vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.2 Velocity and acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.3 Derivative of a Dot Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.4 Length of a curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3 Scalar Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4 Vector fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Gradient and directional derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5.1 Directional derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5.2 The gradient of a scalar field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.5.3 Geometrical description of the gradient . . . . . . . . . . . . . . . . . . . . . . . . . 24

iii
iv

2.5.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.6 Properties of the gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.7 The divergence of a vector field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.8 The Laplacian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.9 The curl of a vector field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.10 Vector Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3 Integrals of vector and scalar fields 31

3.1 Line integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.1.1 Definition of line integral of a vector field . . . . . . . . . . . . . . . . . . . . . . . . 31
3.1.2 Parameterisation independence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.1.3 Examples of line integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2 Conservative vector fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3 Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3.1 The normal to a surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.4 Surface integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.4.1 Area of a surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.4.2 Flux integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

4 Integral Theorems 41
4.1 Stokes’ Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2 Green’s theorem (Stokes’ theorem in the plane) . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.3 Proof of Stokes’ theorem (optional) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.4 Conservative Vector Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4.1 Definition and properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4.2 Finding potentials for conservative vector fields . . . . . . . . . . . . . . . . . . . . . 48
4.4.3 Summary: four equivalent ways of characterising a conservative vector field . . . . . 48
4.5 Divergence theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.6 Vector potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.7 Summary of the integral theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

II Complex variable theory 53

5 Introduction to complex variable theory 57

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.2 Revision of complex numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.3 Continuity and differentiability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.4 Cauchy-Riemann equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.5 Useful functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.5.1 Multi-functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

6 Paths and integrals in the complex plane 67

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.2 Definition of path integral for a complex function . . . . . . . . . . . . . . . . . . . . . . . . 67
6.3 Path integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.3.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.3.2 Fundamental theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
6.4 Cauchy’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.5 Cauchy’s Integral Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.6 Series expansions of complex functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6.6.1 Taylor expansion of analytic functions . . . . . . . . . . . . . . . . . . . . . . . . . . 76
MATH2045 - Vector calculus v

6.6.2 Laurent series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

6.7 Singularities and residues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.7.2 Types of isolated singularities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.7.3 Classification of singularities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.7.4 Residues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
6.8 Cauchy’s Residue Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.9 Evaluation of real integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
6.9.1 Integrals of rational functions in [0, 2π] . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.9.2 Integrals on infinite subsets of the real axis . . . . . . . . . . . . . . . . . . . . . . . 86
6.10 Problem solving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

A Useful formulae & Revision material 89

 Part I

Vector Calculus

1
4
Chapter 1

Vectors - Revision

1.1 Vectors and Scalars

Some quantities, like temperature and pressure, are identified by their magnitude, They are called scalar quan-
tities, i.e., they are described by numbers:
Scalar = number.

Other quantities, like wind velocity are identified by both magnitude (size or length) and direction. Such
quantities are represented geometrically by arrows:

Vector = number (its size) and direction.

1.2 Symbols for vectors

−−→
The vector that joins point A with point B (see Figure 1.1) is indicated by AB. Another, more common symbol
is an underlined character v. Another symbol is − →v . In typed texts both are usually represented as a boldface
character, v. This is the notation used in these notes.
The modulus (magnitude, size, length) of a vector v is is the length of the arrow and is denoted by the symbol
−−→
|v|, mod(v) or |AB|.

1.3 Operations on vectors

We can add two numbers, we can multiply them. What can we do with vectors?

u3
u4 u5
u1 u2
A

Figure 1.1: The vector that joins Figure 1.2: Which of these vectors are equal? See Exam-
the point A with B. ple 1.1.

5
6

v u+v
v
u u

Figure 1.3: Graphical representation of the sum of the vectors u and v.

1.3.1 Equality
Two vectors are equal if they have the same length and the same direction.

Example 1.1 State which of the vectors in Figure 1.2 are equal.

1.3.2 Vector Addition

To obtain the sum of two vectors u and v draw a parallelogram that has sides u and v. The sum u + v is the
vector that joins the tail of u with the vertex of v (see Figure 1.3).

1.3.3 Scalar Multiplication (multiplication by a number)

If v is a vector and t is a real number (=scalar) then the scalar multiple tv is a vector with

1. magnitude |t| times that of v;

2. the same direction as v if t > 0;

3. the opposite direction to v if t < 0;

4. zero length and no direction if t = 0. In this last case we denote it by the zero vector, 0.

Example 1.2 Given a vector v draw the vector 2v and −v.

1.4 Standard Basis Vectors and vector components

For the moment we have worked with vectors without introducing a coordinate systems. All vector operations
are defined, but they are cumbersome to implement. Introducing a Cartesian coordinate system makes vector
operations “easy” to implement. The key to this advantage is the concept of coordinate basis vectors. In the
Cartesian plane there are two vectors that are very important:

1. The vector i from the origin to (1,0), i.e. a vector of length 1 in the x direction;

2. The vector j from the origin to (0,1), i.e. a vector of length 1 in the y direction.

They are called the standard basis vectors in the plane (see Figure 1.4). They allow us to represent two
dimensional vectors using pairs of numbers, the the x and y components of the vector: every vector v in the
plane can be written as a linear combination of the standard basis vectors, i.e. there exists two numbers a and b
such that
v = ai + bj.
The numbers a and b are the components of v along the Cartesian basis vectors. Figure 1.5 is an example of
this idea. We can indicate the vector v using the components as

v = (a, b) .

Writing vectors in terms of components makes the following rule obvious:

 7

y y

(0,1) j
i+2
3
j v= 2j

i x
(1,0) 3i x

Figure 1.4: The standard basis vectors in two Figure 1.5: Representation of the vector v =
dimensions. 3i + 2j, which can also be written as v =
(3, 2), on the standard basis vectors in two di-
mensions.

3 B y
A (3,3)

AB
v

1 A

x
1 4 x B (−1,0)

Figure 1.6: See Example 1.3. Figure 1.7: See Example 1.4.

−−→
The vector AB joining the two points A and B is given by:
−−→ −→ −−→ −−→ −→
AB = −OA + OB = OB − OA ,
−→ −−→
where OA and OB are the vectors from the origin to the points A and B, respectively.

Example 1.3 Find the component representation of the vector that joins A = (1, 1) with B = (4, 3) (see
Figure 1.6).

Example 1.4 Find the component representation of the vector v that joins A = (3, 3) with B = (−1, 0) (see
Figure 1.7).

1.5 Components and Operations

Once we have the components of a vector, we can easily evaluate vector operations.
8

y y

2u
v2
v

v
u

u+

2u 2
u2
u2 u
u1 x
x
u1 v1 2u 1

Figure 1.8: Component representation of the Figure 1.9: Component representation of the
sum of two vectors. product of a vector with a scalar.

1.5.1 Length in Component Form

p
This is a direct application of Pythagoras’ theorem so that if v = v1 i + v2 j then |v| = v12 + v22 .

Example 1.5 Compute the length of the vector v = 4i + 3j.

A vector of length 1 is called a unit vector. The unit vector in the v direction is denoted by v̂.
Given a vector v, the unit vector in the direction of v is

1
v̂ = v.
|v|

Example 1.6 Compute the unit vector in the direction of v = 4i + 3j.

1.5.2 Sum in Component Form

Suppose u = (u1 , u2 ) and v = (v1 , v2 ). Then the sum of u and v has components:

u + v = (u1 i + u2 j) + (v1 i + v2 j)
= (u1 + v1 )i + (u2 + v2 )j ,

so that u + v has components (u1 + v1 , u2 + v2 ) (see Figure 1.8).

Example 1.7 Use the components representation to sum the two vectors u = (1, 1) and v = (−1, −2).

1.5.3 Scalar multiplication in component form

Suppose u = (u1 , u2 ). The scalar multiplication tu has components:

u = (u1 i + u2 j)
tu = t(u1 i + u2 j) = tu1 i + tu2 j ,

so that tu has components (tu1 , tu2 ) (see Figure 1.9).

Example 1.8 Find the component representation of the vector tu with t = −1 and u = (2, −1).
 9

v (−2,4) y

(2,1)
θ
u x

Figure 1.10: Geometrical interpretation of the Figure 1.11: The vectors used in Exam-
dot product: see equation (1.1). ple 1.10: they are orthogonal and their dot
product is zero.

1.6 Dot Product of Two Vectors

1.6.1 Definition of Dot Product
The dot product (also called the scalar product) of two vectors is defined as follows. Given two vectors:

u = u1 i + u2 j = (u1 , u2 )
v = v1 i + v2 j = (v1 , v2 )

we define their dot product u · v to be the sum of the products of their corresponding components:

u · v = u1 v 1 + u2 v 2 .

Note that this product is a number (scalar) and not a vector.

Example 1.9 Compute the dot product of u = (2, 1), and v = (1, −3).

Example 1.10 Compute the dot product of u = (2, 1) and v = (−2, 4).

Remark - Example 1.10 shows that u · v can be zero even though both u and v are non-zero.

1.6.2 Properties of Dot Product

1. u · v = v · u;

2. u · (v + w) = u · v + u · w;

3. v · v = |v|2 ;

4. (tu) · v = t(u · v).

The dot product has a geometrical interpretation: if θ is the angle between the directions of u and v with
0 ≤ θ ≤ π) (see Figure 1.10), then
u·v
u · v = |u||v| cos(θ) =⇒ cos(θ) = . (1.1)
|u||v|
A consequence of this result is that two non–zero vectors are orthogonal (perpendicular) to one another if and
only if their dot product is zero. In fact, if θ = π/2 then
π
θ= =⇒ cos(θ) = 0 =⇒ u · v = 0.
2
10

z
v
P(x,y,z)
u
k z
j y
i
θ s x
y

Figure 1.12: Geometrical interpretation of Figure 1.13: The Cartesian coordinate system
the scalar projection of a vector: see equa- in three dimensions with the standard basis
tion (1.2). vectors i, j and k.

Conversely, provided that |u| =

6 0 and |v| 6= 0 then
π
u · v = 0 =⇒ cos(θ) = 0 =⇒ θ = .
2
So the vectors u and v in Example 1.10 are orthogonal (see Figure 1.11).

Example 1.11 Find the angle between i + j and i − 3j.

1.6.3 Projections
It is useful to project one vector onto another, i.e., to find the component of a vector u along another vector v.
From Figure 1.12 we can see that the scalar projection s of a vector u in the direction of a nonzero vector v is
the dot product of u with a unit vector in the direction of v:
 
u·v u·v v
s = |u| cos(θ) = |u| = =u· = u · v̂ . (1.2)
|u| |v| |v| |v|

The vector projection of u in the direction of v is the scalar multiple of a unit vector v̂ (in the direction of v)
by the scalar projection of u in the direction of v:

uv = sv̂ = (u · v̂)v̂ .

Example 1.12 Compute the scalar and vector projection of u = (2, 3) on v = (1, 1).

1.7 Vectors in more than 2 dimensions

1.7.1 Three dimensions
In three dimensions we can use the Cartesian coordinate system x, y, z and represent a vector by its 3 com-
ponents along the coordinate axes (see Figure 1.13). All the operations defined for vectors in 2D are valid for
vectors in 3D:

1. Length: If v = (1, 2, −2) then p

|v| = 12 + 22 + (−2)2 ;

2. Sum: If v = (v1 , v2 , v3 ) and w = (w1 , w2 , w3 ) then

v + w = (v1 + w1 , v2 + w2 , v3 + w3 ) ;
 11

3. Scalar product: If v = (v1 , v2 , v3 ) and w = (w1 , w2 , w3 ) then

v · w = v1 w1 + v2 w2 + v3 w3 = |v||w| cos(θ).

So if v = (1, −1, 0) and w = (2, −2, −3) then

v · w = 1 × 2 + (−1) × (−2) + 0 × (−3) = 4 .

4. Standard basis vectors: The standard basis vectors in 3D are indicated by i, j and k (see Figure 1.13),
so that if v has components v1 , v2 , v3 along the x, y and z axes respectively, then

v = v1 i + v2 j + v3 k .

1.7.2 Higher dimensions

In principle we can define vectors in any dimensional space, e.g., v = (1, 2, −4, 6), w = (2, 3, 0, −3) are four
dimensional vectors. Higher dimensional spaces are important for specifying more complicated properties. For
example the dynamical state of a classical particle moving in 3 dimensions is 6 dimensional: 3 components
to represent the position and 3 components to represent the velocity. Everything we have so far defined can
be generalised to higher dimensions in ways similar to how we went from from 2D to 3D. The sum of two
4-vectors is just the sum of the respective components, e.g., v + w = (3, 5, −4, 3). Their dot product is
v · w = 1 × 2 + 2 × 3 + (−4) × 0 + 6 × (−3) = −10.

1.8 Cross and Triple Product

These two operations on vectors are defined only in three dimensions.

1.8.1 Cross Product

Definition 1.1 (Cross product) Given two vectors

u = u1 i + u2 j + u3 k and v = v1 i + v2 j + v3 k, (1.3)

their cross product (or vector product) is the vector

u × v = (u2 v3 − u3 v2 ) i − (u1 v3 − u3 v1 ) j + (u1 v2 − u2 v1 ) k . (1.4)

Properties:

1. u × v is perpendicular to both u and v (see Figure 1.14);

2. |u × v| is the area of the parallelogram with sides u and v:

|u × v| = |u||v| sin(θ) ;

3. The direction of u × v is the normal to the parallelogram given by the ‘right hand screw rule’ going from
u to v.

4. u × u = 0;

5. u × v = −v × u, i.e. the cross product is anti-symmetric.

6. (u + v) × w = u × w + v × w.
12
11111111111111111111111111
00000000000000000000000000
00000000000000000000000000
11111111111111111111111111
00000000000000000000000000
11111111111111111111111111
00000000000000000000000000
11111111111111111111111111
u v 00000000000000000000000000
11111111111111111111111111
00000000000000000000000000
11111111111111111111111111
v
00000000000000000000000000
11111111111111111111111111
00000000000000000000000000
11111111111111111111111111
00000000000000000000000000
11111111111111111111111111
w
00000000000000000000000000
11111111111111111111111111
00000000000000000000000000
11111111111111111111111111
00000000000000000000000000
11111111111111111111111111
00000000000000000000000000
11111111111111111111111111
u
00000000000000000000000000
11111111111111111111111111
00000000000000000000000000
11111111111111111111111111
v
00000000000000000000000000
11111111111111111111111111
00000000000000000000000000
11111111111111111111111111
00000000000000000000000000
11111111111111111111111111 u
00000000000000000000000000
11111111111111111111111111

Figure 1.14: Geometrical interpretation of the Figure 1.15: Geometrical interpretation of the
cross product of two vectors. triple product of three vectors.

The most convenient way to compute cross products is using the determinant formula for u × v. If u =
u1 i + u2 j + u3 j and v = v1 i + v2 j + v3 j, then
 
 i j k 

u × v =  u1 u2 u3  , (1.5)
 v1 v2 v3 

where the determinant is normally expanded along the first row.

Example 1.13 Compute the cross product of the basis vectors i, j and k.

Example 1.14 Compute the cross product of u = 3i and v = 5k.

Example 1.15 Compute the cross product of u = 2i + j − 3k and v = −2j + 5k.

1.8.2 Triple Product

Definition 1.2 (Triple product) The triple product of three three dimensional vectors u, v, and w is the scalar

u · (v × w) . (1.6)

The triple product can be conveniently computed using a determinant rule. If u = u1 i + u2 j + u3 j, v =

v1 i + v2 j + v3 j and w = w1 i + w2 j + w3 j, then
 
 u1 u2 u3 
 
u · (v × w) =  v1 v2 v3  . (1.7)
 w1 w2 w3 

The triple product has a geometrical meaning: its absolute value is the volume of the parallelepiped spanned
by the three vectors u, v and w (see Figure 1.15). Consequently u, v and w are only coplanar if their triple
product is zero.

1.8.3 Cross Products and Mechanics (optional)

Angular motion: If a point in a rigid body is rotating about an axis with angular velocity ω (yes a vector!),
then the linear velocity of a particle with position vector r, relative to the axis of motion is v = ω × r
Angular Momentum: Suppose a planet of mass m is in orbit at position r with linear velocity v. Take the
centre of rotation (e.g., the sun) as being the origin of coordinates. The angular momentum of the planet is then
L = r × (mv).
Torque: The torque T of a force F applied to a point with position vector r with respect to the origin is
T = r × F.
 13

P0
P v
r0
r
x

Figure 1.16: Line through r 0 and parallel to a vector v.

1.9 Planes and lines in 3D

We can use vectors to find equations for planes and lines in three dimensions.

1.9.1 Lines in 3D
Given a point P0 : (x0 , y0 , z0 ) and a vector v = ai + bj + ck , there is one and only one line through P0
parallel to v (see figure 1.16). If P : (x, y, z) is a general point on this line, then from figure 1.16 you can see
that r − r 0 is parallel to v. Thus there is a number t such that

r − r 0 = tv ⇒ r = r 0 + tv. (1.8)

This is the vector equation of the line and the vector r can be considered as a function of the real parameter t,
i.e. r(t). All the points on the line can be obtained by varying t from −∞ to +∞. The vector v is called the
direction vector of the line.
We can break the vector equation up into a parametric form by considering the components of a general
point. Let v = ai + bj + ck then

r = xi + yj + zk = x0 i + y0 j + z0 k + t(ai + bj + ck)
= (x0 + ta)i + (y0 + tb)j + (z0 + tc)k,

so that we have:
x = x0 + at,
y = y0 + bt, −∞ < t < +∞.
z = z0 + ct,

These are called the parametric equations of the line, since they involve a parameter, t. To obtain the Cartesian
equation of the line we simply eliminate t between the above equations and end up with the odd-looking
“double-equals” expression, which is shorthand for two equations:

x − x0 y − y0 z − z0
= = , a, b, c 6= 0.
a b c

Example 1.16 Find the line through P : (1, 1, 1) parallel to v = −i + j + 2k in vector, parametric and
Cartesian forms.

Example 1.17 Find the line through P : (1, 0, 1) parallel to v = i + 2j in vector, parametric and Cartesian
forms.
14

z
n
P0
r − r0
r0 P
r
x

Figure 1.17: Plane through a point P0 and perpendicular to a vector n.

1.9.2 Planes in 3D
We can use vectors to find equations for planes in 3D. The easiest case is the one we will always deal with:

Find the equation of the plane which passes through a point P0 and which is perpendicular to a
non-zero vector n.

The vector n is called the normal vector to the plane (see figure 1.17). From the figure, we see that all the
points P of the planes are such that r − r 0 is perpendicular to n. Expressed mathematically, we therefore have,

n · (r − r 0 ) = 0 (1.9)

This is the vector equation for the plane. It can also be written in scalar form using the Cartesian representa-
tions:
P0 : (x0 , y0 , z0 ) , r 0 = x0 i + y0 j + z0 k,
P : (x, y, z) , r = xi + yj + zk,
n = Ai + Bj + Ck.
Using these substitutions in equation (1.9) we can write the equation for the plane in Cartesian coordinates:

0 = n · (r − r 0 )
= (Ai + Bj + Ck) · [(x − x0 )i + (y − y0 )j + (z − z0 )k]
= A(x − x0 ) + B(y − y0 ) + C(z − z0 ).

Hence the equation for the plane in Cartesian coordinates is

Ax + By + Cz = D, (1.10)

where D = Ax0 + By0 + Cz0 .

If at least one of the coefficients A, B, C, is non-zero, an equation of this type represents the equation of a
plane perpendicular to
n = Ai + Bj + Ck. (1.11)

Example 1.18 Find the equation in Cartesian coordinates of the plane through (2, 0, 1) and perpendicular to
n = 3i − 2j − 2k.

Example 1.19 Find a vector normal to the plane 2x − 3y − 4z = 0.

Example 1.20 Find the equation of the plane that passes through the three points P = (1, 1, 0), Q = (0, 2, 1)
and R = (3, 2, −1).
 15

Observation - In the same way that equation (1.8) can be considered as defining a line as a vector function r(t)
that depends on a real parameter t, we can consider the vector equation for the plane, equation (1.9), as a vector
function of two parameters, r(s, t). To see that this is the case, call u1 and u2 two vectors perpendicular to n
and that are not parallel to each other. Any point in the plane perpendicular to n and through the point P can
be written as
−−→
r(s, t) = OP + su1 + tu2 , (1.12)
with s and t to real parameters. Equation (1.12) is not the most practical way to write the equation of the
plane, but it introduces a very powerful notation, vectors functions of two variables, that we will use in the next
chapter to discuss how to represent surfaces.
16
Chapter 2

Vector and scalar valued functions and their

derivatives

2.1 Curves and vector valued functions of a single variable

A vector valued function is a vector r which depends on a real variable, which we will denote t. We write such
a vector valued function as r(t). Sometimes t will take on all real values and sometimes t will be over some
interval, I. If we take a < t < b then I is an open interval and we write I = (a, b) however if we include the
end points so that a ≤ t ≤ b then I is a closed interval and we write I = [a, b].
More formally we can think of a vector valued function as a map

I → R3 : t 7→ r(t). (2.1)

For most of theses notes we assume, unless specified otherwise, that the map is smooth. This means that one
can differentiate the map as often as one likes. However, in some of the theorems we need to be careful about
the degree of differentiability of functions. A continuous function is said to be C 0 , a differentiable function
with continuous first derivative is said to be C 1 , whilst a twice differentiable function with continuous second
derivative is sad to be C 2 , and so on.
As t varies between a and b the vector r(t) traces out a set of points (see figure 2.1). We call this set of points
a curve and we refer to the map (2.1) as a parameterised curve. This is because it is possible to obtain the
same curve in R3 with a different parameterisation. Let f : R → R be a monotonic increasing (smooth)
function. Let f (c) = a and f (d) = b and define I˜ = [c, d]. Then we may define a new vector valued function
r̃(s) = r[f (s)] ≡ r(t) for c ≤ s ≤ d. So that

I˜ → R3 : s 7→ r̃(s)

r(t1 )
y
r(t0 )

r(ti−1 )

r(tk )
a b
r(tN )
t0 t1 t2 tk tN

Figure 2.1: As t varies in the interval [a, b] the vector function r(t) outlines a curve γ in R3 .

17
18

is also a parameterised curve. However as s varies between c and d, r̃ traces out the same set of points as r.
Thus r and r̃ are simply different parameterisations of the same curve.
Observation - From a physical and geometrical point of view, as t varies the function r(t) describes the motion
of a point P in Rn , whose coordinates are the components of r(t). The vector r(t) is the position vector of the
point P . In physics examples t is often time.

2.2 Derivative of a vector

2.2.1 Definition
Definition 2.1 (Derivative of a vector) The derivative of a parameterised curve r(t) is defined as the limit
dr r(t + h) − r(t)
= lim . (2.2)
dt h→0 h
If we give the components of r(t) as

r(t) = x(t)i + y(t)j + z(t)k, (2.3)

then the derivative has components

dr dx dy dz
= i+ j+ k. (2.4)
dt dt dt dt
Geometrically, the direction of the derivative is tangent to the curve r(t).

2.2.2 Velocity and acceleration

We have already mentioned that we can interpret the function r(t) as the position vector of a point P . Similarly,
the derivative of the position vector is the velocity of the particle1 ,
dr dx dy
v(t) ≡ = i+ j. (2.5)
dt dt dt
This quantity is also a vector, since we need to specify not only the scalar speed, but also the direction of
motion.
The corresponding speed of the particle is given by the length of the velocity vector. In two dimensions the
formula is: s 
dx 2
 2
dy
v(t) = |v(t)| = + . (2.6)
dt dt
The velocity vector, being the derivative of the position vector, is tangent to the trajectory of the particle (see
Figure 2.3).
Finally, the acceleration of the particle (also a vector) is the derivative of the velocity, or the second derivative
of the position vector:
dv d2 r d2 x d2 y
a= = 2 = 2i+ 2j. (2.7)
dt dt dt dt
Example 2.1 Compute the velocity, speed, acceleration of the particle position vector

r(t) = t3 i − (4t2 − 8t + 1)j . (2.8)

Hence compute the particle position, velocity, speed and acceleration at t = 1.

Does the acceleration in Example 2.1 cause the particle to speed up or slow down? To answer this, it is useful
to consider the derivative of a dot product.
1
For ease of notation in this and the following section we work in two dimensions. However, all results obtained are also valid in
Rn .
 19

y P v(t)
r(t)

r = x i+ y j
x
x

Figure 2.2: Position vector r of a point P with Figure 2.3: The derivative v(t) of the position
coordinates (x, y). vector r(t) is tangent to the curve described
by r(t).

2.2.3 Derivative of a Dot Product

Let u = u1 (t)i + u2 (t)j and v = v1 (t)i + v2 (t)j then u · v = u1 (t)v1 (t) + u2 (t)v2 (t). We now compute the
derivative of this dot product:

d d
(u · v) = [u1 (t)v1 (t) + u2 (t)v2 (t)]
dt dt
du1 dv1 du2 dv2
= v 1 + u1 + v 2 + u2
dt dt dt dt
du1 du2 dv1 dv2
= v1 + v 2 + u1 + u2 ,
dt dt dt dt

so that
d du dv
(u · v) = ·v+u· . (2.9)
dt dt dt
We can use this formula to find out how the speed of an object changes when it accelerates. We start by writing
the speed in terms of dot-products:

speed ≡ v(t) = |v| = (v · v)1/2 .

We now compute the derivative of v(t):

d d
|v| = (v · v)1/2
dt dt
1 d
= (v · v)−1/2 (v · v)
2 dt
 
1 dv dv
= (v · v)−1/2 ·v+v·
2 dt dt
dv
= (v · v)−1/2 ·v
dt
= a · v̂ ,

so that if a · v̂ > 0 then the speed is increasing, while if a · v̂ < 0 then the reverse happens.

Example 2.2 Determine if the speed of the position vector discussed in Example 2.1, equation (2.8), is increas-
ing (acceleration) or decreasing (deceleration).
20

2.2.4 Length of a curve

 
dr  dr 
If we think of the parameter t representing time, then is the velocity and v(t) =   is the speed. This
dt dt
gives us the following formula for the length of the curve between the point A with position vector r(t0 ) and
the point B with position vector r(t1 ):

t1 t1
  Z t1 " 2  2  2 #1/2
 dr  dx dy dz
Z Z
L= v(t) dt =   dt =
 dt  + + dt. (2.10)
t0 t0 t0 dt dt dt

What happens if we use some other parameterisation of the same curve? Let t = f (s) and r̃(s) = r[f (s)] =
r(t) as above. Then by the chain rule
dr̃(s) dr(t) dt
= .
ds dt ds
Let f (s0 ) = t0 and f (s1 ) = t1 . Then the length of the curve from A to B using the s parameterisation is given
by
Z s1   Z s1  
 dr̃(s)   dr(t) dt 
L̃ =  ds  ds =
 
 dt ds  ds
 
s0 s0
Z s1   Z t1  
 dr(t)  dt  dr(t) 
=  dt  ds ds =
 
 dt  dt = L.
 
s0 t0

Hence the length of the curve is independent of the parameterisation. This feature of the integral is called
reparameterisation invariance. [Note that to go between the second and third steps we used the fact that t
increases as s increases, so dt/ds is positive.]

2.3 Scalar Fields

Definition 2.2 (Scalar field) A scalar field is a function

φ : D ⊆ Rn → R : x ∈ Rn 7→ φ(x) ∈ R. (2.11)

In other words, a scalar field is a scalar valued function that depends upon position. An example is the
temperature in a room. In 2-dimensional space a scalar field is a map

φ : D ⊆ R2 → R : (x, y) 7→ φ(x, y). (2.12)

An example of a scalar field in 2 dimensions is given by

1 2
φ(x, y) = 1 − (x2 + y 2 ) + (x + y 2 )2 . (2.13)
40

Visualisation of scalar fields

To visualise a two–dimensional scalar field, φ(x, y), we can plot the graph of the surface given by z = φ(x, y),
as in figure 2.4, for example: at each point (x, y) in the R2 plane we associate a third coordinate z = φ(x, y).
An alternative is to plot the contour lines or level surfaces given by z = c, with c an appropriate constant.
We plot the curves of the function φ(x, y) in equation (2.13) for a number of equally spaced contours in
figure 2.5. Note that in drawing these contour lines we must assume that c is larger than the minimum value of
φ(x, y), i.e. c ≥ −9. Note that even though we have plotted the contour lines of equation (2.13) equally spaced
in value, their relative distance varies. In particular they are denser the further away we are from the origin,
indicating that the surface is steeper there.
 21

5
10

10 0

y
φ(x, y)

5 0
0 5
-5
-5 0 -5

0 -5
y -5 0 5
x 5 -5 x

Figure 2.4: Surface plot of φ(x, y) defined in Figure 2.5: Contour lines of the function φ(x, y)
equation (2.13). defined in equation (2.13).

Figure 2.6: Level surfaces of the function φ(x, y, z) = x + y + z.

Example 2.3 Given the scalar field

f (x, y) = 4 − (x − 1)2 − y 2 , (2.14)

plot the graph z = f (x, y) and the contour lines f (x, y) = c.

We now consider scalar functions in 3 dimensions. These are maps

φ : D ⊆ R3 → R : (x, y, z) 7→ φ(x, y, z). (2.15)

The graphs of this function w = φ(x, y, z) are 3-dimensional hypersurfaces in 4-dimensional space - which are
not so easy to draw! Instead we draw the level surfaces given by

φ(x, y, z) = c,

with c an appropriate constant. As with contour lines we can visualise the function by drawing a series of
appropriately chosen level surfaces. These are 2-dimensional surfaces in R3 so drawing them is feasible.

Example 2.4 Draw the level surfaces of the function φ(x, y, z) = x + y + z.

22

Figure 2.7: Plot of the vector field F (x, y) = yi− Figure 2.8: Vector field plot (arrows) and field
xj (arrows) and its field lines (dashed circles). lines (dashed lines) for the vector field F (r) =
−kr̂/r2 (see Example 2.7).

2.4 Vector fields

Definition 2.3 (Vector field) A vector field is a function

F : D ⊆ Rn → Rn : x ∈ D 7→ F (x) ∈ Rn . (2.16)

In other words, a vector field is a vector valued function that depends upon position. In 2 dimensions it is a
map
F : D ⊆ R2 → R2 : (x, y) 7→ F (x, y) = F1 (x, y)i + F2 (x, y)j

Visualisation of vector fields

How do we visualise a vector field?

One option is to draw at the point (x, y) the vector F (x, y). This is the vector plot of the field.

Example 2.5 Draw the vector plot of the vector field

F (x, y) = yi − xj . (2.17)

Another way to represent vector fields is to use field lines.

Definition 2.4 (Field lines) A differentiable curve r(t) ∈ Rn , t ∈ [a, b] ⊆ R, is a field line of an n-dimensional
vector field F (x), if
dr
= F [r(t)], (2.18)
dt
i.e. if the vector field F at r(t) is the vector tangent to the curve there.

Geometrically, field lines are curves that are everywhere parallel to the vector field. They are also known as
lines of force 2 .
2
Field lines were first constructed by the Michael Faraday (1791-1867) as a way of picturing magnetic interactions. He is best
known for his law of magnetic induction. Faraday, who was not at all mathematical, inspired the Scottish mathematical physicist James
Clerk Maxwell (1830-1879) to develop his famous equations describing electromagnetism.
 23

The procedure to find the field lines is as follows3 . We start from the definition of field line, i.e. a curve
r(t) = x(t)i + y(t)j whose tangent vector is the vector field F (x, y):

dx

dr
 = Fx [x(t), y(t)],
dt

= F [x(t), y(t)] =⇒ (2.19)
dt  dy = F [x(t), y(t)].

y
dt
This is a set of coupled differential equations for x(t) and y(t). In general it is not possible to simplify it any
further. However, if the field line can be expressed as a function y(x) then it is possible to eliminate t from
equations (2.19) and write a single ordinary differential equation for y(x):

dy dy dt Fy
= = . (2.20)
dx dt dx Fx

Example 2.6 Find the field lines of the vector field F used in Example 2.5, equation (2.17).

Observations:

1. The label on each field line is a constant of integration.

2. Example 2.5 turns out to be easy to integrate. For general Fx (x, y), Fy (x, y), and in higher dimensions
dimensions, the calculation is much harder.

3. The integration does not tell us which direction the field line goes. This must be determined by, for
example, plotting the vector field at one point of a field line.

Example 2.7 Let r = xi + yj denote a generic point in R2 . Let the vector field F (r) be defined by

r k
F (r) = −k = − 2 r̂.
|r|3 r

This represents an inverse square force law (such as gravitation or electrostatic). Draw the vector plot and the
field lines of this field.

2.5 Gradient and directional derivatives

2.5.1 Directional derivative
Let φ(r) be a scalar field. If we now try to define the derivative of φ then we must take into account the fact
that this derivative depends on what direction we move in.

Definition 2.5 (Directional derivative) Let v̂ be some fixed unit vector. Then the directional derivative of a
scalar field φ in the direction of v̂, at the point r 0 , is given by

∂φ φ(r 0 + tv̂) − φ(r 0 )

(r 0 ) ≡ ∇v̂ φ(r 0 ) = lim . (2.21)
∂ v̂ t→0 t
This measures the change in φ in the direction of v̂.

Observation - The definition 2.5 can be extended to a non–unit vector v as

∂φ φ(r 0 + tv̂) − φ(r 0 )
≡ ∇v φ(r 0 ) = lim , (2.22)
∂v t→0 t
where v̂ is the unit vector in the direction of v.
3
Again, for ease of notation we work in two dimensions, but the results are generic.
24

2.5.2 The gradient of a scalar field

The formal definitions of directional derivative, equation (2.21) or (2.22), are not very useful when it comes to
doing calculations so we will convert them into something that we can use, by working backwards. To do this
we think of r 0 and v̂ as fixed and define a real valued function f (t) = φ(r 0 + tv̂). Then
∂φ φ(r 0 + tv̂) − φ(r 0 ) f (t) − f (0)
= lim = lim = f ′ (0). (2.23)
∂v t→0 t t→0 t
We can write f (t) as:
f (t) = φ(r 0 + tv̂) = φ(x, y, z) , (2.24)
where
x = r1 + tv̂1 ,
y = r2 + tv̂2 ,
x = r3 + tv̂3 ,
with r 0 = r1 i + r2 j + r3 k and v̂ = v̂1 i + v̂2 j + v̂3 k. Differentiating (2.24) with respect to t using the chain
rule gives
 
df  ∂φ dx ∂φ dy ∂φ dz 
= + +
dt t=0 ∂x dt ∂y dt ∂z dt t=0
(2.25)
∂φ ∂φ ∂φ
= v̂1 + v̂2 + v̂3 .
∂x ∂y ∂z
We now introduce a vector field of fundamental importance in vector calculus, the gradient of a scalar field.
Definition 2.6 (Gradient of a scalar field in Cartesian components) The gradient of a differentiable scalar
field φ(x, y, z), is the vector field with Cartesian components
∂φ ∂φ ∂φ
∇φ = i+ j+ k. (2.26)
∂x ∂y ∂z
The symbol ∇ is called gradient or nabla. An alternative notation for ∇φ is grad φ.
Comparing equations (2.23), (2.25) and (2.26) we see that we can write the directional derivative as
∂φ
= v̂ · ∇φ . (2.27)
∂v
Note that the dot product on the right hand side of this equation involves the unit vector v̂, not the generic
vector v.
Example 2.8 Let φ(x, y, z) = xy + z 2 . Determine the rate of change of φ in the direction of v = 3i + 4k at
the point (x, y, z) = (1, 2, 1).

2.5.3 Geometrical description of the gradient

Gradient and steepest ascent
When calculating ∇φ using equation (2.26) we need to know φ as a function of the Cartesian coordinates x, y
and z. However we often need to use other coordinate systems, so we now give a geometrical definition of ∇φ
which enables us to calculate it in any coordinate system.
Given a scalar field φ(r) in what direction is it increasing most? Let v̂ be a unit vector then
∇v̂ φ = v̂ · ∇φ = |v̂||∇φ| cos(θ) = |∇φ| cos(θ) . (2.28)
This is maximum when θ = 0 and minimum when θ = π, so that
− |∇φ| ≤ ∇v̂ φ ≤ |∇φ| . (2.29)
We have therefore established the following theorem:
 25

y tan
ge
nt
pla
ne
11
00
00
11
v
P

φ ( r) = c 0

x
Figure 2.9: Level surface (solid curve) of a function φ(r). On this surface φ(r) = c0 , with c0 a given constant.
The vector v is tangent to the level surface at P : it belongs to the tangent plane at P and is, hence, perpendicular
to ∇φ there.

Theorem 2.1 The directional derivative ∇v̂ φ is maximised when v̂ points in the same direction ∇φ and min-
imised when it points in the opposite direction. Furthermore its maximum and minimum values are given by
±|∇φ|.

Gradients and level surfaces

Consider a scalar field φ(x, y, z) and point P with coordinates (x0 , y0 , z0 ). Let φ(x0 , y0 , z0 ) = c0 . The level
surface of φ through the point P is given by the set of points (x, y, z) such that φ(x, y, z) = c0 . Now let v be a
tangent vector at P , i.e. v is an element of the tangent plane at P (see figure 2.9): since v is tangent to the level
surface φ(x, y, z) = c0 , the derivative of φ in the v direction is zero. Thus
∇v φ = 0 =⇒ v · ∇φ = 0
=⇒ the tangent vector v is perpendicular to ∇φ
=⇒ ∇φ is perpendicular to the tangent space at P
=⇒ ∇φ is normal to the level surface φ = c0 .
We have therefore established the following:

Theorem 2.2 The gradient of a differentiable scalar field φ, ∇φ, is normal to the surfaces φ = const and
points in the direction of increasing φ.

A coordinate independent definition of the gradient of a scalar function

We now use theorems 2.1 and 2.2 to obtain a new formula for the gradient. By Theorem 2.2, we know that ∇φ
is normal to the surface φ = const, i.e.
∇φ = k n̂, (2.30)
where k = |∇φ| and n̂ is the unit vector normal to the surface φ = c0 in the direction of increasing φ.
On the other hand, we know from Theorem 2.1 that the maximum value of the directional derivative is in the
direction of n̂. Thus
∂φ
|∇φ| = . (2.31)
∂ n̂
Combining equations (2.30) and (2.31) we obtain
∂φ
∇φ = n̂. (2.32)
∂ n̂
26

This geometric definition is coordinate independent and allows the computation of the gradient in any coordi-
nate system.

2.5.4 Summary
1. Gradient (geometric definition).
The gradient of the scalar field φ is the vector field ∇φ given by

∂φ
∇φ = n̂, (2.33)
∂n

∂φ
where n̂ is the unit normal to the surfaces φ = const in the direction of increasing φ and is the
∂n
directional derivative of φ in the n̂ direction.

2. Gradient (Cartesian definition).

If the scalar field φ is given in Cartesian coordinates by φ(x, y, z) then the gradient of φ in Cartesian
coordinates is
∂φ ∂φ ∂φ
∇φ = i+ j+ k. (2.34)
∂x ∂y ∂z

3. Directional Derivative.
The directional derivative of a scalar field φ in the direction of a vector v is defined by equation (2.22)
and can be calculated as

∂φ
= v̂ · ∇φ (2.35a)
∂v
∂φ ∂φ ∂φ
= v̂1 + v̂2 + v̂3 (in Cartesian coordinates). (2.35b)
∂x ∂y ∂z

In this module we will not use the gradient in other coordinate systems; if you take MATH2044, you will need
to use the gradient in polar and cylindrical coordinates (see auxiliary non-examinable notes).

Example 2.9 Let a general point be denoted r = xi + yj + zk, and let r = |r|. Consider the scalar field
φ(r) = |r| = r. Determine ∇φ.

Example 2.10 Calculate the equation of the tangent plane to the surface S given by x3 y − yz 2 + z 5 = 9 at
the point P given by (3, −1, 2).

2.6 Properties of the gradient

(1) Linearity: ∇(φ + ψ) = ∇φ + ∇ψ.
∂ ∂φ ∂ψ
Proof : Use Cartesian coordinates and the fact that (φ + ψ) = + .
∂x ∂x ∂x

(2) Leibniz 4 (or product) rule: ∇(φψ) = ψ∇φ + φ∇ψ.

∂ ∂φ ∂ψ
Proof : Use Cartesian coordinates and the fact that (φψ) = ψ +φ .
∂x ∂x ∂x
4
Gottfried Wilhelm Leibniz (1646-1716), distinguished German mathematician and philosopher (inter alia), credited as (indepen-
dent) co-inventor of calculus.
 27

2.7 The divergence of a vector field

In the previous section we saw how to calculate a vector field ∇φ by taking the derivative of a scalar field. In
this section we show how to calculate a scalar field by taking the derivative of a vector field F . We call this
derived scalar field the divergence of F and denote it divF or ∇ · F .
The divergence, being the dot product of two vectors, is coordinate independent. In this module we focus on its
expression in Cartesian coordinates. We can compute its expression in this coordinate system in two steps:
Step 1: First introduce the (linear) differential operator nabla (or gradient)
∂ ∂ ∂
∇= i+ j+ k. (2.36)
∂x ∂y ∂z
This definition of ∇ is quite general. For example, we can write the gradient of φ as the differential operator
nabla applied to the field φ, i.e.
 
∂ ∂ ∂ ∂φ ∂φ ∂φ
∇φ = i+ j+ k φ= i+ j+ k, (2.37)
∂x ∂y ∂z ∂x ∂y ∂z
which is indeed how we defined ∇φ in equation (2.26).
Step 2: Now let F be a vector field with components
F = F1 i + F2 j + F3 k . (2.38)
Then we may write its divergence as the dot product of the differential operator nabla, equation (2.36), with the
vector field F :
∂F1 ∂F2 ∂F3
divF = ∇ · F = + + . (2.39)
∂x ∂y ∂z
To summarise:
Definition 2.7 (Divergence in Cartesian coordinates) The divergence of a differentiable vector field F is a
scalar field indicated with ∇ · F or divF . Its value in Cartesian coordinates is
∂F1 ∂F2 ∂F3
∇·F = + + . (2.40)
∂x ∂y ∂z

Observations
1. The divergence is a linear operator, so that if F and G are two vector fields, then
∇ · (F + G) = ∇ · F + ∇ · G. (2.41)

2. Let φ be a scalar field and F a vector field then

∇ · (φF ) = (∇φ) · F + φ∇ · F . (2.42)
Proof - Use Cartesian coordinates to calculate both sides. This rule is a very plausible generalisation of
the product rule.
3. If F is constant then ∇ · F ≡ 0. The converse is not true.
Example 2.11 Compute the divergence for the special case where F is given by
F = φc
where φ is a scalar field and c is a constant vector.
Example 2.12 Show that the divergence of F = y 2 zi + xzj − y 2 k is everywhere zero.
Observation - Look at this example twice! For F = y 2 zi + xzj − y 2 k, not constant, ∇ · F = 0. This example
shows that divF = 0 does not imply that F is constant.
Why is this interesting? The divergence is a kind of vector generalisation of a derivative. Normally when a
derivative of something is zero, the something is constant. But not here!
28

2.8 The Laplacian

If we start with a scalar field φ we may first take the gradient to obtain ∇φ. This is a vector field so we may
take its divergence to obtain a scalar field, ∇ · (∇φ). We call this scalar field the Laplacian5 of φ.

Definition 2.8 The Laplacian of a twice differentiable scalar field φ is denoted ∇2 φ and defined by

∇2 φ = div(gradφ) = ∇ · (∇φ) .

It is quite straightforward to compute its expression in Cartesian coordinates. The gradient of φ in this coordi-
nate system is
∂φ ∂φ ∂φ
∇φ = i+ j+ k.
∂x ∂y ∂z
Taking its divergence we get
     
∂ ∂φ ∂ ∂φ ∂ ∂φ
∇ · (∇φ) = + +
∂x ∂x ∂y ∂y ∂z ∂z
2 2
∂ φ ∂ φ ∂ φ 2
= + 2 + 2.
∂x2 ∂y ∂z
Thus
∂2φ ∂2φ ∂2φ
∇2 φ = + 2 + 2. (2.43)
∂x2 ∂y ∂z
Comments
1. ∇2 is the scalar operator
∂2 ∂2 ∂2
 
+ + . (2.44)
∂x2 ∂y 2 ∂z 2
So ∇2 φ is scalar if φ is scalar, and ∇2 F is vector if F is vector.

2. The operator ∇2 , and the function ∇2 φ are both scalars. They are therefore independent of the frame of
reference. They can be calculated in any coordinate systems, e.g. spherical or cylindrical coordinates,
but in this module we will only use their expression in Cartesian coordinates, equations (2.44) and (2.43)
respectively.

Example 2.13 Let φ(x, y, z) = ex y 3 sin(z), calculate ∇2 φ.

2.9 The curl of a vector field

We have already considered a differential operator called the divergence that may be symbolically written
as “∇·”. In this section we consider a differential operator called the “curl” or “rotation” which may be
symbolically written as “∇×”. As its name suggests the “curl” measures the extent to which a point particle
rotates as it moves along the field lines of the vector field F .

Definition 2.9 Given a differentiable vector field F we define the curl of F as the vector field ∇ × F .

Note that this differential operator maps a vector field in another vector field. As usual, we limit our attention
to the expression of this operator in Cartesian coordinates.
We define the vector operator curl as the operator
 
∂ ∂ ∂
∇× = i +j +k × (2.45)
∂x ∂y ∂z
5
Pierre-Simon Laplace (1749-1827), French mathematician, physicist and astronomer. Also famous for Laplace Transforms, pre-
dictions in celestial mechanics and probability distributions.
 29

As stated before, this vector operator maps a vector function onto another vector function.
We use the determinant formula for the cross product, equation (1.5), to compute the curl of F = F1 i +
F2 j + F3 k as  
 i j k 

 ∂ ∂ ∂ 
∇ × F =  , (2.46)
 ∂x ∂y ∂z 
 F1 F2 F3 

where the expansion is always along the first row and the operators in the second row act on the vectors in the
third row. Evaluating this, we obtain
     
∂F3 ∂F2 ∂F1 ∂F3 ∂F2 ∂F1
curlF = ∇ × F = − i+ − j+ − k. (2.47)
∂y ∂z ∂z ∂x ∂x ∂y

Example 2.14 Calculate ∇ × F for F = xy 2 i + ez j + y sin(x)ez k.

Example 2.15 Let φ = x2 yz 3 . Show that ∇ × (∇φ) = 0.

In this example we find that ∇ × ∇φ = 0. In fact this is not just chance. It is a general result, and works for
all suitably continuous scalar fields φ(r), as we now show.

Theorem 2.3 Let φ be a C 2 (i.e.twice differentiable with continuous second derivative) scalar field. Then:

curl(gradφ) = ∇ × ∇φ = 0.

Proof: We use equation (2.26) to write the gradient of φ:

∂φ ∂φ ∂φ
∇φ = i+ j+ k.
∂x ∂y ∂z

Therefore
 
 i j k 
 ∂ ∂ ∂ 

∇ × ∇φ =  ∂x ∂y ∂z 
 
 ∂φ ∂φ ∂φ 
 
∂y ∂z 
 ∂x2

∂2φ
 2
∂2φ
 2
∂2φ
  
∂ φ ∂ φ ∂ φ
= − i+ − j+ − k
∂y∂z ∂z∂y ∂x∂z ∂z∂x ∂x∂y ∂y∂x
= 0,

where we have used the C 2 condition to neglect the order of differentiation in the mixed derivatives, so that,
∂2φ ∂2φ
for example, = .
∂y∂z ∂z∂y

Example 2.16 Let F = x2 ey i + x ln zj + (x + z)k. Calculate G = ∇ × F and show that ∇ · G =

∇ · (∇ × F ) = 0.

In this example ∇ · (∇ × F ) = 0. Again this is not just chance but is a general result:

Theorem 2.4 Let F be a C 2 vector field, then

div(curlF ) = ∇ · (∇ × F ) = 0 .
30

Proof: Let F = F1 i + F2 j + F3 k, then

 
 i j k 

 ∂ ∂ ∂ 
G = ∇ × F = 
 ∂x ∂y ∂z 

 F1 F2 F3 
     
∂F3 ∂F2 ∂F1 ∂F3 ∂F2 ∂F1
= − i+ − j+ − k.
∂y ∂z ∂z ∂x ∂x ∂y

Now ∇ · G is given by

∂G1 ∂G2 ∂G3

∇·G= + +
∂x ∂y ∂z
2
∂ F3 2
∂ F2 ∂ 2 F1 ∂ 2 F3 ∂ 2 F2 ∂ 2 F1
= − + − + −
∂x∂y ∂x∂z ∂y∂z ∂y∂x ∂z∂x ∂z∂y
= 0,

where we have again used the C 2 condition to ensure that the mixed second derivatives commute.
Observation - One has to be careful with replacing a vector by an operator and just carrying on regardless. For
example, note that although a × b = −b × a, it is emphatically not the case that ∇ × F = −F × ∇. Not
only are they not equal, they are not even the same kind of object. ∇ × F is a vector field, but F × ∇ is a
differential operator, as we can see by substituting in and working through:
 
 i j k 

 F F2 F3 
F × ∇ =  1
 ∂ ∂ ∂ 
 ∂x ∂y ∂z 
     
∂ ∂ ∂ ∂ ∂ ∂
= i F2 − F3 + j F3 − F1 + k F1 − F2 .
∂z ∂y ∂x ∂z ∂y ∂x

Exercise: Show that (F × ∇)φ = F × (∇φ).

2.10 Vector Identities

In this section we collect together some useful vector differential identities. They are mostly generalisations of
algebraic vector identities and the differentiation product rule. Learn them and be able to prove them. Some of
these will turn up in example sheets or assignments, but some may also occur in an exam.

∇(φψ) = ψ∇φ + φ∇ψ (2.48a)

∇ · (φF ) = ∇φ · F + φ∇ · F (2.48b)
∇ × (φF ) = ∇φ × F + φ∇ × F (2.48c)
∇ · (F × G) = G · ∇ × F − F · ∇ × G (2.48d)
∇ × (∇φ) = 0 (2.48e)
∇ · (∇ × F ) = 0 (2.48f)
2
∇ · (∇φ) = ∇ φ (2.48g)
2
∇ × (∇ × F ) = ∇(∇ · F ) − ∇ F . (2.48h)

In equation (2.48h) we have used the notation ∇2 F = ∇2 F1 i + ∇2 F2 j + ∇2 F3 k.

Chapter 3

Integrals of vector and scalar fields

3.1 Line integrals

Let F (r) be a (smooth) vector field and let γ(t) for t0 ≤ t ≤ t1 be a curve parameterised by t. Let r(t) denote
the position vector of the point on the curve given by γ(t).

3.1.1 Definition of line integral of a vector field

In this section we give a definition of Z
F · dr (3.1)
γ
the line integral of F along the curve γ. We will see below that this integral depends upon the curve γ, but not
upon the parameterisation of the curve - only upon the set of points on the curve.
In the usual way when considering integrals we break the curve γ up into a series of N → ∞ small straight
line segments (see figure 3.1). Now let δr k denote the line segment joining r k to r k+1 , and let F k = F (r k ).
Then we define the line integral by
Z N
X
F · dr = lim F k · δr k . (3.2)
γ N →∞
k=1

This definition is consistent with usual definitions of what is meant by an integral in analysis. However this
definition is not very useful when it comes to calculations. In order to calculate a line integral we convert it to
an ordinary 1-dimensional integral.

r(t1 )
y γ

r(t0)
r(tk)

a b r(tk+1)
r(t N )
t0 t1 t2 tk tN
x

Figure 3.1: The curve γ is approximated by a piecewise linear curve in which the k-th segment joins the points
r(tk ) and r(tk+1 ). As the partition of the interval [a, b] becomes smaller and smaller, i.e. as δt ≡ tk+1 −tk → 0,
the piecewise linear curve becomes a better and better approximation of γ: in the limit δt → 0 its length is
defined as the length of γ.

31
32

We start by looking at the parameterised curve

γ : t 7→ r(t), a ≤ t ≤ b.

and approximate it with a piecewise linear curve (see figure 3.1) in which the k-th segment joins the points
r(tk ) and r(tk+1 ). Using Taylor’s theorem we can write each segment as
dr
(tk )δt + O (δt)2 .
 
δr k ≡ r k+1 − r k =
dt
We also look at the value of the vector field along the curve

F = F [r(t)] = F (t),

so that the vector field becomes a function of the parameter t. Combining these the term in the sum on the right
hand side of equation (3.2) becomes
N
dr
Z X
F · dr = lim Fk · (tk )δt, (3.3)
γ N →∞ dt
k=1
 
dr
(δt)2
 
where we have not written the terms O for ease of notation. Notice that F [r(t)] · is simply some
dt
function f (t) of t. Hence, the right hand side of equation (3.3) is now a function of t only and its limit is an
ordinary integral over t of f (t):
N b
dr dr
X Z
lim Fk · (tk )δt = F [r(t)] · (t) dt. (3.4)
N →∞ dt a dt
k=1

Combining equation (3.4) with (3.3) we now have an “easy” way to compute the line integral of a vector field
over a curve, namely
Z b
dr
Z
F · dr = F [r(t)] · dt. (3.5)
γ a dt

3.1.2 Parameterisation independence

Note that the parameterisation r(t) is not unique. We could have used a completely different parameterisation
r̃(s), in which every point on γ is associated with a different number s 6= t. The line integral is independent of
the parameterisation.

Example 3.1 Verify that the following two functions both represent the section of the unit circle in the first
quadrant traversed counter-clockwise:

r 1 (t) = cos(t)i + sin(t)j 0 ≤ t ≤ π/2,

r 1 (t) = cos(2t)i + sin(2t)j 0 ≤ t ≤ π/4.

Hence verify that the line integral of the vector field

F = −yi + xj,

over the quarter circle is the same for the two parameterisations.

The result of this example is always true, as proved in the following theorem:

Theorem 3.1 The formula (3.5) for the line integral of a vector field F (x) over a piecewise differentiable
curve γ is independent of the parameterisation chosen for γ.
 33

2
γ2

z
1
γ1

0
0 0
y 0.5 1 2
1
x
Figure 3.2: Geometry for Examples 3.2 (γ1 , solid curve) and 3.3 (γ2 , dashed line). The two curves join the
origin (blue dot) to the point (2, 1, 3) (red dot).

Proof: We assume that the curve γ is parameterised by a function r(t), with a ≤ t ≤ b, and a function r̃(s),
with c ≤ s ≤ d. Let f : R → R be a monotonic increasing (smooth) map from [c, d] to [a, b] such that f (c) = a
and f (d) = b. Therefore, r̃(s) = r[f (s)] for c ≤ s ≤ d. In terms of the r̃(s)-parameterisation the line integral
is given by
Z d
dr̃
Z
F · dr = F [r̃(s)] · ds . (3.6)
γ s=c ds
However by the chain rule
dr̃ dr dt
= .
ds dt ds
So changing variable in the integral (3.6) we get
d
dr̃
Z Z
F · dr = F [r̃(s)] · ds
γ s=c ds
d
dr dt
Z
= F [r̃(s)] ·ds
s=c dt ds
Z b
dr
= F [r(t)] · dt ,
t=a dt

which agrees with equation (3.5). Hence the line integral is independent of the parameterisation of the curve as
claimed.

3.1.3 Examples of line integrals

Example 3.2 Let
F = 2xyi + (x2 − z 2 )j − 3xz 2 k . (3.7)
Z
Evaluate the line integral F · dr, where γ1 is the curve given by
γ1

r(t) = 2ti + t3 j + 3t2 k, 0≤t≤1 (3.8)

going from the point (0, 0, 0) to the point (2, 1, 3) (see figure 3.2).
34

Example 3.3 Calculate the line integral for the vector field F in equation (3.7) of the previous example, but
this time along the curve γ2 , the straight line connecting (0, 0, 0) to (2, 1, 3) (see figure 3.2).

These examples show the important fact that in general the value of the line integral depends upon the path
taken, not just the end points.

Example 3.4 Let F = ∇φ where φ = xy 2 z. Calculate the line integral of F for both of the curves used in
examples 3.2 and 3.3, respectively.

In this example the two line integrals give the same answer, even though the routes are different. In fact this is
true for the line integral of any vector field that is the form ∇φ.

Theorem 3.2 If F is a gradient vector field so that F = ∇φ, for some C 1 scalar field φ, then the line integral
Z
F · dr only depends upon the end points A and B of γ, but not on the path taken. In fact
γ

Z
∇φ · dr = φ(r B ) − φ(r A ) ,
γ

where φ(r A ) and φ(r B ) are the values of φ at the start and end points of the curve γ.

Proof: Let γ be parameterised by r(t), t0 ≤ t ≤ t1 then

 t1 
dr
Z Z
∇φ · dr = ∇φ[r(t)] · dt .
γ t=t0 dt

We compute the various terms in this integral. The gradient of φ is

∂φ ∂φ ∂φ
∇φ = i+ j+ k.
∂x ∂y ∂z

If we write
r(t) = x(t)i + y(t)j + z(t)k,

with t0 ≤ t ≤ t1 then
dr dx dy dz
= i+ j+ k.
dt dt dt dt
Hence
dr ∂φ dx ∂φ dy ∂φ dz dφ(r(t))
∇φ · = + + = ,
dt ∂x dt ∂y dt ∂z dt dt
where the last equality follows form the chain rule. Thus

t1
dφ(r(t))
Z Z
∇φ · dr = dt
γ t=t0 dt
= [φ(r(t))]tt=t
1
0

= φ[r(t1 )] − φ[r(t0 )]
= φ(r B ) − φ(r A ),

which proves the proposition.

 35

3.2 Conservative vector fields

Why is the result of the previous section significant? One important physics application of line integrals is to
find out the work done by a force. For a constant force F moving in a straight line from A to B, the work done
is “the force times the distance moved in the direction of the force”. The mathematical formula for this is

W = F · r,
−−→
where r = AB. However if the force and path are changing, we have to divide the path into small sections,
and then take the limit
XN Z
W = lim F k δr k = F (r) · dr
N →∞ γ
k=1

in order to obtain the formula for the work done by the vector field F (r) in moving from A to B along the path
γ.
If we can find a scalar field such that
F = −∇φ ,
[Note the minus sign!] then we say that φ is a potential for the force F and the change in φ simply represents
the change in potential energy. In such a situation any loss of potential energy results in a gain in kinetic, i.e.
motion, energy (and vice-versa) so that the total energy is conserved. For this reason a gradient vector field is
often called a conservative vector field, and we will use this name from now on in this module. In section 4.4
of these notes we will look at conditions which guarantee that a vector field is conservative.

3.3 Surfaces
We start by giving a number of different definitions of a 2-dimensional surface in three dimensional space R3 .
There are essentially three different ways of specifying a surface:

1. As the graph of a function of 2 variables, z = f (x, y);

2. As a level surface of a function of 3 variables F (x, y, z) = c;

3. As a parameterised surface (s, t) 7→ [x(s, t), y(s, t), z(s, t)].

For the purpose of calculating surface integrals we will mostly use definition (3) and think of surfaces as
(smooth) maps
D ⊆ R2 → R3 : (s, t) 7→ r(s, t), (3.9)
where D ⊂ R2 is some 2-dimensional set which specifies the region in which the parameters s and t lie. In
many of the applications this will simply involve s0 ≤ s ≤ s1 and t0 ≤ t ≤ t1 so that D will be given by the
rectangle [s0 , s1 ] × [t0 , t1 ].
If we draw a picture of this map we see that the lines s = const. and t = const. are simply the coordinate
lines on the surface (see figure 3.3).

Example 3.5 Show that the curved surface of a cylinder of radius a and height h can be parameterised as

a cos(s)i + a sin(s)j + tk, (3.10)

with −π ≤ s < π and 0 ≤ t ≤ h.

Example 3.6 Show that a sphere of radius a with centre the origin can be parameterised as

r(s, t) = a cos(s) sin(t)i + a sin(s) sin(t)j + a cos(t)k, (3.11)

with −π ≤ s < π and 0 ≤ t ≤ π.

36

t z

r(s, t)

s y

Figure 3.3: Surface, parameterised by lines s constant (dashed lines) and t constant (dotted lines).

z z
z
1 r(s, t)
t
t = t0
r(s, t)
t s y y

s x x
y s = s0

Figure 3.4: Surface of a cylinder Figure 3.5: Surface of a sphere (see Example 3.6). The left panel
(see Example 3.5). shows the geometrical meaning of the two parameters s and t.
The right panel show the lines of constant t (dashed red) and con-
stant s (dotted red).

∂r ∂r
× δsδt
∂s ∂t
∂r
δt
∂t

r(s0 , t)
∂r
δs dA
∂s

r(s0 + δs, t)
r(s, t0 ) r(s, t0 + δt)

Figure 3.6: The partial derivatives are each tangent to one of the coordinate lines (constant s dotted, constant t
dashed). Their cross product is perpendicular to the surface and its modulus is the area dA of the patch.
 37

3.3.1 The normal to a surface

The parameterisation of a surface allows us to calculate an expression for the normal to a general surface.
∂r ∂r
The vector is tangent to the curve t =const, while the vector is tangent to the curve s =const, as shown
∂s ∂t
in figure 3.6. A normal to the surface is therefore given by
∂r ∂r
n= × . (3.12)
∂s ∂t
Note that there are infinitely many normal vectors to each point of the surface. If we specify the length of the
vector, e.g. we request the unit normal, there are only two vectors left, pointing in opposite directions. If the
surface is closed, the unit normal is often taken outwards from the inside of the volume enclosed.

Example 3.7 Compute the normal to the curved surface of a cylinder of radius a and height h,

r(s, t) = a cos(s)i + a sin(s)j + tk, (3.13)

with 0 ≤ s ≤ 2π and 0 ≤ t ≤ h.

Example 3.8 Compute the unit normal to a sphere of radius a and centre the origin,

r(s, t) = a cos(s) sin(t)i + a sin(s) sin(t)j + a cos(t)k, (3.14)

with 0 ≤ s ≤ 2π and 0 ≤ t ≤ π.

3.4 Surface integrals

3.4.1 Area of a surface
In order to calculate surface integrals we first need to find an expression for the element of surface area dA. To
do this we look again at the picture of the coordinate lines on a surface (see figure 3.6). Suppose we start at the
point r(s, t) and move to the nearby point r(s + δs, t). Then the vector connecting these two points is

δr 1 = r(s + δs, t) − r(s, t) .

Now by Taylor’s theorem

∂r
r(s + δs, t) = r(s, t) + δs (s, t) + O[(δs)2 ] .
∂s
Hence
∂r
δr 1 = δs (s, t) + O[(δs)2 ] .
∂s
Similarly if we move in the t direction to the nearby point r(s, t+δt) then the vector measuring the displacement
from r(s, t) is
δr 2 = r(s, t + δt) − r(s, t) .
Again using Taylor’s theorem we have
∂r
r(s, t + δt) = r(s, t) + δt (s, t) + O[(δt)2 ]
∂t
and hence
∂r
δr 2 = δt (s, t) + O[(δt)2 ] .
∂t
Thus the area of the parallelogram with sides δr 1 and δr 2 is
 
 ∂r ∂r 
δA = |δr 1 × δr 2 | ≈   ×  δsδt .
∂s ∂t 
38

In the limit of finer and finer coverings of the surface, i.e. δt, δs → 0, we get
 
 ∂r ∂r 
dA =  ×  ds dt . (3.15)
∂s ∂t 
To summarise:
Definition 3.1 (Area of a surface) The area of a differentiable surface S parametrised by
r(s, t) : D ⊆ R2 → R3 , (3.16)
is defined as Z  
 ∂r ∂r 
Z
Area of S ≡ dA =  ×  ds dt . (3.17)
D ∂s ∂t 

S

Example 3.9 Use the expression (3.15) for the area element to calculate the area of the curved surface of a
cylinder of radius a and height h.

Observation - If we discuss surface integration in the abstract, we generally use one integration sign. But all
concrete examples involve two integrations, one over s and the other over t. So in concrete examples, you will
see two integration signs, indicating the limits of the integrations.

Example 3.10 Use the expression (3.15) for the area element to calculate the area of a sphere of radius a.

3.4.2 Flux integrals

Flux integrals are a very common concept in scientific applications of vector analysis. A flux integral measures
the “flow” of a vector field F through a surface S.
Definition 3.2 (Flux integral) Given a differentiable surface S with unit normal n̂ and a vector field F con-
tinuous on S, we define the flux of F through S in the direction of
Z
F · n̂ dA , (3.18)
S

Observation - If we think of F = ρv, where ρ is the mass volume density of a fluid (in kg/m3 ) with velocity
v (in m/s), then the flux integral measures the net rate of flow of the fluid through the surface S in kg/s.
As with line integrals the method of calculating flux integrals is to use the parameterisation of the surface to
convert them into ordinary integrals. We introduce the vector element of area
dS = n̂ dA (3.19)
and we write the flux integral (3.18) as
Z Z
F · n̂ dA = F · dS . (3.20)
S S
Then dS points in the same direction as n̂ and has modulus |dS| = dA. From this we see that (see also
comment 3 below)  
∂r ∂r
dS = × ds dt . (3.21)
∂s ∂t
Using this formula for dS we may write the flux integral (3.20) as
Z x 
∂r ∂r

F · dS = F [r(s, t)] · × ds dt , (3.22)
S ∂s ∂t
D

where the integrand is a function of s and t: equation (3.22) thus transforms the flux integral into an ordinary
two-dimensional integral. The sign is chosen so that the the cross product is parallel to n̂.
Observations:
 39

z
2

0 0.5
-0.5 0
0 -0.5
0.5
1 y
x

Figure 3.7: Surface of the helicoid defined in equation (3.23).

1. Note the single integration sign in front of the dS, but the double integration in front of ds dt. Some
authors always use double integration signs.
2. The formula for the flux integral depends upon the parameterisation. However one can show, as with the
case of the line integral, that the final answer is independent of the parameterisation chosen. We do not
give the details of the calculation here, but not that the essential point is that if we calculate dS using a
different parameterisation, (s̃, t̃), it differs from the original one by being multiplied by the the Jacobian
determinant of the transformation from (s, t) to (s̃, t̃). This is exactly the factor one needs when going
from ds dt to ds̃dt̃.
3. Strictly speaking, as we have defined it, there is a sign ambiguity in the definition, and the direction of
dS is not uniquely defined. We need some extra information to do this, such as dS goes from “left to
right”, or “from inside the material to outside”’, before we can decide what the sign is. In particular, this
means that in equation (3.21) we may need to swap the order of the cross-product in order for the normal
vector to point in the “correct” direction.

Example 3.11 Let S be the parameterised surface given by

r(s, t) = s cos(t)i + s sin(t)j + tk, (3.23)

with 0 ≤ s ≤ 1 and 0 ≤ t ≤ 2π (this is a twisted ribbon or “helicoid”, see figure 3.7). Let F (r) be the vector
field given by
F = xi + yj + (z − 2y)k.
Evaluate the flux integral of F through S with unit normal in the positive z-direction.

Flux integrals for surfaces expressed in explicit form

The general expression (3.21) for the surface element can be simplified considerably in the special case that
the surface is the graph of a function z = f (x, y), i.e. the surface is expressed in explicit form. To calculate the
flux integral in this case we take x and y themselves to be the parameters used to describe the surface. Hence
the parameterised surface is given by

r(x, y) = xi + yj + f (x, y)k, (x, y) ∈ D

Then
∂r ∂f
=i+ k,
∂x ∂x
∂r ∂f
=j+ k.
∂y ∂y
40

Hence
 
∂r ∂r
dS = × dx dy
∂x ∂y
 
 i j k 
 
 1 0 ∂f 
 
=  ∂x  dx dy
 0 1 ∂f 
 
 ∂y 
 
∂f ∂f
= − i− j + k dx dy.
∂x ∂y

Thus for the case of a surface which is the graph of a function we have the formula
Z x 
∂f ∂f

F · dS = F (x, y, f (x, y)) · − i − j + k dx dy .
S ∂x ∂y
D

Use symmetries
We end this section by making a very important point. The methods outlined in this section give a calculation
scheme. But, in fact, it can often save time to think a bit about the geometry of a flux integral before going
ahead and mechanically doing the calculation. There can sometimes be a shortcut due to symmetries of the
problem, as we can see in the next example.

Example 3.12 Let F = xi + yj + zk, and S the surface of the sphere x2 + y 2 + z 2 = a2 . Calculate the flux
integral of F across the surface in the outward direction.
Chapter 4

Integral Theorems

In this section we look at the relationship between line integrals, flux integrals and volume integrals using grad,
div and curl.

4.1 Stokes’ Theorem

We start by looking at the relationship between the line integral round a closed loop and the flux flowing through
the loop. This theorem goes under the name of Stokes’ theorem1 .

Theorem 4.1 (Stokes) Let S be a bounded, piecewise smooth, orientable, surface in R3 , with boundary ∂S,
which consists of a finite number of piecewise C 1 simple closed curves. Let F be a C 1 vector field whose
domain includes S. Then Z I
(∇ × F ) · dS = F · dr .
S ∂S

Observations

1. We will prove a restricted version of this theorem in section 4.2. The proof of Stokes’ theorem is in the
(optional) section 4.3.

2. The surface does not have to be flat.

1
Sir George Gabriel Stokes (1819-1903), Irish applied mathematician, Lucasian Professor of Mathematics at the University of
Cambridge, known particularly for his work in fluids. In addition to appearing in his theorem, his name is preserved in Stokes’ Law
(the terminal velocity of a falling particle in a viscous fluid medium), the Navier-Stokes equations (the governing equations in fluid
dynamics), and various Stokes phenomena in optics.

n̂
n̂ ¶S2
¶S3

¶S ¶S1

Figure 4.1: Relation between orientation of the Figure 4.2: Generalisation of the example in
unit normal n̂ to an orientable surface S and the figure 4.1 to the case of a non-simply connected
direction in which its boundary ∂S is traversed, surface. The inside and outside boundaries are
according to the right hand rule. traversed in opposite directions.

41
42

(0,0,9)

z=9- x2-y2

(0,3,0)

(-3,0,0) (3,0,0)

(0,-3,0)
S: x2+y2=9

Figure 4.3: Graph of the surface S used in Example 4.1.

I
3. The symbol indicates that the line integral is round a closed curve. This line integral round ∂S is
computed in a counter-clockwise direction relative to the normal n to S used to define dS, as shown in
the diagram in figure 4.1. If the opposite normal is chosen both integrals therefore change sign.

4. In figure 4.1 the boundary of S consists of just one component. However the theorem allows for the
boundary to consist of several components (e.g. three as shown in figure 4.2). In such a case the line
integral is the sum of the line integrals for the various components, so in the example above
I I I I
F · dr = F · dr + F · dr + F · dr , (4.1)
∂S ∂S1 ∂S2 ∂S3

where the direction of integration round the outer boundary ∂S1 counter-clockwise relative to the normal
n to S. Note that for inside boundaries, the direction of integration is reversed.

5. Stokes’ Theorem contains some fine print concerning the nature of:

(a) the surface, which has to be piecewise smooth and orientable. Orientable means that one can choose
a consistent choice of normal vector at every point on the surface; a counterexample is the famous
Möbius strip. Piecewise smooth means that the surface consists of a finite number of pieces each
of which can be represented in parametric form r(s, t) with r(s, t) a continuously differentiable
function.
(b) the boundary, which has to possess a finite number of piecewise C 1 simple closed curves. Again
piecewise C 1 means that the boundary consists of a finite number of curves, each one of which
can be parameterised by a differentiable function. A closed curve is one for which the start and
end points are the same. A simple curve is one which does not cross itself; a figure of eight is an
example of a curve which is not simple.
(c) the vector field F , which has to be C 1 on a domain that includes S, i.e. the vector function is
continuously differentiable on the surface.

You can fine more detailed explanations of the precise meaning of these terms in textbooks. Luckily,
in most physical applications (and Stokes was really a physicist in modern language), the conditions for
Stokes theorem are satisfied.

Example 4.1 Let S be the paraboloid

z = 9 − x2 − y 2 , z ≥ 0, (4.2)

see figure 4.3. Verify Stokes’ theorem for the vector field

F = (2z − y)i + (x + z)j + (3x − 2y)k. (4.3)

 43

Example 4.2 Let S̃ be the disc x2 + y 2 ≤ 9, z = 0. Calculate

Z
(∇ × F ) · dS
S̃

where F is as in Example 4.1, see equation (4.3), and n̂ is the upward unit normal.

The method used to solve this example is often useful, so we state it in the general case as a consequence of
Stokes’ theorem.

Corollary 4.1 (Surface deformation theorem) Let S and S̃ be two bounded, piecewise smooth, orientable,
surfaces in R3 , with the same boundary (consisting of a finite number of piecewise C 1 simple closed curves).
Let F be a C 1 vector field whose domain includes both S and S̃. Then
Z Z
(∇ × F ) · dS = (∇ × F ) · dS.
S S̃

4.2 Green’s theorem (Stokes’ theorem in the plane)

Stokes’ theorem states that Z I
(∇ × F ) · dS = F · dr . (4.4)
S ∂S
In this section we look at this relation in the case that the vector field F depends only upon x and y and both it
and the surface S lie in the (x, y)-plane. Since F lies in the (x, y)-plane and only depends upon x and y, then

F = F1 (x, y)i + F2 (x, y)j (4.5)

Then
 

 i j k 
 ∂ ∂ ∂ 
∇ × F = 
 ∂x ∂y ∂z 
 F1 (x, y) F2 (x, y) 0 
  (4.6)
∂F2 ∂F1
= 0i + 0j + − k
∂x ∂y
 
∂F2 ∂F1
= − k.
∂x ∂y
Also for a surface lying in the (x, y)-plane

dS = n̂dA = k dx dy . (4.7)

Hence Z  
∂F2 ∂F1
Z
(∇ × F ) · dS = − dx dy .
S S ∂x ∂y
On the other hand, if we only consider points in the (x, y)-plane then

r = xi + yj =⇒ dr = dxi + dyj,

so that I I
F · dr = (F1 dx + F2 dy) .
∂S ∂S
Hence in the xy-plane Stokes’ theorem becomes
Z  
∂F2 ∂F1
I
− dx dy = (F1 dx + F2 dy) .
S ∂x ∂y ∂S
44

(x, y + )
y
C+

x
∂S = C + − C −
C−

(x, y − )

Figure 4.4: Integration loop used to prove Green’s theorem. The boundary ∂S of the flat region S can be into
two (or more) curves, each of which is the graph of a function of the x coordinate. In the case of the contour
in this figure, it can be split into two parts: the top is the graph of the function y + (x), while the bottom is the
graph of the function y − (x).

This result is known as Green’s theorem 2 :

Theorem 4.2 (Green’s theorem) Let S be a bounded region in R2 , with boundary ∂S which consists of a
finite number of piecewise C 1 simple closed curves. Let F be a C 1 vector field in R2 whose domain includes
S. Then
Z  
∂F2 ∂F1
I
− dx dy = (F1 dx + F2 dy) .
S ∂x ∂y ∂S

where the line integral is traversed according to the right hand rule.

Note - The main interest in Green’s theorem is that it can be proved quite easily (see next section) and that it
can be used to prove Stoke’s theorem. How to do this is sketched in the (optional) section 4.3.
Proof We need to show that:
Z  
∂F2 ∂F1
I
− dx dy = (F1 dx + F2 dy) .
S ∂x ∂y ∂S

We assume that we can split the boundary ∂S into two curves; in the example in figure 4.4, C + is the upper
curve and C − the lower curve. Let (x, y + (x)) denote points on the graph of the upper curve and (x, y − (x))
denote points on the graph of the lower curve. Then if we traverse the closed curve in a clockwise direction we
move from left to right on the bottom curve and from right to left on the upper curve. Hence

Z Z b
F1 dx = F1 [x, y − (x)] dx
ZC Zx=a
−
a
F1 dx = F1 [x, y + (x)] dx
C+ x=b
Z b
=− F1 [x, y + (x)] dx.
x=a

2
George Green (1793-1841) was a self-taught mathematician, who was originally a miller in Nottinghamshire, before attending
Cambridge University at the age of 40. Green’s theorem predates Stokes’s theorem, which from this point of view can be regarded as a
generalisation of Green’s theorem.
 45

Thus
I Z Z
F1 dx = F1 dx + F1 dx
C C− C+
Z b
F1 [x, y + (x)] − F1 [x, y − (x)] dx

=−
x=a
Z b
y + (x)
=− [F1 (x, y)]y=y− (x) dx
x=a
Z "Z y+ (x) #
∂F1
=− dy dx.
y=y − (x) ∂y

Hence
∂F1
I Z
F1 dx = − dx dy.
C S ∂y
Likewise
∂F2
I Z
F2 dy = dx dy.
C S ∂x
Thus Z  
∂F2 ∂F1
I I
F1 dx + F2 dy = − dx dy,
C C S ∂x ∂y
as required.

4.3 Proof of Stokes’ theorem (optional)

It is possible to derive Stokes’ theorem from Green’s theorem. This follows from the fact that if one parame-
terises the surface S in Stokes’ theorem in terms of s and t by r(s, t), with (s, t) ∈ D, then one finds (using
either some techniques from index notation, namely the contracted ε-identity, or brute force) that
Z  
∂Ft ∂Fs
Z
(∇ × F ) · dS = − ds dt,
S D ∂s ∂t
∂r ∂r
where Fs = F · and Ft = F · . It then follows from Green’s theorem in the (s, t)-plane that
∂s ∂t
Z  
∂Ft ∂Fs
I
− dsdt = (Fs ds + Ft dt).
D ∂s ∂t ∂D
On the other hand,
∂r ∂r
dr = ds + dt,
∂s ∂t
so that F · dr = Fr ds + Ft dt and, hence,
I I
(Fs ds + Ft dt) = F · dr .
∂D ∂S
Putting all this together we get
Z  
∂Ft ∂Fs
Z
(∇ × F ) · dS = − du dv
S D ∂s ∂t
I
= (Fs ds + Ft dt) (by Green’s theorem)
I∂D
= F · dr.
∂S
Hence we have used Green’s theorem in the st-plane to prove Stokes’ theorem:
Z I
(∇ × F ) · dS = F · dr .
S ∂S
46

B
γ1
γ2

γ3
γ4
A

Figure 4.5: Four different paths that connect two points A and B. Note that all curves are piecewise differen-
tiable and oriented, but they need not be smooth: in this example γ4 is not smooth as it has a corner.

4.4 Conservative Vector Fields

4.4.1 Definition and properties
In section 3.2 we introduced the concept of conservative vector field as a vector field that can be written as the
gradient of a (differentiable) scalar field. In this section we use Stokes’ theorem to identify simple criteria to
determine if a vector field is conservative. In doing this we also prove a variety of properties of conservative
vector fields. We start by defining a vector field with path-independent integrals; we will then show that such
fields are conservative.

Definition 4.1 A continuous vector field F has path-independent line integrals in the region U if
Z Z
F · dr = F · dr (4.8)
γ1 γ2

for any two curves γ1 and γ2 which lie in the set U and both start at the same point A, and end at both end at
the same point B (see figure 4.5). The curves need to be piecewise differentiable and oriented.

Theorem 4.3 Let F be a continuous vector field. Then F has path-independent integrals in U if and only if
I
F · dr = 0
γ

for all closed, simple and piecewise differentiable curves γ in U .

Proof: Suppose that F is a path-independent vector field and γ is a closed curve. Then we may split it up into
two curves γ1 and γ2 that start at A and end at B. Then γ is just γ1 followed by −γ2 (i.e. γ2 traversed in the
opposite direction). Hence I Z Z
F · dr = F · dr − F · dr = 0 .
γ γ1 γ2
Conversely suppose the line integral round any closed loop vanishes and γ1 and γ2 are two curves that start and
finish at the same point. Then γ1 followed by −γ2 (i.e. γ2 traversed in the opposite direction) is just γ so that
Z Z I Z Z
F · dr − F · dr = F · dr = 0 =⇒ F · dr = F · dr .
γ1 γ2 γ γ1 γ2

We now give a theorem that relates path-independent vector fields to conservative (or gradient) vector fields.

Theorem 4.4 Let F be a continuous vector field in a region U Then there exists a differentiable scalar field φ
such that F = ∇φ if and only if F has path independent integrals in U .
 47

Proof:
F = ∇φ ⇒ path-independent integral
We have already shown that if F = ∇φ then
Z Z
F · dr = (∇φ) · dr = φ(r B ) − φ(r A ),
γAB γAB

and hence the integral is independent of the path.

Path-independent integral ⇒ ∃φ such that F = ∇φ.
If F has path independent integrals then we define a scalar field φ by
Z
φ(r) = F · dr , (4.9)
γ

where γ is some path from some (arbitrary) fixed point A to the point P with position vector r. Note that
because the vector field F has path independent integrals this is well defined without having to specify the path
γ.
We now show by direct computation that
∂φ ∂φ ∂φ
= F1 , = F2 , = F3 , i.e. ∇φ = F .
∂x ∂y ∂z
We start from equation (4.9), repeated here for convenience:
Z
φ(x, y, z) = F · dr .
γ

Then Z Z
φ(x + h, y, z) = F · dr + F · dr ,
γ γ1

where γ1 is a straight line parallel to the x-axis going from x to x + h. Hence

Z x+h
φ(x + h, y, z) − φ(x, y, z) = F · dr
x
Z x+h
= F1 dx (since dr = i dx on γ1 )
x
= hF1 (x, y, z) + O(h2 ) .

Hence  
∂φ φ(x + h, y, z) − φ(x, y, z)
= lim = lim [F1 (x, y, z) + O(h)] = F1 (x, y, z) .
∂x h→0 h h→0

Similarly
∂φ ∂φ
= F2 , and = F3 .
∂y ∂z

Observation - Different choices of the arbitrary initial point A simply change φ by a constant. But this has no
effect on ∇φ.
This theorem tells us that a conservative vector field is equivalent to one with path-independent integrals.
Unfortunately this condition is not very easy to check as we need to know it is true for any possible path! The
next theorem, however, gives a more useful criterion, which is easy to check.

Theorem 4.5 Let F be a differentiable vector field defined on a simply connected region U in R3 . Then there
exists a C 2 scalar field φ such that F = ∇φ iff (i.e. if and only if) ∇ × F = 0.
48

Proof: We know from the vector identity (2.48e) that ∇ × (∇φ) = 0, so if F = ∇φ then we know that
∇ × F = 0.
On the other hand, if ∇ × F = 0 then we know from Stokes’ theorem that
Z I
0= (∇ × F ) · dS = F · dr,
S ∂S

where S is some spanning surface with boundary ∂S. Note that such a spanning surface S exists precisely
because of the simply-connected condition. Hence by Theorem 4.3 F has path-independent integral, and thus,
by Theorem 4.3, F = ∇φ, as required.
Observation - The fine print in this theorem is again important to note. In particular the simply connected
condition (which means we can continuously shrink any closed path down to a point) is required - see the
exercise sheets for a counterexample to this theorem when U is not simply connected. Recall that C 2 means
that second derivatives are well-defined.

4.4.2 Finding potentials for conservative vector fields

In the previous section we showed how to test for a conservative vector field in this section we show how one
can actually find a potential for the vector field.

Example 4.3 Let F = [2xy + cos(2y)]i + [x2 + 2y − 2x sin(2y)]j. Show that F is conservative and find a
scalar field φ such that ∇φ = F .

Observation - The method used to solve this example always works to find φ if it exists. But if we make a
mistake, and think erroneously that we have verified that a φ exists (but in fact it doesn’t), then this calculation
will be inconsistent; none of the possible φ we obtain from F1 will be consistent with the actual F2 , for example.

4.4.3 Summary: four equivalent ways of characterising a conservative vector field

In all the summary statements below we assume that the vector field is defined in a simply connected region U
and is suitably differentiable.

1. Path-independent integral for all paths γi starting at A and ending at B, generic points in U :
Z Z
F · dr = F · dr (4.10)
γ1 γ2

2. Vanishing integral round all closed curves γ contained in U :

I
F · dr = 0 (4.11)
γ

3. Existence of a potential:
F = ∇φ . (4.12)

4. The vector field is curl free3 :

∇ × F = 0 ∀x ∈ U . (4.13)
3
Vector fields which have vanishing curl are called irrotational.
 49

S1 a

S3
h

S2

Figure 4.6: Surface of the cylinder used in Example 4.4.

4.5 Divergence theorem

Stokes’ theorem relates a flux integral of the curl of a vector field over a surface S to a line integral round the
boundary ∂S. In this section we consider an integral theorem which relates a volume integral of the divergence
of a vector field over a region V to a flux integral over the boundary surface ∂V . This theorem is known as the
Divergence theorem, or sometimes as Gauss’s theorem 4 .

Theorem 4.6 (Gauss’s theorem or the divergence theorem) Let V be a bounded solid region in R3 with
boundary surface ∂V which consists of a finite number of piecewise smooth, closed orientable surfaces (with
the orientation chosen so that the normals point out of the surface). Let F be a C 1 vector field then
Z {
∇ · F dV = F · dS. (4.14)
V ∂V
{
The symbol is used to indicate that the integral is over a closed surface. The proof is omitted, but can be
found in a textbook.

Example 4.4 Let F be the vector field F = xi + yj + zk. Let S be the closed surface of the cylinder radius
a and height h, with the centre of the base at the origin (see figure 4.6). Calculate the flux integral
{
Φ= F · dS . (4.15)
S

In fact the result of this example is quite general:

Corollary 4.2 Let S be a closed surface. Then

{
r · dS = 3 × volume of region enclosed by S.
S

Proof: ∇ · r = 3. Therefore, by the divergence theorem:

{ Z
r · dS = 3 dV = 3 × volume of region enclosed by S.
S V

Z
Example 4.5 Use the divergence theorem to calculate F · dS where F = 2xy 2 i + z 3 j − x2 yk and S is
S
the hemisphere x2 + y 2 + z 2 = a2 , z ≥ 0 (see figure 4.7).
50

(0,0,a)

(0,a,0)
~
S

(-a,0,0) (a,0,0)

(0,-a,0) ~
S : x2 + y2 £ a2; z = 0
S : x2 + y2 + z 2 = a2

Figure 4.7: Volume and surfaces used in Example 4.5.

Example 4.6 Show that the equation ∇2 φ = 0 in D ⊂ R3 , with boundary condition n · ∇φ = 1 on ∂D, has
no solution. Here n is the outward unit vector to the surface ∂D that bounds the region D.

4.6 Vector potential

For a C 2 scalar field we know that ∇ × (∇φ) = 0 so that if F = ∇φ then ∇ × F = 0. However we have
also seen that in a simply connected region the converse is true and if ∇ × F = 0 there exists a scalar field φ
such that F = ∇φ.
In this section we examine the consequences of the identity ∇ · (∇ × F ) = 0. We have seen in section 2.10
that
G = ∇ × F =⇒ ∇ · G = 0.

However, in a simply connected region one can also show (although we do not give the proof in this course)
that
∇ · G = 0 =⇒ there exists a vector field F such that G = ∇ × F .

We call such a vector field F a vector potential for G.

Both scalar and vector potentials are of great importance in theoretical physics, particularly in the theories of
fluids and electromagnetism.

4.7 Summary of the integral theorems

Theorem 4.1, p. 41 [Stokes’ theorem] - Let S be a bounded, piecewise smooth, orientable, surface in R3 , with
boundary ∂S, which consists of a finite number of piecewise C 1 simple closed curves. Let F be a C 1 vector
field whose domain includes S. Then
Z I
(∇ × F ) · dS = F · dr
S ∂S

where the unit normal and direction of the line integrals are chosen consistently by following the right hand
rule.
Right hand rule - “The chosen direction of circulation around a boundary is such that your left arm is on the
surface and the unit normal points from your feet to your head.” In the case of an external boundary this is
equivalent to the rule that states: “wrap your right hand around the boundary and the normal is in the direction
of your thumb” or the “screw rule”. Note, however, that the circulation around a hole in the surface is in the
opposite direction to that on the outer boundary.
4
Carl Friedrich Gauss (1777-1855), German mathematician and physicist, sometimes regarded as the greatest of all mathematicians.
His contributions are too numerous to even approximate in this footnote.
Theorem 4.2, p. 44[Green’s theorem] - Let S be a bounded region in R2 , with boundary ∂S which consists of
a finite number of piecewise C 1 simple closed curves. Let F be a C 1 vector field in R2 whose domain includes
S. Then Z  
∂F2 ∂F1
I
− dx dy = (F1 dx + F2 dy) .
S ∂x ∂y ∂S
where the line integral is traversed according to the right hand rule.

Theorem 4.6, p. 49 [Divergence (or Gauss’) theorem] - Let V be a bounded region in R3 with boundary sur-
face ∂V which consists of a finite number of piecewise smooth, closed orientable surfaces (with the orientation
chosen so that the normals point out of the surface). Let F be a C 1 vector field defined in V then
Z {
∇ · F dV = F · dS.
V ∂V

Conservative vector fields - The following statements are equivalent conditions for a differentiable vector field
F to be conservative in a connected region U (see Section 4.4.3):

1. There exists a C 2 function φ defined in U such that F = ∇φ everywhere in U .

2. The line integral of F between any two points A and B in U is path-independent.

3. The line integral of F on any closed loop wholly contained in U is zero.

4. The curl of F is zero everywhere in U and U is simply connected.

52
 Part II

Complex variable theory

53
56
Chapter 5

Introduction to complex variable theory

5.1 Introduction
In this part of the module we will study differentiation and integration of complex functions. Much of the
theory resembles that of real functions. However, there are two main differences:

(a) Differentiability is a stronger condition in C than in R, e.g. f (x) = x|x| is a differentiable function
R → R, but g(z) = z|z| is not a differentiable function C → C. In C, differentiable once implies
differentiable n times for all n ≥ 1, but this is false in R: for example, f (x) = x|x| has no second
derivative at the origin.
Rb
(b) The value of a definite integral a f (z) dz may depend on the chosen path in C from a to b. This can
cause complications, but can also make some integrals easier Rto evaluate: a technique called the Calculus
b
of Residues allows one to evaluate certain definite integrals a f (z) dz without needing to find an anti-
derivative F (z) for f (z). This is not too dissimilar from what he have seen in vector calculus for line
integrals of conservative vector fields.

We will derive both sets of results in the next chapter. Here, instead, we recap the main properties of complex
numbers and complex functions.

5.2 Revision of complex numbers

Complex numbers arise as a natural extension of the number system:

• Within the natural numbers (x ∈ N) one can solve 2 + x = 3, but not 3 + x = 2;

• Within the integer numbers (x ∈ Z) one can solve 3 + x = 2, but not 2x = 3;

• Within the rational numbers (x ∈ Q) one can solve 2x = 3, but not x2 = 2;

• Within the real numbers (x ∈ R) one can solve x2 = 2, but not x2 = −2;

• Within the complex numbers (x ∈ C) one can solve any polynomial equation.

A complex number z can be written in Cartesian form z = x + iy where x and y are real numbers and
i2 = −1. We call x the real part of z and denote it by Re z = ℜ(z). Similarly, we call y the imaginary part of
z and denote it by Im z = ℑ(z).
A complex number can be represented as a vector in a complex plane, called the Argand diagram, see
Figure 5.1. This represents the complex plane by a Real and an Imaginary axis (horizontal and vertical
respectively, referring to x and y). Therefore a complex number z = x + iy is represented by a point with
coordinates (x, y). Alternatively, we can also think that it is identified by a position vector with components x
and y.

57
58

ℑ(z)
z = x + iy
y

θ ℜ(z)
x

z̄ = x − iy

Figure 5.1: Argand diagram - A complex number is represented by a point in the complex plane. Its complex
conjugate is its mirror image with respect to the horizontal (real axis).

Once we have introduced this geometrical representation for complex numbers, we can exploit what we
have learned in vector calculus to obtain different representations of complex numbers. For example, the
complex number z could be represented using polar coordinates

x = r cos(θ) and y = r sin(θ), (5.1)

where r is the distance from the origin and θ is the angle between the position vector that identifies z and the
real axis. In complex variable notation the distance is called the modulus of z and is represented as r = |z|.
From figure 5.1 we can easily see that p
|z| = x2 + y 2 . (5.2)
The angle θ is the argument of z, θ = arg(θ). Note that arg(z) is not uniquely defined: we can add any integer
multiple of 2π to arg(z) without altering z. To avoid confusion it is sometimes helpful to define the principal
value of the argument as Arg(z) where −π < Arg(z) ≤ π. From figure 5.1 we could be tempted to write that
y
θ = arctan . (5.3)
x
However, we must be very careful in applying this formula as the inverse tangent function has range (−π/2, π/2)
and so it is not a faithful representation of either Arg(z) or arg(z). A more accurate equation is
 y
 arctan
 x ≥ 0,
Arg(z) = x (5.4)
y
 arctan
 +π x < 0.
x
The polar representation of a complex number z can be made very compact. Using equation (5.1) we can write

z = x + iy = r cos(θ) + ir sin(θ) = reiθ , (5.5)

where the last equality can be proved using Taylor’s theorem, theorem 6.11. The complex exponential notation
is very powerful and the complex exponential has many convenient properties:

1. The product of two complex numbers zj = rj exp(iθj ), with j = {1, 2}, is

z1 z2 = r1 r2 ei(θ1 +θ2 ) .

2. The ratio of two complex numbers zj = rj exp(iθj ), with j = {1, 2}, is

z1 r1
= ei(θ1 −θ2 ) .
z2 r2
 59

3. |eiθ | = 1.
Z 2π
4. eiθ dθ = 0.
0

On the other hand, the polar notation is not very convenient when adding complex numbers.
Since z = reiθ = eln(r) eiθ = eln(r)+iθ , the complex natural logarithm function log is given by

log(z) = log(|z|) + i arg(z). (5.6)

log can also be written as ln – in this module all of these will denote logarithm with base e.

Complex conjugate
The complex conjugate of z is denoted with an asterisk or bar and is defined to be

z̄ ≡ z ∗ = x − iy = re−iθ . (5.7)

From the definition it follows that

|z̄| = |z| , arg(z̄) = − arg(z) and z z̄ = r2 . (5.8)

In the complex plane z̄ is a reflection of z in the x-axis.

Triangle inequalities
The triangle inequalities apply to any two complex numbers (z1 , z2 ):

|z1 + z2 | ≤ |z1 | + |z2 |, (5.9a)

|z1 − z2 | ≥ ||z1 | − |z2 ||. (5.9b)

Subsets of the complex plane

We can generalise the concepts of open and closed intervals from R to C. If z0 ∈ C and ε > 0, define the
neighbourhood of z0 as a disc, radius ε, centred at z0 :

Nε (z0 ) = {z ∈ C | |z − z0 | < ε}. (5.10)

If S ⊆ C, an element z0 ∈ S is an interior point of S if there exists ε > 0 such that Nε (z0 ) ⊆ S, i.e. all points
close to z0 are also in S. A subset S ⊆ C is open if every z0 ∈ S is an interior point.

Example 5.1 Determine which of the following set is open. Explain your answer.

(a) S = {z ∈ C | |z| < 1}, (b) S ′ = {z ∈ C | |z| ≤ 1}. (5.11)

A subset S ⊆ C is closed if its complement C \ S is open, i.e. if, whenever z0 6∈ S, there exists ε > 0 such that
Nε (z0 ) is disjoint from S.

Example 5.2 Determine which of the set in Example 5.1, equation (5.11), is closed. Explain your answer.

A circle of radius r with centre a in the complex plane has the equation

|z − a| = r, (5.12)

where a can be complex and r is real and positive. The inside of the circle

|z − a| < r, (5.13)
60

Open disk ℑ(z)

Closed disk
a4
|z − a2 | ≤ r2
a2
r4
r2
|z − a4 | < r4

ℜ(z)
Circle
Punctured
a3
open disk
r1 r3
a1

0 < |z − a3 | < r3
|z − a1 | = r1

Figure 5.2: Examples of disks in the complex plane.

is called an open disk. If we include the boundary it becomes a closed disk:

|z − a| ≤ r. (5.14)
If we remove the point a from the disk it is called a punctured disk - such a disc is usually open,
0 < |z − a| < r, (5.15)
but could be closed,
0 < |z − a| ≤ r. (5.16)
A graphical representation of these subsets is shown in figure 5.2.

5.3 Continuity and differentiability

A mapping f which maps a complex number z to a unique complex number w, that is w = f (z), is a function.
There are also mappings such that w is not unique; these multifunctions are not true functions but mappings. For
example, log is not a proper function of z, since arg(z) is not unique. Here we will concentrate on functions.
Definition 5.1 (Continuous function) Let f be a function defined on an open set S ⊆ C and let z0 ∈ S. Then
f is continuous at z0 if
lim f (z) = f (z0 ), (5.17)
z→z0
i.e. for each ε > 0 there exists δ > 0 such that
|z − z0 | < δ ⇒ |f (z) − f (z0 )| < ε. (5.18)

If f and g are continuous at z0 then so are f ± g, f g and, provided g(z0 ) 6= 0, f /g. We say that f is continuous
on S if it is continuous at every z0 ∈ S.
Definition 5.2 (Complex derivative) We say that f is differentiable at z0 , with derivative f ′ (z0 ), if
f (z) − f (z0 )
f ′ (z0 ) = lim (5.19)
z→z0 z − z0
exists finite.
 61

ℑ(z)

z0

ℜ(z)

Figure 5.3: In the definition of complex derivative, equation (5.19), the limit must be independent of the path
taken to reach the point z0 .

Observation - Notice that for the limit to exist, it must be independent of the direction in which h = (z − z0 )
is taken to zero (see figure 5.3). This is true also for real functions, but on the real axis we can approach a point
x0 in only two directions, from the left or the right. In the complex plane any path that is on the domain of
definition of
If f and g are differentiable at z0 then so are f ± g, f g and, provided g(z0 ) 6= 0, f /g, with the usual formulae
for their derivatives. For example,
g ′ (z0 )
 
d 1
=− 2 . (5.20)
dz g(z) g (z0 )
Definition 5.3 (Holomorphic function) We say that f is holomorphic on S if it is differentiable at every z0 ∈
S; it is entire if it is holomorphic on C.

For example, a polynomial function of degree n can be written as

n
X
f (z) = ak z k (5.21)
k=0
with ak ∈ C. Such a function is holomorphic throughout C and we have
n
X
′
f (z) = kak z k−1 (5.22)
k=1
as usual.

5.4 Cauchy-Riemann equations

Theorem 5.1 (Cauchy-Riemann equations) Let f (z) = u(x, y) + iv(x, y) with z = x + iy, with u, v, x and
y real. If f is differentiable at z0 then
∂u ∂v ∂v ∂u
= and =− (5.23)
∂x ∂y ∂x ∂y
at z0 .
Proof - We write the definition of derivative, equation (5.19), in terms of the real and imaginary parts of f
and z:
df f (z + h) − f (z)
= lim
dz h→0 h
(5.24)
u(x + δx, y + δy) + iv(x + δx, y + δy) − u(x, y) − iv(x, y)
= lim ,
h→0 δx + iδy
62

where
h = δx + iδy. (5.25)
As f is differentiable the limit holds for directions along which h tends to zero. We take first δy = 0 (so,
approach with h tending to zero along the x-axis). Then h = δx, so that

df u(x + δx, y) + iv(x + δx, y) − u(x, y) − iv(x, y)

= lim
dz δx→0  δx 
u(x + δx, y) − u(x, y) v(x + δx) − v(x, y)
= lim +i (5.26)
δx→0 δx δx
∂u ∂v
= + .
∂x ∂x
If, instead, we take δx = 0 (i.e., approach with h to zero along the y-axis), then h = iδy and we obtain

df u(x, y + δy) + iv(x, y + δy) − u(x, y) − iv(x, y)

=
dz iδy
(5.27)
∂u ∂v
=−i + .
∂y ∂y

Comparing equations (5.26) and (5.27) we obtain that

ux + ivx = f ′ (z) = vy − iuy . (5.28)

Taking the real and imaginary parts of this expression and equating them we obtain the Cauchy-Riemann
equations, equation (5.23).
Now assume that u(x, y) and v(x, y) are twice differentiable, and that their first and second derivative are
continuous. (In fact, as we shall see later, if f (z) is analytic, u and v are automatically guaranteed to have
continuous partial derivatives of all orders!) By taking another derivative of the Cauchy-Riemann equations,
we obtain
uxx = vyx = −uyy and vyy = uxy = −vxx (5.29)
and hence
uxx + uyy = 0 and vxx + vyy = 0. (5.30)
We find that u(x, y) and v(x, y) both satisfy Laplace’s equation. Such functions are called harmonic.
Observation 1 - While if f (z) is analytic then its real and imaginary must satisfy the Cauchy–Riemann re-
lations, the converse is not true. However, if the two real functions u(x, y) and v(x, y) obey the Cauchy–
Riemann equation at (x0 , y0 ) and have continuous first derivatives there, then the complex function f (z) =
u(x, y) + iv(x, y), with z = x + iy is differentiable at z0 = x0 + iy0 .
Observation 2 - Notice the similarity with conservative vector fields. If a vector field F is conservative at x0
then ∇ × F = 0 there. For a two dimensional vector field this condition is identical to equation (5.23). The
converse is true only if F is differentiable in a disk around x0 .
Observation 3 - The Cauchy-Riemann equations in polar form, with z = reiθ , are:

∂u 1 ∂v ∂v 1 ∂u
= and =− . (5.31)
∂r r ∂θ ∂r r ∂θ
The polar form of the Cauchy-Riemann equations is useful whenever the original function f is more naturally
expressed in terms of polar coordinates.

Example 5.3 Show that f (z) = z 2 is holomorphic for all z ∈ C.

Example 5.4 Determine whether f (z) = z̄ is differentiable anywhere in C.

 63

5.5 Useful functions

In this section we quickly review some fundamental complex functions.

Polynomials
A complex polynomial has form f (z) = nk=0 ak z k , where each ak ∈ C; the usual rules for algebra, differ-
P
entiation and integration apply. One case show (for example, in the third year Complex Variables module) that
a polynomial of degree n has n roots in C, counting multiple roots (Fundamental Theorem of Algebra; this is
false in R, e.g. f (x) = x2 + 1.

Rational functions
A rational function has form f (z) = p(z)/q(z) where p(z) and q(z) are polynomials. It is defined wherever
q(z) 6= 0; the usual rules for algebra, differentiation and integration apply.

Exponential function
The exponential function is defined as
∞
z
X zk
exp(z) ≡ e = . (5.32)
k!
k=0
This power series converges for all z ∈ C, so it is differentiable term-by-term:
∞ ∞ ∞
X kz k−1 X z k−1 X zk
exp′ (z) = = = = exp(z). (5.33)
k! (k − 1)! k!
k=1 k=1 k=0

We will discuss power series and their convergence in section 6.6.

Basic identities

exp(z1 + z2 ) = exp(z1 ) exp(z2 ), (5.34a)

exp(z1 )
exp(z1 − z2 ) = , (5.34b)
exp(z2 )
exp(z)n = exp(nz) for n ∈ Z, (5.34c)
x
exp(x + iy) = e [cos(y) + i sin(y)], (5.34d)
exp(z + 2πi) = exp(z). (5.34e)

Equation (5.34e) states that the exponential is periodic in the imaginary part of its argument with period 2π.
These properties also imply that the exponential is an infinite-to-one map, C → C \ {0} and a bijection
R × (−π, π] → C \ {0}. The origin of C is excluded as the exponential function never takes the value zero.

Hyperbolic functions
The hyperbolic functions are defined in terms of the exponential. Using the series definition of this last function,
equation (5.32), it is relatively straightforward to obtain the series expansion of the hyperbolic functions.
The two fundamental hyperbolic functions are the hyperbolic cosine
∞
ez + e−z X z 2k
cosh(z) = = , (5.35)
2 (2k)!
k=0

and the hyperbolic sine

∞
ez − e−z X z 2k+1
sinh(z) = = . (5.36)
2 (2k + 1)!
k=0
64

They are one the derivative of the other:

cosh′ (z) = sinh(z) and sinh′ (z) = cosh(z). (5.37)

As the exponential they are both periodic functions in the imaginary part of their argument with period 2π:

cosh(z + 2πi) = cosh(z), and sinh(z + 2πi) = sinh(z). (5.38)

The other hyperbolic functions, such as tanh(z), are defined in the usual way.

Trigonometric functions
The trigonometric functions are also defined in terms of the exponential, but with argument iz. The two
fundamental trigonometric functions are the cosine
∞
eiz + e−iz X (−1)k z 2k
cos(z) = = , (5.39)
2 (2k)!
k=0

and the sine

∞
eiz − e−iz X (−1)k z 2k+1
sin(z) = = . (5.40)
2i (2k + 1)!
k=0

They have derivatives

cos′ (z) = − sin(z), and sin′ (z) = cos(z). (5.41)
The usual trigonometric identities hold also for complex arguments, for example

cos(z1 + z2 ) = cos(z1 ) cos(z2 ) − sin(z1 ) sin(z2 ), (5.42)

sin(z1 + z2 ) = sin(z1 ) cos(z2 ) + cos(z1 ) sin(z2 ), (5.43)

The complex sine and cosine are periodic in the real part of their argument with period 2π:

cos(z + 2π) = cos(z) and sin(z + 2π) = sin(z) . (5.44)

It is possible to write trigonometric functions in terms of hyperbolic functions and vice versa, either as complex
or real functions. In the first case:

cos(iz) = cosh(z) and sin(iz) = i sinh(z). (5.45)

In the second case we write z = x + iy and expand the complex sine function in its real and imaginary part,

sin(z) = sin(x + iy) = sin(x) cos(iy) + cos(x) sin(iy) = sin(x) cosh(y) + i cos(x) sinh(y), (5.46)

and collect the two so that

ℜ[sin(z)] = sin(x) cosh(y) and ℑ[sin(z)] = cos(x) sinh(y). (5.47)

Similar relations hold for the real and imaginary parts of cos(z). The other trigonometric functions, such as
tan(z), are defined in the usual way.

5.5.1 Multi-functions
The final two types of mappings are strictly not functions (since they are multi-valued), but multi-functions.
While we will compute their values in this module, in general their integration can only be dealt with using
methods studied in third year modules.
 65

Logarithm
The formal definition of logarithm is that w is the logarithm of z, indicated with w = log(z), or with w = ln(z),
if and only if z = exp(w). If we write z = reiθ and apply the usual properties of the logarithm, we can write

log(z) = log(r) + iθ = log(|z|) + i arg(z), (5.48)

This is clearly a many-valued since so is arg(z). The principal value

Log(z) = log(|z|) + i Arg(z) (5.49)

with
− π < Arg(z) ≤ π, (5.50)
is continuous on S = C \ {z | ℜ(z) ≤ 0}, but not on the half-line ℜ(z) ≤ 0 where the value of the Arg(z)
function changes by ±2π. It is also holomorphic on S with derivative
1
Log′ (z) = . (5.51)
z
More generally, for any a, b ∈ R with 0 < b − a ≤ 2π we can define a single-valued branch of log(z) on
S = {z = reiθ | r > 0, a < θ < b}, with derivative 1/z.

Powers
We define the power of a complex base a ∈ C to an exponent z ∈ C as

az = exp[z log(a)]. (5.52)

This is in general many-valued as so is log(a). We define its principal value as

P.V. [az ] = exp[zLog(a)], (5.53)

with Log(a) the principal value of the logarithm. This function is holomorphic on S = C \ {z | ℜ(z) ≤ 0} and
has derivative Log(a) az .

Roots
Roots are powers of z with exponents of the form 1/n, with n ∈ N+ :
√
n
z ≡ z 1/n . (5.54)

Roots are most easily computed in polar notation. If we substitute z = reiθ+i2mπ , with m ∈ Z in equa-
tion (5.54) we obtain
z 1/n = r1/n eiθ/n ei2mπ/n . (5.55)
Therefore, there are n distinct roots of z, one for each value of m = 0, 1, 2, . . . n − 1. These have modulus
r1/n = |z|1/n and argument θ/n + i2mπ/n. On the Argand diagram, the roots are all on a circle centred at the
origin with radius |z|1/n . The spokes that join them to the origin are separated by an angle θ/n.

Example 5.5 Find all the fourth roots of z = 1 + i.

66

ℑ(z)
ei(π/16+π/2)

π/2

eiπ/16
π/16

ℜ(z)
ei(π/16+2π/2)
21/8

ei(π/16+3π/2)

Figure 5.4: Representation of the Argand diagrams of the four fourth roots of (1 + i), see Example 5.5.
Chapter 6

Paths and integrals in the complex plane

6.1 Introduction

We are used to integrating a real function along the real line: for example, integrating a real function f (x) as
x ranges between α and β and finding the (signed) area underneath its graph. In the first part of this module
we have extended this concept to line integrals of vector fields along a curve in Rn . In this case we saw that,
in general, the integral depends on the path and on the end points. Only for conservative vector fields was the
path unimportant.
In this chapter we want to extend the concept of integral of a real function on a subset of the real axis, to the
integral of a complex function over a generic path in the complex plane. In doing this we will see that there are
many analogies with two–dimensional vector fields. For example, we will see that holomorphic functions are
very similar to conservative vector fields. However, while for vector fields we were particularly interested in
conservative fields, in the case of complex integration the functions that are not holomorphic (i.e. the “fields”
that are not “conservative”) are of most interest.

6.2 Definition of path integral for a complex function

Definition 6.1 A curve in C is a smooth function γ : [a, b] ⊆ R → C.

Definition 6.2 A path γ is a finite collection of joint curves, γ = (γ1 , γ2 , . . . , γn ) such that for all 1 ≤ i ≤
n − 1, the final point of γi coincide with the initial point of γi+1 .

If γ(a) = γ(b) then the path γ is closed. If γ(t1 ) = γ(t2 ) only if t1 = t2 for all t ∈ (a, b) then the path γ is
simple, which means that it does not intersect itself. Note that this does not include end points.
Since a closed path starts where it finishes we need to also define its orientation. By convention we take the
orientation to be in an anticlockwise direction. An alternative name for path is contour and one can equally
refer to path integrals or contour integrals, for example.
Finally, we should note that a curve or path is defined as an application, not as a set of points in the complex
plane. The latter is the range of the path γ and not the path. In particular, different paths may have the same
graph (similarly to what discussed for curves in Rn in section 2.1, page 17. Having made this clear, we will,
however, refer to a path by the name of its graph, e.g. a circle centred at the origin of radius 1, when there is no
ambiguity. Examples of curves, paths (contours) are shown in figure 6.1.

67
68

ℑ(z)

Curve
(not simple) Path/contour
Curve (simple)
(closed, not simple)

Curve Path/contour
(closed and simple) (closed and simple)

ℜ(z)

Figure 6.1: Curves, paths and contours in the complex plane. The small circles indicates the points where
curves are joined to form a path. The function that parameterises the path, z(t), will not have continuous
derivative there.

6.3 Path integrals

6.3.1 Definitions
Definition 6.3 Let γ(t) = x(t) + iy(t) with t ∈ [a, b] ⊆ R and x(t), y(t) ∈ R, be a path γ ∈ C. The length
of γ is defined as
Z Z bp Z b
L(γ) = ds = x′ (t)2 + y ′ (t)2 dt = |γ ′ (t)| dt, (6.1)
γ a a
provided this integral exists.
One can prove that the integral is defined if γ is smooth in [a, b], i.e. x(t) and y(t) have continuous derivatives
in [a, b]. Here ds is called the line element (as in vector calculus) and is a shorthand notation for
p
ds = x′ (t)2 + y ′ (t)2 dt . (6.2)
Observation - Equation (6.1) is a definition, but it can be justified as a limiting process exactly as the formula
for the length of a curve studied in Vector Calculus, equation (2.10) and section 3.1.1.
If the path γ has corners (e.g. it is a square), i.e. if there are finitely many points a = t0 < t1 < · · · <
tn = b in [a, b] such that x(t) and y(t) have continuous derivatives on each sub-interval (tk , tk+1 ), where
k = 0, . . . , n − 1, then the length of γ is
n−1
X Z tk+1 p
L(γ) = x′ (t)2 + y ′ (t)2 dt. (6.3)
k=0 tk

Definition 6.4 Let γ(t) = x(t) + iy(t) with t ∈ [a, b] ⊆ R and x(t), y(t) ∈ R, be a path γ ∈ C and let f (z)
be a complex function defined on γ. The path integral of f (z) over γ is defined as
Z Z b
f dz = f [γ(t)]γ ′ (t) dt , (6.4)
γ a

provided that the limit exists.

 69

ℑ(z)

r
γ

ℜ(z)

Figure 6.2: The circle Ca;r = {z : |z − a| = r} = {z(t) = a + reit : 0 ≤ t < 2π}.

If the path γ has corners (e.g. it is a square), i.e there are finitely many points a = t0 < t1 < · · · < tn = b in
[a, b] such that x(t) and y(t) have continuous derivatives on each sub-interval (tk , tk+1 ), where k = 0, . . . , n −
1, then the integral of f (z) on γ is

Z n−1
X Z tk+1
f dz = f [γ(t)]γ ′ (t) dt. (6.5)
γ k=0 tk

Example 6.1 Let f (z) = 1/z and let C be a segment of the unit circle, from exp(ia) to exp(ib), parameterised
by γ(t) = exp(it) for t ∈ [a, b]. Compute
dz
Z
. (6.6)
γ z

Example 6.2 Integrate the function (z − a)n around a circle Ca;r with radius r and centre a (see figure 6.2),
when n is an integer, i.e., find
Z
In = (z − a)n dz. (6.7)
Ca;r

Example 6.3 Let γ be the unit square, with corners at 0, 1, 1 + i and i. Evaluate
Z
z 2 dz . (6.8)
γ

6.3.2 Fundamental theorems

Having defined path integrals, we can now start studying their properties. The first is that the sign of the path
integral depends on the direction the path is traversed (similarly to line integrals of vector fields, section 3.1.1).

Theorem 6.1 Let −γ denote the contour γ in the reverse direction, i.e. parametrized by γ(a + b − t) for
t ∈ [a, b]. Then
Z Z
f (z) dz = − f (z) dz. (6.9)
−γ γ
70

Proof - The proof is based on introducing a new variable τ = a + b − t in the integral:

Z b
d
Z
f (z) dz = f [γ(a + b − t)] γ(a + b − t) dt
−γ a dt
Z a  
d
= f [γ(τ )] − g(τ ) (−dτ )
dτ
Zb a
d
= f [γ(τ )] g(τ ) dτ (6.10)
b dτ
Z b
d
=− f [γ(τ )] g(τ ) dτ
dτ
Za
= − f (z) dz.
γ

The next theorem generalises to complex functions the relation between integration and differentiation already
seen for real variables.

Theorem 6.2 (The Fundamental Theorem of Calculus) Let f have an anti-derivative F (i.e. F ′ = f ) in a
region D containing a contour γ from γ(a) = z0 to γ(b) = z1 . Then
Z
f (z) dz = F (z1 ) − F (z0 ). (6.11)
γ

Proof -
Z Z b
f (z) dz = F ′ [γ(t)]γ ′ (t) dt
γ a
(6.12)
b
d
Z
= F [γ(t)] dt = F [(γ(b)] − F [γ(a)] = F (z1 ) − F (z0 ).
a dt

Corollary 6.1 Let f have a single-values anti-derivative F in a region D containing a closed contour γ. Then
Z
f (z) dz = 0. (6.13)
γ

Proof - Apply Theorem 6.2 with z0 = z1 ∈ γ.

Observation 1 - Corollary 6.1 explains the result in Example 6.3, showing that γ z 2 dz = 0 where γ is the
R

unit square: we can applyR Corollary 6.1 with F (z) = z 3 /3 and D = C. More generally, if n 6= −1 and C is
any closed contour, then C z n dz = 0 since we can take F (z) = z n+1 /(n + 1).
Observation 2 - The preceding remark fails when n = −1, since log(z) is not single-valued (see figure 6.3).
Let f (z) = 1/z as in Example 6.1, on a segment γ of the unit circle S 1 from exp(ia) to exp(ib). The integral
has value
1
Z
dz = log[exp(ib)] − log[exp(ia)] = i(b − a), (6.14)
γ z

as before. However, if b = a + 2π we have a problem: γ(b) = γ(a), but

log[γ(b)] = ib 6= log[γ(a)] = ia (6.15)

and
1
Z
dz = log(exp(ib)) − log(exp(ia)) = i(b − a) = 2iπ. (6.16)
γ z
 71

ℑ[log(z)]
2

1
-1
0
-1 0
1

ℜ(z) ℑ(z)

Figure 6.3: Graph of the imaginary part of log(z), as z moves along the curve γ(t) = exp(it), with t ∈ [0, 4π].
Even though γ(t) = γ(t + 2π) log[γ(t)] 6= log[γ(t + 2π)].

More generally, if a closed contour γ winds k times around 0 for some integer k, then dividing γ into suitable
sub-contours and adding the integrals along each give
1
Z
dz = 2kπi. (6.17)
γ z
We call k the winding number of γ around the origin; it can be positive or negative, as γ goes around the
origin in the positive or negative direction.
ObservationR x 3 - In the case of real functions, every continuous function f has an anti-derivative, namely
F (x) = a f (t) dt where a is constant. However, not every continuous complex function f has an anti-
derivative.

Example 6.4 Show that f (z) = |z|2 has no anti-derivative.

The function f (z) = |z|2 = x2 + y 2 , where z = x + iy, is continuous everywhere. Suppose that there exists
a differentiable function F (z) satisfying F ′ (z) = f (z). If we decompose F into real and imaginary parts
(F = u + iv) then using the Cauchy-Riemann equations gives

F ′ (z) = ux + ivx = vy − iuy . (6.18)

Since f is a real-valued function of z, F ′ (z) must be real, and so the imaginary parts in equation (6.18) must
be zero, i.e.
vx ≡ 0 and uy ≡ 0. (6.19)
So vy is a function of y only, and ux is a function of x only. Looking back at equation (6.18) we must conclude
that F ′ (z) is both a function of y only and of x only, and so it must be a constant. But f (z) is not constant and
so there is no such F .

We conclude this section with two theorems that are of fundamental importance for the evaluation of inte-
grals using complex variable methods.

Theorem 6.3 If f : [a, b] → C is continuous then

Z b  Z b
 
R≡  f (t) dt ≤ |f (t)| dt. (6.20)
a a
72

Proof - Assume that the left hand side, defined as R ∈ R, is non-zero1 . Then set
Rb b
f (t)dt 1
Z
W = R a  = f (t)dt . (6.21)
 b
 a f (t) dt
 R a

By construction |W | = 1 and
Z b Z b Z b
−1 −1
ℜ W −1 f (t) dt,
 
R=W f (t) dt = W f (t) dt = (6.22)
a a a

where, in the last step, we have used the fact that R is real. We can use the fact that the real part of a complex
number is smaller than its modulus to complete the proof:
Z b Z b 
Z b
−1 W −1 f (t) dt =
 
R= ℜ W f (t) dt ≤ |f (t)| dt, (6.23)
a a a

where in the last step we have used that |W | = 1.

Theorem 6.4 (The Estimation Lemma) If |f (z)| ≤ M for all z on a contour γ of length L then
Z 
 
 f (z) dz  ≤ M L. (6.24)
 
γ

Proof - Let γ be parameterised by γ : [a, b] → C. Then

Z  Z b  Z b Z b
   ′
 ′
 f (z) dz  = 
   f [γ(t)]γ (t) dt ≤
 |f [γ(t)]| |γ (t)| dt ≤ M |γ ′ (t)| dt = M L. (6.25)
γ a a a

Example 6.5 Define IR as

z+1
Z
IR = dz, (6.26)
CR z3 − 1

where CR is the circle |z| = R of radius R > 1 centred at the origin. Show that

lim IR = 0. (6.27)
R→∞

6.4 Cauchy’s Theorem

Cauchy’s theorem is one of the most important theorems in mathematics:

Theorem 6.5 (Cauchy’s theorem) If f is holomorphic inside and on a closed contour γ, then
Z
f (z) dz = 0. (6.28)
γ
1
If R = 0 then the theorem is trivial because the right hand side of equation (6.20) is positive, as it is the integral of a positive
quantity.
 73

Proof - The proof is based on Green’s theorem (see p. 44). Let f (z) = u + iv with u, v ∈ R, and let γ be
parameterised by γ(t) = x(t) + iy(t) for a ≤ t ≤ b, with x(t), y(t) ∈ R. Then
Z Z b
f (z) dz = (u + iv)(x′ + iy ′ ) dt
γ a
Z b
= ((ux′ − vy ′ ) + i(uy ′ + vx′ )) dt
Za Z (6.29)
= (u dx − v dy) + i (u dy + v dx)
γ γ
x  ∂v ∂u
 x  ∂u ∂v 
= − + dx dy + i − dx dy,
∂x ∂y ∂x ∂y
D D

with the last step justified by applying Green’s Theorem (Theorem 4.2 on p. 44) to the vector fields (u, −v) =
u − iv = f and (v, u) = v + iu = if , where D is the region enclosed by γ, so that γ is the boundary ∂D
of D. But both integrands
R are identically zero in D, since f satisfies the Cauchy-Riemann equations there,
equations (5.23), so γ f (z) dz = 0, as required.
Observation
R - If f (z) is an entire function, such as a polynomial, or exp(z), sin(z), cos(z), sinh(z) or cosh(z),
then γ f (z) dz = 0 for every closed contour C, since f is holomorphic everywhere.
The following theorem is a first application of Cauchy’s theorem:

Theorem 6.6 (Deformation theorem) If f is holomorphic in a region D, and γ1 and γ2 are closed contours
in D which can be continuously deformed into each other within D, then
Z Z
f (z) dz = f (z) dz. (6.30)
γ1 γ2

Proof - Let γ0 be a path in D joining γ1 to γ2 , and let γ be the closed contour

γ = γ1 + γ0 + (−γ2 ) + (−γ0 ), (6.31)

where + means “followed by” and − means the reverse path (as in Theorem 6.1). Then f is holomorphic on
and within γ, so Cauchy’s Theorem gives
Z
0 = f (z) dz
γ
Z Z Z Z
= f (z) dz + f (z) dz + f (z) dz + f (z) dz
γ1 γ0 −γ2 −γ0
Z Z Z Z (6.32)
= f (z) dz + f (z) dz − f (z) dz − f (z) dz
γ1 γ0 γ2 γ0
Z Z
= f (z) dz − f (z) dz,
γ1 γ2

using Theorem 6.1 twice.

R may be “holes” within D, where f is not holomorphic (or even undefined), so it
Observation - Note that there
is not necessarily true that γi f (z) dz = 0 for i = 1 or 2; however, since γ1 can be continuously deformed to
γ2 within D, such holes cannot lie between γ1 and γ2 , and hence cannot be within γ.

Example 6.6 Show that the integral of the function f (z) = 1/z around any closed path γ with winding number
k, i.e. that loops k times round the origin, but does not pass through it, is equal to
dz
Z
= 2πik. (6.33)
γ z
74

Holomorphic functions have many features of conservative vector fields, studied in section 3.2. We con-
clude this introductory section to path integrals, by exploring this connection. First of all we extend to the
complex plane the definition of simply connected domain already introduced in vector calculus:
Definition 6.5 A region D ⊆ C is simply connected if any closed contour in D can be continuously deformed,
within D, to a point.
Intuitively, a simple connected domain in C has no “holes”, unlike the domain in Example 6.6.
We now state the equivalent of some of the conservative vector fields theorems studied in section 3.2 and
summarised in section 4.4.3.
Theorem 6.7 Let f be holomorphic in a simply connected region D ⊆ C. Then
R
1. γ f (z) dz = 0 for every closed contour γ in D;
Rz
2. if z0 , z1 ∈ D then z01 f (z) dz is independent of the chosen path in D from z0 to z1 .
Proof - The first statement is a direct consequence of Cauchy’s theorem, Theorem 6.5. The second statement is
a consequence of the Deformation theorem, Theorem 6.6: as D is simply connected, any path γ1 that joins z0
to z1 can be deformed in any other path with the same end points. By theorem 6.6 all these path integrals have
the same value, i.e. their value is independent of the path taken.

6.5 Cauchy’s Integral Formula

Having studied in the previous section functions that are holomorphic on and within a closed contour γ, we
now study functions that are not so well behaved. What we will find is that their integrals have remarkable
properties.
Theorem 6.8 (Cauchy’s Integral Formula) If f is holomorphic inside and on a simple closed contour γ, and
a is inside γ, then
1 f (z)
Z
f (a) = dz. (6.34)
2πi γ z − a
Proof - Omitted.
Observation 1 - This theorem states that the values of f inside γ are determined by its values on γ! This is not
so strange if we remember that f (z) is holomorphic on and inside γ. We have seen in section 5.4 that the real
and imaginary parts of f (z), u(x, y) and v(x, y), satisfy Laplace’s equation, see equation (5.30):
∇2 u = 0 and ∇2 v = 0.
You will study in the second semester that these equations have one and only one solution if the values of u
and v are specified on the boundary, i.e. if f (z) is known on γ. Therefore, we can see the Cauchy’s Integral
Formula, theorem 6.8, as a different way to see this property of Laplace’s equation.
Observation 2 - Using integrals on the boundary to find the values of a function inside the boundaries, as in
equation (6.34), is a widely used class of numerical techniques to solve partial differential equations that goes
under the name of Boundary Elements Methods.

Example 6.7 If |a| < 1 then since f (z) = sin(z) is an entire function, Theorem 6.8 gives
sin(z)
Z
dz = 2πi sin(a). (6.35)
S z−a
1

Here S 1 denotes a circle of unit radius, centred on the origin. However, if |a| > 1 then applying Cauchy’s
Theorem, Theorem 6.5, gives
sin(z)
Z
dz = 0, (6.36)
S1 z − a
as the function g(z) = sin(z)/(z − a) is holomorphic on and in S 1 is |a| > 1.
 75

Cauchy’s integral formula can be generalised to the derivatives of f (z):

Theorem 6.9 (Cauchy’s integral formula for the nth derivative) If f is holomorphic inside and on a simple
closed contour γ, and a is inside γ, then f is n times differentiable at a for each integer n ≥ 0, with

n! f (z)
Z
f (n) (a) = dz. (6.37)
2πi γ (z − a)n+1

Proof - Omitted.
Observation - An immediate consequence of this theorem is that if a function is holomorphic in a neighbour-
hood of a then it also differentiable many times there.

Example 6.8 If |a| < 1 then Theorem 6.9 with n = 1 gives

sin(z)
Z
dz = 2πi cos(a). (6.38)
S 1 (z − a)2

Here S 1 again denotes a circle of unit radius, centred on the origin. However, if |a| > 1 then applying Cauchy’s
Theorem, Theorem 6.5, gives
sin(z)
Z
dz = 0, (6.39)
S 1 (z − a)2
as the function g(z) = sin(z)/(z − a)2 is holomorphic on and in S 1 is |a| > 1.

Example 6.9 If |a| < 1 and n is divisible by 4 then Theorem 6.9 gives

sin(z) 2πi sin(a)

Z
n+1
dz = . (6.40)
S 1 (z − a) n!

There are similar results for other integers n, replacing sin(z) with − sin(z) or ± cos(z) as n = 4m + 2 or
4m ± 1 for some integer m.

Corollary 6.2 (Cauchy’s inequalities) If f is holomorphic inside and on a circle γ, with centre a and radius
R, and |f (z)| ≤ M for all z ∈ γ, then
 (n)  M n!
 
f (a) ≤ n . (6.41)
R
Proof - Omitted.
The last theorem of this section states a rather surprising property of holomorphic functions:

Corollary 6.3 (Liouville’s Theorem) If f is holomorphic and bounded on C, then f is constant: every bounded
entire function is constant.

Proof - Omitted.
Observation - The corresponding result for real functions is false: for instance f (x) = sin(x) is differentiable
and bounded on R, but not constant.

6.6 Series expansions of complex functions

Many properties of complex functions are related to the fact that it is possible to represent them as power
series. This is particular useful when evaluating complex integrals, as we will see in section 6.8. Here we
start by studying the series expansion of holomorphic functions and then move on to consider non-holomorphic
functions.
76

6.6.1 Taylor expansion of analytic functions

First of all we define where a (positive) power series converges:

Definition 6.6 (Radius of convergence) Consider the power series

∞
X
an (z − z0 )n , (6.42)
n=0

with z, z0 and an all in C. A real number R > 0 is called the radius of convergence of the series if the series
converges absolutely for all |z − z0 | < R and diverges otherwise. If the series converges for z ∈ C the radius
R is considered infinite.

Example 6.10 The radius of convergence of

∞
X zn
exp(z) = (6.43)
n!
n=0

is infinity. However, the radius of convergence of the series (“Geometric series”)

∞
X
zn (6.44)
n=0

is one and the series converges to f (z) = 1/(1 − z) for |z| < 1.

Definition 6.7 (Analytic function) A complex function f (z) is analytic at z0 ∈ C if it can be represented as a
power series near z0 . i.e. if there exists R > 0 and an ∈ C, with n ∈ N, such that
∞
X
f (z) = an (z − z0 )n , (6.45)
n=0

for |z − z0 | < R.

The following two theorems show that for complex functions being analytic is the same as being holomor-
phic, i.e. being infinitely many times differentiable is equivalent to being represented as a (converging) power
series.
The first theorem states that if a function is analytic then it is holomorphic:
n
P∞
Theorem 6.10 Let n=0 an (z − z0 ) be a power series with z0 ∈ C and an ∈ C for all n ∈ N . If the power
series n=0 an (z − z0 )n converges to f (z) with radius of convergence R, then f is holomorphic on this disc,
P∞
with
∞
X
f ′ (z) = nan (z − z0 )n−1 . (6.46)
n=1

Proof - Omitted.
The second theorem states that if a function is holomorphic on a disc then it is also analytic there.

Theorem 6.11 (Taylor’s Theorem) If f is holomorphic on a disc |z − z0 | < R, then for all such z we have
∞
X f (n) (z0 )
f (z) = (z − z0 )n . (6.47)
n!
n=0

This power series is called the Taylor expansion of f (z) around z0 .

 77

ℑ(z)

RoC = 1

a=3 ℜ(z)
a=0
RoC = 2

f (z) = (1 − z)−1

Figure 6.4: Regions of convergence of the series expansions of the function f (z) = (1 − z)−1 around z = 0
and z = 3. The boundary of the convergence disk is fixed by the singularity of f (z) at z = 1 (indicated by a
diamond in the diagram).

“Proof” - To motivate (rather than prove) this result, suppose that f (z) = n an (z−z0 )n on the disc |z−z0 | <
P
R for some constants an . By Theorem 6.10 f is holomorphic and the series can be differentiated term by term.
If we differentiate k times we get
∞
X n!
f (k) (z) = an (z − z0 )n−k , (6.48)
(n − k)!
n=k

for |z − z0 | < R. Putting z = z0 gives f (k) (z0 ) = k!ak , so (renaming the dummy variable)

f (n) (z0 )
an = (6.49)
n!
for all n ≥ 0.
Observation - These theorems do not hold for real functions. A real function can be differentiable infinitely
many times, but not have a power series representation. For example,

exp(−1/x2 )

x 6= 0,
f (x) = (6.50)
0 x = 0,

is differentiablePinfinitely many times, with f (n) (0) = 0 for all n ≥ 0 (exercise!). If it could be represented as
a power series an xn near x0 = 0 we would have an = f (n) (0)/n! = 0 for all n ≥ 0, so f (x) = 0 for all x
near 0, which is clearly false since f (x) > 0 for all x 6= 0.

1
Example 6.11 Find the Taylor Series of f (z) = 1−z about z = 0.

1
Example 6.12 Find the Taylor Series of f (z) = 1−z about z = 3.

The results of Examples 6.11 and 6.12 become intuitive if we sketch the convergence disks on an Argand,
as in figure 6.4.
78

The binomial series

The binomial series often offers the quickest way in deriving power series, so it is important to remember it:

k(k − 1) 2 k(k − 1)(k − 2) 3

(1 + z)k =1 + kz + z + z + ...
2 3!
k(k − 1)(k − 2) . . . (k − (n − 1)) n
+ z + ... (6.51)
n!
∞
X k(k − 1) · · · [k − (n − 1)] n
=1 + z ,
n!
n=1

which converges for |z| < 1. A special case which arises often is that already mentioned above, namely
∞
X
(1 − z) −1
= zn. (6.52)
n=0

It is possible to write a binomial expansion also for |z| > 1 by rewriting the expansion (6.51) in terms of 1/z:

1 k
 
k k
(1 + z) = z 1 +
z
" ∞
# (6.53)
k
X k(k − 1) · · · [k − (n − 1)] 1
=z 1+ .
n! zn
n=1

Note that this is a power series in negative powers of z. It is a first example of power series with both positive
and negative powers of z. We will see in the next section that this are essential to describe functions that are
not holomorphic.

6.6.2 Laurent series

Definition 6.8 (Laurent series) A series of the type
+∞
X
cn (z − z0 )n , (6.54)
n=−∞

with z, z0 and cn in C is called a Laurent series in z around z0 .

We say that a Laurent series converges to a function f (z) if the series of positive and negative powers
converge separately, i.e. if we can write f (z) = g(z) + h(z), with
+∞
X
g(z) = cn (z − z0 )n (6.55)
n=0

and
−1
X
h(z) = cn (z − z0 )n . (6.56)
n=−∞

The Laurent series converges in an open annulus D = {z ∈ C : R1 < |z| < R2 }, i.e. there exists a unique
inner radius R1 and a unique outer radius R2 that determine the largest annulus around z0 in which the series
converges. Outside this annulus the series diverges.
A function that is holomorphic in an annulus, but not on the corresponding disk, cannot be represented by
a Taylor series. It can, however, be represented by a Laurent series:
 79

Theorem 6.12 (Existence of the Laurent series) If f is holomorphic in an annulus R1 < |z − z0 | < R2 then
f is represented by a Laurent series in this region, i.e.
+∞
X
f (z) = cn (z − z0 )n , (6.57)
n=−∞

for R1 < |z − z0 | < R2 . The Laurent series in a given annulus is unique.

Proof - Omitted.

Laurent series expansion

We will see in section 6.8 that the Laurent series expansion is an extremely important theoretical tool that will
allow us to easily evaluate path integrals and, also, real integrals. However, contrary to the Taylor expansion,
which is of immediate practical use, there is seldom any need to evaluate the Laurent series expansion of a
function. If required to do so, one can resort to a range of techniques.
First of all, it is always possible to compute the coefficients using the formula

1 f (z)
I
ck = dz, (6.58)
2πi γ (z − z0 )k+1

where k ∈ Z and γ is a simple closed path around z0 wholly contained in the annulus of convergence of the
series. We can obtain this result by using the expansion (6.57) to write
+∞
f (z) X
= cn (z − z0 )n−k−1 . (6.59)
(z − z0 )k+1 n=−∞

Integrating term by term we obtain

+∞
f (z)
I X I
dz = cn (z − z0 )n−k−1 dz. (6.60)
γ (z − z0 )n+1 n=−∞ γ

By Corollary 6.1 all integrals on the right hand side are zero, except for n − k − 1 = −1, i.e. n = k, in which
case the integral is 2πi, as discussed in Observation 2 below Corollary 6.1. Hence,

f (z)
I
2πick = k+1
dz, (6.61)
γ (z − z0 )

from which equation (6.58) immediately follows.

Often, however, equation (6.58) is not of immediate use and it is often easier to resort to Taylor’s theorem
and the binomial expansion as seen in the examples below.

Example 6.13 Find the Laurent series for

sin(z)
f (z) = (6.62)
z2
around z = 0.

Example 6.14 Find the Laurent series for

1
f (z) = (6.63)
(z − 1)(z − 2)

around z = 0.
80

Example 6.15 [G. James, Engineering Mathematics, Example 1.21, p.50] Determine the Laurent series ex-
pansion of the function
f (z) = z 3 exp(1/z) (6.64)
about

1. z = 0;
2. z = a, with a a finite, non-zero complex number;
3. z = ∞.

6.7 Singularities and residues.

6.7.1 Introduction
In the previous section we have seen that points where the integrand is singular, e.g. is not defined, can lead be
used to connect the value of the function there to a path integral, see for example Cauchy’s Integral Formula,
Theorem 6.8. We would now like to expand on this point in order to obtain some very powerful tools to evaluate
complex and real integrals. The first step on this journey is to defined what is meant for a function to be singular
and then how to classify different types of singularities.

Definition 6.9 (Regular and singular points) A point a is a regular point of f (z) if f is analytic at a. The
point a is a singularity of f (z) if a is a limit point of regular points which is not itself regular.

Definition 6.10 (Isolated singularity) If f (z) is holomorphic in a region 0 < |z − a| < R for some R > 0,
but not at a, then f has an isolated singularity at a.

If there exists no R > 0, however small, such that f (z) is analytic on {z : 0 < |z − a| < R}, then a is
called a non-isolated essential singularity, but we shall not be dealing with these. An example of this singular
behaviour is the function f (z) defined as
X ∞
f (z) = zn.
n=0
This converges for all |z| < 1. Hence any point a, with |a| = 1, i.e. on the circumference of radius 1, is a
singularity of f (z), according to the definition 6.9. However, there is no disk around a, however small, where
f (z) is holomorphic. Hence a is non-isolated.

6.7.2 Types of isolated singularities

− a)n for f near
P∞
There are three types of isolated singularities, distinguished by the Laurent series −∞ cn (z
a (which exists by Theorem 6.12, with R1 = 0 and R2 = R).

Removable singularity
If there are no negative powers of z − a in the Laurent series, that is, cn = 0 for all n < 0, then
∞
X
f (z) = cn (z − a)n → c0 as z → a, (6.65)
n=0

so by defining (or redefining) f (a) = c0 we obtain a function which is holomorphic at a (by Theorem 6.10,
since it is represented by a convergent power series near a), and hence holomorphic for |z − a| < R.

Example 6.16 Let f (z) = z for z 6= 0, and f (0) = 1. The singularity at 0 can be removed by redefining
f (0) = 0.

Example 6.17 Determine the singularities of f (z) = sin(z)/z.

 81

Poles
Definition 6.11 (Pole) A singular point a ∈ C is called a pole of f (z) if the Laurent series expansion of f (z)
at a has only a finite number of negative powers. The order m of the pole is defined as the exponent of the most
negative power with non–zero coefficient:
m = max{n ∈ N : c−n 6= 0} . (6.66)

Equivalently, the function (z − a)m f (z) is holomorphic near a, and has a non-zero limit c−m as z → a.
A pole of order m = 1 is called a simple pole. A pole of order 2 is called a double pole. Poles of order
three and higher are usually referred to as such, although can be labelled triple etc.

Example 6.18 If k ∈ N then the function f (z) = 1/z k has a pole of order k at 0.

Example 6.19 The function f (z) = cosec(πz) has simple poles at the points a ∈ Z, where sin(πz) =
1/cosec(πz) has simple zeros.
ez
Example 6.20 Determine the singularities of f (z) = .
z2

Essential singularities
Here there are infinitely negative powers of z − a in the Laurent series, and it can be shown that |f (z)| has no
limit as z → a.

Example 6.21 The function

∞
X 1 −n
f (z) = exp(1/z) = z (6.67)
n!
n=0
is holomorphic at all z 6= 0, but has an essential singularity at 0. If z = x ∈ R then |f (z)| → ∞ as x → 0+,
but |f (z)| → 0 as x → 0−. Most essential singularities arising in this module are related to this case; for
example, sinh(1/z) has analogous behaviour as z → 0.

6.7.3 Classification of singularities

One often needs to find and classify the singularities of a complex function when computing an integral over
the function. Denote the complex function f (z) and suppose that it can be expressed as the ratio of two complex
functions g(z)/h(z) (almost all functions appearing in this module can be expressed in this way.) If either g(z)
or h(z) have an essential singularity at z0 then f (z) has an essential singularity. If this is not the case then we
can classify the singularities of f (z) in an algorithmic manner. A sample of possible cases is as follows:

1. If g(z0 ) = 0 and h(z0 ) is non-zero (finite) then f (z) has a zero at z = z0 .

2. If g(z) is singular at z0 and h(z0 ) 6= 0 then the singularity of f (z) is the same as the singularity of g(z).
3. If h(z) is singular at z0 and g(z0 ) 6= 0 then f (z) has a removable singularity at z0 .
4. If h(z0 ) = 0 and g(z0 ) 6= 0, then f (z) has a pole of order m, where m is the order of the zero of h(z) at
z0 . In particular, if h(z) is a polynomial then m is the multiplicity of the zero at z0 . Otherwise, it is the
order of the lowest non–zero power of the Taylor expansion of h(z) at z0 .
5. If both g(z) and h(z) are zero at z0 then f (z) is singular there:
(a) The singularity is removable if
g(z)
lim = f0 with |f0 | < +∞. (6.68)
z→z0 h(z)
In this case we can make f (z) holomorphic in a neighbourhood of z0 by defining f (z0 ) = f0 .
82

(b) The singularity is a pole of order m ≥ 1, where

m = max{n ∈ N : lim (z − z0 )n f (z) 6= 0.} (6.69)

z→z0

Alternatively, m = mh − mg , where mg and mh are, respectively, the lowest order terms of the
Taylor expansions of g(z) and h(z) at z = z0 .

Example 6.22 Classify the singularities of

ez
f (z) = . (6.70)
(z 2 + 1)

Example 6.23 Classify the singularities of

1
f (z) = , (6.71)
(z − a)k

with a ∈ C and k a positive integer.

Example 6.24 Classify the singularities of

cos( 12 πz)
f (z) = . (6.72)
(z − 1)

6.7.4 Residues
In this section we introduce the concept of residue and explain how to compute its value. This is an essential
ingredient for the evaluation of complex integrals, as we will see in the next section.

Definition 6.12 (Residue) If f has an isolated singularity at a, then the residue res(f, a) of f at a is the
coefficient c−1 of (z − a)−1 in the Laurent series of f around a.

Note that a removable singularity has no residue, since c−1 = 0 as part of c−1 = c−2 = c−3 = . . . = 0.
Both poles and essential singularities have residues since c−1 6= 0 in general, although we can have a residue
of zero in these cases (but recall that non-removable singularities would always contain at least one non-zero
negative coefficient in their Laurent expansion).
It is important for us to calculate residues quickly as they are used a great deal in complex integration. If
one knows the Laurent series, then one can immediately extract the residue. However, in general this is a very
cumbersome procedure. The following theorem offers a much quicker route:

Theorem 6.13 If f has a pole of order m ≥ 1 at a then

1 dm−1
res(f, a) = lim m−1 (z − a)m f (z). (6.73)
(m − 1)! z→a dz

Proof - The function

(z − a)m f (z) = c−m + c−m+1 (z − a) + · · · + c−1 (z − a)m−1 + c0 (z − a)m · · · (6.74)

is holomorphic near a, and its (m − 1)th derivative,

m!
(m − 1)! c−1 + c0 (z − a) + · · · , (6.75)
1!
has limit (m − 1)! c−1 as z → a.
 83

Observation 1 - When m = 1 (simple poles) equation (6.73) simplifies to

res(f, a) = lim (z − a)f (z). (6.76)

z→a

When m = 2 (double poles) this simplifies to

d 
(z − a)2 f (z) .

res(f, a) = lim (6.77)
z→a dz

Observation 2 - In the case of essential singularities the formula given in Theorem 6.13 does not apply, and
the only way to obtain the residue is by deriving the Laurent series (up to c−1 ).

Example 6.25 Classify the singularities and find the corresponding residues of the function

(z + 1) exp(z)
f (z) = . (6.78)
(z − 1)3

z2
Example 6.26 Find the residues of all the real singularities of f (z) = (z−2)(z 2 +1)
.

1
Example 6.27 Find the residue of f (z) = z 2 (z+2)3
at z = −2.

cos(z) + z 3
Example 6.28 Find the residue of all singularities of f (z) = .
sin(z)

6.8 Cauchy’s Residue Theorem

In this section we apply residues to evaluate line integrals. The first step is to evaluate residues using line
integrals:

Theorem 6.14 If γ is a closed contour in a region 0 < |z − a| < R where f is holomorphic, and γ has winding
number 1 around a, then
1
Z
res(f, a) = f (z) dz. (6.79)
2πi γ

Proof - This is a direct consequence of equation (6.58), page 79, with k = −1.
The following theorem is fundamental for the evaluation of complex (and, ultimately, real) integrals. It
generalises theorem 6.14 to the case where γ encloses any finite number of singularities:

Theorem 6.15 (Cauchy’s Residue Theorem) If f (z) is holomorphic inside and on a simple closed contour
γ, except for a finite set of isolated singularities at points a1 , . . . , an inside γ, then
Z n
X
f (z) dz = 2πi res(f, ak ). (6.80)
γ k=1

Proof - For each k = 1, . . . , n let γk be a small circle inside γ enclosing ak but no other singularities, and
let Pk be a path inside γ joining γ to γk ; we can arrange that these circles and paths do not meet each other
except where each Pk meets γk (see figure 6.5 for an example of such a contour). By travelling around γ in
the positive direction, and for each k taking a detour along Pk , around γk in the negative direction, and back
along Pk to γ, we obtain a closed contour K such that f is holomorphic inside and on K. Cauchy’s Theorem
therefore implies that Z
f (z) dz = 0. (6.81)
K
84

ℑ(z)

−P2 P2
γ2
a2 γ
P1
a1
−P1
γ1
a3
γ3

−P3 P3

ℜ(z)

Figure 6.5: ”Proof” of the residue theorem: the function f (z) is holomorphic in the shaded area and, hence, its
integral over the path created by joining γ with all the circles γk and straight paths ±Pk is zero.

This integral is the sum of the integrals around γ, along each Pk in either direction, and around each γk . For each
k the integrals along Pk in either direction cancel by Theorem 6.1, and the integral around γk is −2πi res(f, ak )
by Theorems 6.1 and 6.14, so
Z Z Xn
f (z) dz = f (z) dz − 2πi res(f, ak ), (6.82)
K γ k=1

giving the result.

Observation - When evaluating integrals using Cauchy’s Residue Theorem, one often makes use of the De-
formation Theorem, i.e. by Cauchy’s Theorem the integral remains unchanged when the contour is changed,
provided that the singularities inside the contour remain unchanged. Figure 6.6 illustrates on the left several
contours which contain no singularities, and therefore the integral around all such contours is zero. On the right
of Figure 10 are shown two contours which are equivalent because they both contain the same singularities. If
an explicit parameterisation of a contour is needed, then a symmetric contour such as a circle will in general be
easier to work with than an irregular contour.
Example 6.29 Evaluate
dz
Z
, (6.83)
γ (z − 1)(z − 2)
where γ is a circle with centre at the origin and radius (a) 1/2, (b) 3/2 and (c) 5/2.
Example 6.30 Evaluate the integral
z+1
Z
IR = dz, (6.84)
CR z3 − 1
where CR is the circle |z| = R of radius R > 1 centred at 0.

6.9 Evaluation of real integrals

Residues can be used to evaluate several types of real integrals over an interval. In this module we study two
cases: integrals of rational functions over the interval [0, 2π] and integrals of real functions over an infinite or
semi-infinite real line.
 85

ℑ(z) ℑ(z)

γ3 γ3

γ2 γ2

γ1 γ1

ℜ(z) ℜ(z)

Cauchy’s Theorem is independent of the con- Cauchy’s Residue Theorem is independent of

tour, provided it remains inside an analytic re- the contour provided the singularities inside re-
gion. main unchanged.
Figure 6.6: How changing the integration path affects either Cauchy’s (left) or Cauchy’s Residue (right) the-
Rorems. In bothRcases the diamonds
R represents isolated singularities of an otherwise holomorphic function and
γ1 f (z) dz = γ2 f (z) dz =
6 γ3 f (z) dz.

Observation - Residues, cannot, however, be used to find primitives of real functions, i.e. to solve indefinite
integrals.

6.9.1 Integrals of rational functions in [0, 2π]

Let Z 2π
I= Q[cos(θ), sin(θ)] dθ, (6.85)
0
where Q is a rational function2 of cos(θ) and sin(θ). Write
z + z −1 z − z −1
cos(θ) = and sin(θ) = , (6.86)
2 2i
where
z = eiθ =⇒ dθ = −i dz. (6.87)
Substituting for cos(θ) and sin(θ) in equation (6.85), we can write the integral as
Z
I = P (z) dz, (6.88)
γ

where P is a rational function of z and γ is the unit circle |z| = 1. A rational function is holomorphic
everywhere apart from finitely many poles, at the zeros of its denominator. If none of these lies on γ then it
follows from Cauchy’s Residue Theorem applied to equation (6.88) that
n
X
I = 2πi res(P, ak ), (6.89)
k=1

where a1 , . . . , an are the poles of P enclosed by γ, that is, with |ak | < 1.

Example 6.31 Use complex variable methods to evaluate

Z 2π
dθ
I= . (6.90)
0 3 − cos(θ)
2
A rational function is the ratio of two polynomials in its argument. In this section Q is the ratio of two polynomials in cos(θ) and
sin(θ).
86

6.9.2 Integrals on infinite subsets of the real axis

We can also use residues to evaluate real integrals of the form
Z +∞ Z R Z 0
I= f (x) dx = lim f (x) dx + lim f (x) dx , (6.91)
−∞ R→∞ 0 R→∞ −R

or Z ∞ Z R
I= f (x) dx = lim f (x) dx. (6.92)
0 R→∞ 0

RNotation
+∞
- All the functions that will be considered in this section converge absolutely, in the sense that
−∞ |f (x)| dx < ∞. In this case it is possible to use, instead of the formal definition given by equation (6.91),
the more compact notation
Z +∞ Z R
I= f (x) dx = lim f (x) dx . (6.93)
−∞ R→∞ −R

We will use this notation for the rest of these notes.

Whether we are dealing with integrals like (6.93), (6.92) or others of similar type, the procedure is as follows:

1. We start by considering the integral for a finite value of R, for example

Z R
IR = f (x) dx, (6.94)
−R

as the integral in the complex plane of complex function f (z) on the path γR that lies along the real axis
and joins −R to R.

2. We then construct additional paths γ1 , γ2 , . . . , γn such that γR = γR ∪ (∪ni=1 γi ) is a simple closed path.

3. We can then apply the residue theorem and write

Z N ′
X
f dz = 2πi res(f, ak ) =⇒
γR k=1
′
(6.95)
Z R N
X Z
=⇒ f (x) dx = 2πi res(f, ak ) − f dz,
−R k=1 ∪n
i=1 γi

where ak , k = 1, 2, . . . , N ′ , with N ′ depending on R, are the isolated singularities of f contained within

γR .

4. Finally, we take the limit R → ∞ of equation (6.95) to evaluate I:

N
X Z
I = 2πi res(f, ak ) − lim f dz, (6.96)
R→∞ ∪n γi
k=1 i=1

where ak , k = 1, 2, . . . , N are the isolated singularities of f contained within the (infinite) closed path
limR→∞ γR .

Clearly this method is advantageous only if the evaluation of the integrals on the right hand side of equa-
tion (6.96) is trivial. Often, this happens either because they are zero in the limit R → ∞ or because they end
up being a multiple of I.

Example 6.32 Evaluate

+∞
x2 + 1
Z
I= dx. (6.97)
−∞ x4 + 1
 87

6.10 Problem solving

In this module we have learnt methods to compute the following types of integrals:

1. Integrals over the complex plane: Z

Iγ = f (z)dz, (6.98)
γ

where γ is a closed contour in the complex plane and f (z) is a function with isolated singularities. If
f (z) has no singularities, then the integral vanishes by Cauchy’s theorem. Functions with non-isolated
singularities and so-called multifunctions such as logarithms have not been considered here.

2. Real integrals of rational periodic functions:

Z 2π
I2 = Q[cos(θ), sin(θ)]dθ, (6.99)
0

where Q is a rational function of its arguments. Substituting z = exp(iθ) the integral can be mapped to
an integral of a rational function P (z) over the unit circle γ in the complex plane:
Z
I2 = P (z)dz (6.100)
γ

which is a special case of type 1.

3. Real integrals over the infinite real line:

Z ∞
I3 = f (x)dx (6.101)
−∞

This integral can be related to an integral over the complex plane, for which the contour γR runs along
the real axis and is then closed with a semicircle in the upper half plane:
Z
I3 = f (z)dz (6.102)
γR

provided that the contribution from the semicircle can be shown to vanish using the Estimation Lemma.
Here f (z) is the function f (x) promoted to the complex plane, i.e. we replace x by z in the function.
The integral I3 is again a special case of type 1.

4. Real integrals over the positive real line:

Z ∞
I4 = f (x)dx (6.103)
0

where f (x) is an even function, i.e. f (−x) = f (x). In such a case

1 ∞
Z
I4 = f (x)dx (6.104)
2 −∞

and therefore can be computed as for type 3.

5. Other types of real integrals over the real or the positive real line in which the contour in the complex
plane is closed using paths on which the integral is a multiple of the integral on the real line. These are
all special cases and have to be dealt with individually (see problem sheet).

We now outline a problem solving strategy for integrals of this type:

88

1. First identify which of the four types of integrals you are being asked to compute. If it is a real integral,
then it needs to be mapped to a complex integral as outlined above. In case of type (2) be very careful
with your algebra: errors in the rational function you obtain may make your later calculation much harder
than it should be!

2. Once you have identified the integral over the complex plane, draw a sketch of the required integration
contour. For type (2) this will always be the unit circle and for type (3) and (4) it will be the semi-circle
in the upper or lower half of the complex plane. There is no definite rule for type (5).

3. All types will be reduced to particular cases of type (1), for a given function f (z) with isolated singulari-
ties. The next step is to find and classify the singularities of the function. You should start by finding the
locations of the singularities and working out which ones lie inside the integration contour. Singularities
which lie outside the integration contour cannot contribute to the integral so you should not waste time
computing their residues! For singularities inside the integration contour, identify their classification:
degree of pole or essential singularity.

4. Compute the residues of singularities inside the integration contour. For essential singularities you will
need to work out the Laurent series but for poles use the formulas given earlier in the notes, equa-
tion (6.73) and the special cases (6.76) and (6.77).

5. Apply Cauchy’s Residue Theorem, Theorem 6.15, to compute the integral, summing over the residues of
the poles within the contour.

6. For types (3) and (4), and possibly (5), use the Estimation Lemma, Theorem 6.4, to argue that the contri-
bution from the semicircle or other additional paths vanishes.

7. Make as many checks on your answer as possible. For example, in all cases where the integral you are
computing is real, the answer must be real: all factors of i must cancel in your final answer. Also, if the
integrand is sign definite, e.g. it is positive, the integral must have the same sign.

Example 6.33 Compute the integral

dz
Z
I= , (6.105)
γ (z 2 − 2az + 1)
where a is real, a > 1 and γ is the unit circle.

Example 6.34 Compute the integral

2π
dt
Z
I2 = (6.106)
0 a − cos(t)
for real a, a > 1.
Appendix A

Useful formulae & Revision material

Functions and Multifunctions

Identities:
cosh(iz) = cos(z), sinh(iz) = i sin(z). (A.1)
Periodic functions:
dz 1 1
If z = eiθ then dθ = and cos(nθ) = (z n + z −n ), sin(nθ) = (z n − z −n ). (A.2)
iz 2 2i
Complex logarithm:
log(z) = log(|z|) + i arg(z). (A.3)

Power Series
∞
k k(k − 1) 2 X k(k − 1) · · · [k − (n − 1)]
(1 + z) = 1 + kz + z + ... = 1 + zn, |z| < 1
2! n!
n=1
∞
z2 X zn
ez = 1+z+ + ... = , z∈C
2! n!
n=0
∞
z2 z3 X zn
log(1 + z) = z− + + ... = (−1)n+1 , |z| < 1
2 3 n
n=1
∞
z2 z4 X z 2n
cos(z) = 1− + + ... = (−1)n , z∈C
2! 4! (2n)!
n=0
∞
z3 z5 X z 2n+1
sin(z) = z− + + ... = (−1)n , z∈C
3! 5! (2n + 1)!
n=0
∞
z2 z4 X z 2n
cosh(z) = 1+ + + ... = , z∈C
2! 4! (2n)!
n=0
∞
z3 z5 X z 2n+1
sinh(z) = z+ + + ... = , z∈C
3! 5! (2n + 1)!
n=0

Residues
Pole of order m:
1 dm−1
res(f, a) = lim m−1 [(z − a)m f (z)] . (A.4)
(m − 1)! z→a dz
Formulae valid only for a simple pole :
res(f, a) = lim (z − a)f (z). (A.5)
z→a

g(z) g(a) 6= 0, g(a)

If f (z) = , and g(z) analytic, then res(f, a) = . (A.6)
h(z) h(a) = 0, h′ (a)

89
90

Contour Integration
Parameterisation of contours:
Z Z b
If γ = {z(t) : a ≤ t ≤ b} then f (z) dz = f [γ(t)] γ ′ (t) dt.
γ a

For a circle of radius R, centred at the origin:

Z Z 2π
it
f Reit iReit dt.


γ = z(t) = Re : 0 ≤ t < 2π and f (z) dz =
γ 0

Cauchy’s Residue Theorem: Z X

f (z) dz = 2πi res(f, ak ). (A.7)
γ Singularities ak
of f (z) inside γ.

Main complex variable theorems

The theorem numbers are those given in the lecture notes. You should memorise the following theorems for the
exam.

Theorem 6.4, p. 72 [Estimation Lemma] - If |f (z)| ≤ M for all z on a contour γ of length L then
Z 
 
 f (z) dz  ≤ M L.
 
γ

Theorem 6.5, p. 72 [Cauchy’s Theorem] - If f is holomorphic inside and on a closed contour γ, then
Z
f (z) dz = 0.
γ

Theorem 6.9, p. 75 [Cauchy’s integral formula for the nth derivative] - If f is holomorphic inside and on a
simple closed contour γ, and a is inside γ, then f is n times differentiable at a for each n ≥ 0, with
n! f (z)
Z
(n)
f (a) = dz.
2πi γ (z − a)n+1

Theorem 6.11, p. 76 [Taylor’s Theorem] - If f is holomorphic on a disc |z − z0 | < R, then for all such z we
have
∞
X f (n) (z0 )
f (z) = (z − z0 )n .
n!
n=0
This power series is called the Taylor expansion of f (z) around z0 .

Theorem 6.12, p. 79 [Existence of the Laurent series] - If f is holomorphic in an annulus R1 < |z −z0 | < R2
then f is represented by a Laurent series in this region, i.e.
+∞
X
f (z) = cn (z − z0 )n ,
n=−∞

for R1 < |z − z0 | < R2 . The Laurent series in a given annulus is unique.

Theorem 6.15, p. 75 [Cauchy’s Residue Theorem] - If f (z) is holomorphic inside and on a simple closed
contour γ, except for a finite set of isolated singularities at points a1 , . . . , an inside γ, then
Z Xn
f (z) dz = 2πi res(f, ak ).
γ k=1
Module Overview
This module aims to provide you with some of the essential mathematical tools to model and understand
the world around us. Vector calculus is essential to describe real and virtual objects that move in space and
time. Complex variables are often the natural setting for many seemingly unrelated problems, from numerical
methods to integral transforms and the study of the response of dynamical systems.
In the first part of this module we build on multivariate calculus studied in the first year and extend it to the
calculus of scalar and vector functions of several variables. Line, surface and volume integrals are considered
and a number of theorems involving these integrals (named after Gauss, Stokes and Green) will be discussed.
In particular Green’s theorem, which gives a formula for the line integral of a vector field in the plane round
a closed curve, is closely related to complex integration considered in the second part of the module. The
integral theorems are also useful in many branches of Applied Mathematics and to describe physical quantities
that vary in space and in time. For example, this module is a pre-requisite for MATH2044, Fields and fluids,
where these methods are used to describe the behaviour of fluids and of electromagnetic fields.
In the second part of this module, we extend our investigation of calculus to functions of a complex variable,
once again building on the material studied in the first year. This theory has both great aesthetic appeal and a
large number of applications. We focus here on the integration of these functions, particularly along curves in
the complex plane. We develop the basic theory and ideas of the integration of a function of a complex variable,
use the main theorems such as Cauchy’s theorem and the Cauchy integral formula, and explore some of their
consequences, such as the evaluation of real integrals.