MA 231 Vector Analysis

Stefan Adams

2010, revised version from 2007, update 02.12.2010

Contents
1 Gradients and Directional Derivatives 1

2 Visualisation of functions f : Rm → Rn 4
2.1 Scalar fields, n = 1 . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Vector fields, n > 1 . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Curves and Surfaces . . . . . . . . . . . . . . . . . . . . . . . 10

3 Line integrals 16
3.1 Integrating scalar fields . . . . . . . . . . . . . . . . . . . . . . 16
3.2 Integrating vector fields . . . . . . . . . . . . . . . . . . . . . 17

4 Gradient Vector Fields 20

4.1 FTC for gradient vector fields . . . . . . . . . . . . . . . . . . 21
4.2 Finding a potential . . . . . . . . . . . . . . . . . . . . . . . . 26
4.3 Radial vector fields . . . . . . . . . . . . . . . . . . . . . . . . 30

5 Surface Integrals 33
5.1 Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2 Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.3 Kissing problem . . . . . . . . . . . . . . . . . . . . . . . . . . 40

6 Divergence of Vector Fields 45

6.1 Flux across a surface . . . . . . . . . . . . . . . . . . . . . . . 45
6.2 Divergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

7 Gauss’s Divergence Theorem 52

8 Integration by Parts 59

9 Green’s theorem and curls in R2 61

9.1 Green’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 61
9.2 Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

10 Stokes’s theorem 72
10.1 Stokes’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . 72
10.2 Polar coordinates . . . . . . . . . . . . . . . . . . . . . . . . . 78

11 Complex Derivatives and Möbius transformations 80

12 Complex Power Series 89

13 Holomorphic functions 99

14 Complex integration 103

15 Cauchy’s theorem 107

16 Cauchy’s formulae 117

16.1 Cauchy’s formulae . . . . . . . . . . . . . . . . . . . . . . . . 117
16.2 Taylor’s and Liouville’s theorem . . . . . . . . . . . . . . . . . 120

17 Real Integrals 122

18 Power series for Holomorphic functions 126

18.1 Power series representation . . . . . . . . . . . . . . . . . . . . 126
18.2 Power series representation - further results . . . . . . . . . . 128

19 Laurent series and Cauchy’s Residue Formula 130

19.1 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
19.2 Laurent series . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

ii
Preface
Health Warning: These notes give the skeleton of the course and are not a
substitute for attending lectures. They are meant to make note-taking easier
so that you can concentrate on the lectures. An important part in vector
analysis are figures and pictures. These will not be contained in these notes.
For any figure which appears on the blackboard in my lectures I leave some
empty space with a reference number which coincides with the number I am
using in the lectures. You can fill the diagrams and figures by your own.
These notes grew out of hand written notes from Jochen Voß who gave
this course 2005 and 2006. I thank him very much for letting me using his
notes.
Any remarks and suggestions for improvements would help to create better
notes for the next year.

Stefan Adams

Motivation
What is Vector Analysis?
In analysis differentiation and integration were mostly considered in one di-
mension. Vector analysis generalises this to curves, surfaces and volumes in
Rn , n ∈ N. As an example consider the “normal” way to calculate a one
dimensional integral: You may find a primitive of a function f and use the
fundamental theorem of calculus, i.e. for f = F 0 we get
Z b
f (x) dx = F (b) − F (a).
a

The value of the integral can be determined by looking at the boundary

points of the interval [a, b]. Does this also work in higher dimensions? The
answer is given by Gauss’s divergence theorem.

Notation
One of the main problems in vector analysis is that there are many books
with all possible different notations. During the whole course I outline alter-
native notations in use. It is one of the objectives to acquaint you with the
different notations and symbols. Note that most of the material originated
from physics and hence many books are using notations and symbols known
by people in physics.

iii
 p
Vectors: x ∈ Rn with x = (x1 , . . . , xn ) and with norm kxk = x21 + · · · + x2n .
Alternative notations are: ~x, x, x and in dimension n = 3 ~r = (x, y, z) for
the vector and |x| or r for the norm of ~x, x, x or ~r.
Properties of the norm are (i)kxk ≥ 0; (ii)kλxk = |λ|kxk for λ ∈ R; (iii)kx +
yk ≤ kxk + kyk for x, y ∈ Rn .
Functions: f : Rm → Rn with component functions f1 , . . . , fn : Rm → R.
Alternative notations are: f~ or f or f .
Partial derivatives: Let ei , 1 ≤ i ≤ n, be the canonical basis vectors in Rn
(i.e. hei , ej i for i, j+1, . . . , n). The partial derivative of a function f : Rn → R
with respect to the i-th direction at point x ∈ Rn is given by

∂f (x) f (x + hei ) − f (x)

= lim , x ∈ Rn .
∂xi h→0 h
Alternative notation is: ∂i f (x).
Scalar
Pn product: The scalar product (dot product) is defined as hx, yi =
n
i=1 i i for x, y ∈ R . Alternative expression is x·y. Recall that hx, yi =
x y
kxkkyk cos θ where θ is the angle between the two vectors x and y. Note that
kyk cos θ is the component of the vector y in direction of the vector x.

iv
Part I: Real Vectoranalysis
1 Gradients and Directional Derivatives
In this section we ask how does a function f : Rn → R change when we move
from x ∈ Rn in direction of a vector y ∈ Rn .

Figure 1

We can reduce the problem to one dimension. We define for the given x ∈ Rn
and y ∈ Rn the function
ϕ(x) : R → R
t 7→ ϕ(x) (t) = f (x + ty).
The change of f in direction y equals the change of ϕ(x) and is thus given by
ϕ0(x) .
Definition 1.1 The directional derivative Dy f (x) of a function f : Rn →
R at the point x ∈ Rn in direction of y ∈ Rn is given by

f (x + ty) − f (x)
Dy f (x) = lim
t→0 t

if the indicted limit exists. Dy f (x) is also called the derivative along the
vector y at the point x.

1
Remark 1.2 (a) With ϕ(x) (t) = f (x + ty) the directional derivative of f
at the point x ∈ Rn in direction y ∈ Rn is given as

Dy f (x) = ϕ0(x) (0).

(b) We have defined the directional derivative via the limit as t → 0 for
the differential quotient. If one takes the limit t ↓ 0, that is t > 0 and
t → 0, one gets the directional derivative from the right hand side. The
same applies to the limit t ↑ 0 for the directional derivative from the
left.

We calculate directional derivatives in the next example.

Example 1.3 Let f : R2 → R, f (x) = x21 + x22 be given. Then for x, y ∈ R2

we have
f (x + ty) = (x1 + ty1 )2 + (x2 + ty2 )2
= x21 + x22 + 2t(x1 y1 + x2 y2 ) + t2 (y12 + y22 )

Hence Dy f (x) = ϕ0(x) (0) = 2hx, yi.

Figure 2

2
Recall ϕ(x) (t) = f (x + ty), t ∈ R, x, y ∈ Rn . The function ϕ(x) : R → R is
the composition of the function g : R → Rn , t 7→ x + ty, and the function
f : g(R) → R (i.e. the restrition of f on g(R) ⊂ Rn ), that is

ϕ(x) (t) = (f ◦ g)(t) = f (g(t)).

The chain rule gives

n
X
0
(f ◦ g) (t) = ∂i f (g(t))gi0 (t)
i=1
n
X
= ∂i f (x + ty)yi ,
i=1

where gi (t) = xi + tyi , t ∈ R, 1 ≤ i ≤ n, are the component functions of g.

Hence we get
n
X
0
Dy f (x) = ϕ(x) (0) = ∂i f (x)yi . (1.1)
i=1

We can write (1.1) in a shorter way with the following definition.

Definition 1.4 Let the function f : Rn → R be differentiable. The map-

ping ∇f : Rn → Rn , x 7→ ∇f (x) = (∂1 f (x), . . . , ∂n f (x)) is called the gra-
dient mapping and the vector ∇f (x) = (∂1 f (x), . . . , ∂n f (x)) is called the
gradient of f at the point x. Alternative notations are grad f, ∇f , ∇f or
∇f = ni=1 ∂x∂f
P
i
ei .

Note that ∇f : Rn → Rn is a “vector” of the component functions ∂i f : Rn →

R, x 7→ ∂i f (x), 1 ≤ i ≤ n.

Definition 1.5 A scalar or a vector quantity is said to be a field if it is a

function of the spatial position. Examples: Let D ⊂ Rm , then f : D → Rn is
called a vector field if n > 1, and f : D → R is called a scalar field.

Examples for vector fields are the magnetic, the electric or the velocity (vec-
tor) field, whereas temperature and pressure are scalar fields.
Our calculation in (1.1) shows that the directional derivative Dy f at any
point x ∈ Rn is linear in y and we only need to know the gradient ∇f (x) in
order to calculate Dy f (x)

Dy f (x) = h∇f (x), yi (1.2)

3
Remark 1.6 Recall the Cauchy-Schwarz inequality

|hx, yi| ≤ kxkkyk for x, y ∈ Rn .

We get
|Dy f (x)| = |h∇f (x), yi| ≤ k∇f (x)kkyk,
and for y ∈ Rn with kyk = 1 we have |Dy f (x)| ≤ k∇f (x)k. Assume that
∇f (x)
∇f (x) 6= 0 and pick y = k∇f (x)k
with kyk = 1. This gives

∇f (x)
Dy f (x) = ∇f (x), = k∇f (x)k.
k∇f (x)k

We conclude that ∇f (x) (under the assumption that ∇f (x) 6= 0) points in

the direction along which f is increasing the fastest. Alternatively you may
prove this via h∇f (x), yi = k∇f (x)k cos θ for any unit vector y having an
angle θ with ∇f (x).

2 Visualisation of functions f : Rm → Rn
Visualisation of functions is hard and needs practice. One of the objectives
of the course is to learn basic techniques to create sketches/figures showing
basic features of a given function.

2.1 Scalar fields, n = 1

We recall the notion of a graph of a function and introduce the notion of a
level set of a function. The graph of a function f : Rm → Rn is easily pictured
for m = n = 1 and m = 2, n = 1, see Figure 3.

4
 Figure 3:

Definition 2.1 Let f : D → R be a function for some domain D ⊂ Rm .

(a) The graph of f is the subset of Rm+1 consisting of all points (x, f (x)), x ∈
D. In symbols,

graph f = {(x, f (x)) ∈ Rm+1 : x ∈ D}.

(b) Let c ∈ R. The set

f −1 (c) = {x ∈ D : f (x) = c}

is called the c-level set of the function f . In dimension m = 2 we

speak also of a level curve and in dimension m = 3 of a level surface.

The behaviour or structure of a function is determined in part by the

shape of its level sets; consequently, understanding these sets is of great help
understanding the functions in question. The idea of level sets is also used in
drawing contour maps, where one draws lines to represent constant altitude.
Walking along such a line would mean walking on a level path. In the case
of a hill rising from the x − y plane, a graph of all level curves gives us a
good idea of the ’height’ function h(x, y), which represents the height of the
hill at point (x, y).

5
Example 2.2 (a) f : R2 → R, (x, y) 7→ f (x, y) = x2 + y 2 .

Figure 4:

(b) f : R3 → R, (x, y, z) 7→ f (x, y, z) = z 2 − x2 − y 2 .

Let c = 0: f −1 (0) is a cone,

f (x, y, z) = 0 ⇔ r2 := x2 + y 2 = z 2 ⇔ r = |z|,
p
where r = x2 + y 2 .

Let c > 0:
√ p
f (x, y, z) = c ⇔ r = z 2 − c with z 2 ≥ c ⇔ z = ± x2 + y 2 + c.

Hence the level set is a hyperboloid of two

√ sheets around the z axis, passing
through the z axis at the points (0, 0, ± c).

6
 Figure 5: cone

Figure 6: hyperboloid (two sheets)

Let c < 0:
√ p
f (x, y, z) = c ⇔ r = z2 − c ⇔ z = ± x2 + y 2 − c.

7
The level set (surface) is the single-sheeted hyperboloid of revolution
√ around
the z axis, which intersects the x − y plane in the circle of radius −c.

Figure 7: hyperboloid (single-sheet)

2.2 Vector fields, n > 1

We are going to sketch vector fields. As an example think about the velocity
vector field inside a fluid or the electric and magnetic field of power currents.
1
2 2
Example 2.3 (a) f : R → R , (x, y) 7→ f (x, y) = 2 , see figure 8 below.
0

8
 Figure 8:

 
2 2 x
(b) f : R → R , (x, y) 7→ f (x, y) = , see figure 9 below.
y

Figure 9:

9
  
2 2 y
(c) f : R → R , (x, y) 7→ f (x, y) = , see figure 10 below.
x

Figure 10:

Vector fields in two dimensions can be visualised by a sketch. In this case

the simplest procedure is to evaluate the vector field at a sequence of points
and draw vectors indicating the magnitude and direction of the vector field
at each point. An example of this procedure is the drawing of wind speeds
and directions on weather maps. In the above example 2.2 the third vector
field at point is f (1, 0) = (0, 1), so at this point a vector of magnitude 1 in
the y-direction is drawn. By considering a few additional points, a sketch of
the vector field can be built up (see figure 10).

2.3 Curves and Surfaces

We study now curves and surfaces, i.e. we consider m < n. We consider
mainly two examples, m = 1, 2. For m = 1 we have parametric curves in Rn
and for m = 2 we have parametric surfaces in Rn . Parametric curves and
surfaces are defined via mappings from subsets of R respectively subsets of
R2 into Rn . In the next example we study the case n = 3. That is we have no
axis available for the arguments of the mappings, we only sketch the range
of these mappings.

10
Example 2.4 (a)
 
cos t
f : [0, 2π] → R3 , t 7→ f (t) =  sin t  .
t

This defines the helix seen in the figure 11 below.  

cos t
Note that for c ∈ R the mapping fc : [0, 2π] → R3 , t 7→ fc (t) =  sin t 
c
defines a circle line in the x − y plane shifted in z direction by c.

(b)  
s cos t
f : [0, 1] × [0, 2π] → R3 , (s, t) 7→ f (s, t) =  s sin t  .
t
This mapping defines the helicoid seen in figure 12.

Note that for fixed parameter t we have lines within the surface of the helicoid
and through any point of the helicoid there is a helix going through that
point. 

Curves can be given in two ways. A parametric curve is a map ϕ : [a, b] →

R , e.g. ϕ(t) = (cos t, sin t) ∈ R2 for t ∈ [0, 2π]. Curves in Rn , i.e. subsets
n

C ⊂ Rn , can be given as the level set of some real valued function, e.g.
consider the function f : R2 → R, (x, y) 7→ f (x, y) = x2 + y 2 . The curve C is
then the level set
C = {(x, y) ∈ R2 : x2 + y 2 = 1},
which is again the circle line in x − y plane.
For a parametric curve ϕ : R → Rn any point ϕ(t) gives the ’position’ at
’time’ t. The derivative with respect to the parameter t gives the velocity
vector ϕ0 (t) at time “time” t. Both are vectors in Rn with the following
components

ϕ(t) = (ϕ1 (t), . . . , ϕn (t)) ∈ Rn and ϕ0 (t) = (ϕ01 (t), . . . , ϕ0n (t)) ∈ Rn .

If ϕ0 (t) 6= 0, then ϕ0 (t) is a tangent vector of the curve. The tangent line
Tϕ(t) at a point ϕ(t) is given by

Tϕ(t) : R → Rn , λ 7→ Tϕ(t) (λ) = ϕ(t) + λϕ0 (t).

This is a straight line through the point ϕ(t) in direction of ϕ0 (t).

11
Figure 11: helix

12
Figure 12: helicoid

13
Definition 2.5 A vector x ∈ Rn is orthogonal to a parametric curve
ϕ : R → Rn at the point ϕ(t) if hx, ϕ0 (t)i = 0, i.e. if it is orthogonal to
the tangent line.

Lemma 2.6 Let f : Rn → R be differentiable and a ∈ Rn . Then

∇f (a) ⊥ {x ∈ Rn : f (x) = f (a)} =: L(f (a)).

Proof. Let ϕ : R → L(f (a)) be a differentiable parametric curve in the

surface L(f (a)) with ϕ(0) = a. We apply the chain rule (see Proposi-
tion 2.8 below) and the notion of differentiability in Definition 2.7. Note
that Df (x0 ) ∈ Lin(Rn , R), x0 ∈ Rn , is a linear mapping given by the (1 × n)-
matrix  
∂f ∂f ∂f
∂x1
(x 0 ) ∂x2
(x 0 ) . . . ∂xn
(x 0 ) ,

and that Dϕ(t) ∈ Lin(R, Rn ) is a linear mapping given by the (n × 1)-matrix

 0 
ϕ1 (t)
 ϕ02 (t) 
 
 · .
 
 · 
ϕ0n (t)

The chain rule gives for the derivative of the composition f ◦ ϕ with respect
to t at t = 0 as

D(f ◦ ϕ)(t = 0) = Df (ϕ(0)) ◦ Dϕ(0),

where the ◦-operation on the right hand side is the matrix product which in
this case is the corresponding scalar product in Rn . With that we get
d
0= f (ϕ(t)) t=0 = h∇f (ϕ(0)), ϕ0 (0)i
dt
= h∇f (a), ϕ0 (0)i.

This implies ∇f (a) ⊥ ϕ0 (0) and ϕ0 (0) is tangent vector, i.e. it is in L(f (a)).
2
Recall the notion of (total) differentiability in the following definition. It
generalises the notion of differentiability for real-valued functions defined on
the real line. Roughly speaking, the existence of the differential quotient is
equivalent to a linear approximation (tangent line) to that function.

14
Definition 2.7 (Differentiability) Let D ⊂ Rn be a domain and f : D →
Rm , m ≥ 1, f = (f1 , . . . , fm ). We say that f is differentiable at x0 ∈ D if
the partial derivatives of f exist at x0 and if there exists a linear mapping
L : Rn → Rm with
kf (x) − f (x0 ) − L(x − x0 )k
lim = 0,
x→x0 kx − x0 k

where the linear mapping L is given by the so-called (m×n)− Jacobi matrix
at the point x0 , i.e.
 ∂f ∂f1 ∂f1

1
∂x1
(x0 ) ∂x 2
(x 0 ) . . . ∂x n
(x 0 )
 ∂f2
 (x0 ) ∂f2 (x0 ) . . .

∂x ∂x
. . . 
L = Df (x0 ) = 
 ...
1 2 .
 ... ... ...  
∂fm
... ... . . . ∂xn (x0 )

The matrix Df (x0 ) is said to be the (total) derivative of f at x0 . We say

that f is differemtiable if it is differentiable at every point of its domain D.
In that case the derivative Df of f is mapping

Df : D → Lin(Rn , Rm ),

where Lin(Rn , Rm ) is the space of linear mappings from Rn to Rm isomorphic

to the space of real (n × m)− matrices.

Proposition 2.8 (Chain rule) Pick n, p, q ∈ N. Let U ⊂ Rn and V ⊂ Rp

be open subset sets and consider mappings f : U → Rp and g : V → Rq . Let
x0 ∈ U be such that f (x0 ) ∈ V . Suppose f is differentiable at x0 and g at
f (x0 ). Then g ◦ f : U → Rq , the composition of g and f , is differentiable at
x0 , and we have

D(g ◦ f )(x0 ) = Dg(f (x0 )) ◦ Df (x0 ) ∈ Lin(Rn , Rq ),

where the ◦-operation on the right hand side means composition of the linear
mappings which corresponds to the matrix product of the matrices describing
the linear mappings.

Notation 2.9 Let a curve C ⊂ Rn be given as a level set of some function or

via some equation. A parametric curve γ : [a, b] → Rn is called a parametri-
sation (resp C 1 parametrisation if γ is C 1 , that is, γ is differentiable and
both, γ and γ 0 are continuous) of the curve C if γ([a, b]) = C. A given curve
can have several parametrisations.

15
3 Line integrals
In this section we want to take integrals along a curve C ⊂ Rn . A curve is a
one-dimensional object in Rn .

3.1 Integrating scalar fields

If we integrate the constant scalar field u : Rn → R, x 7→ u(x) = 1 along a
curve with given parametrisation γ we shall get the length of the γ. And if
we integrate some density along that curve we get the total mass of a thin
curve.

Figure 13:

Definition 3.1 (Scalar line integral) Let γ : [a, b] → Rn be a C 1 parametri-

sation of the curve C ⊂ Rn . The scalar line integral of a scalar field
u : Rn → R along the curve C is given by
Z Z b
u := u(γ(t))kγ 0 (t)k dt. (3.3)
C a

R R R
Alternative notations are: γ u, C u ds and γ u ds and the integral is
sometimes also called the path integral along the path C.

16
Note that the scalar field u needs only to be defined along the curve, i.e.
u : C → R.

Remark 3.2 The length of a parametric curve γ : [a, b] → Rn is given by

Z Z b
1= kγ 0 (t)k dt.
γ a

As an example consider n = 2 and calculate the length of the circle line

 
2 cos t
γ : [0, 2π] → R , t 7→ γ(t) = .
sin t

As  
0 − sin t p
γ (t) = and kγ 0 (t)k = (− sin t)2 + cos2 t = 1,
cos t
R 2π
kγ 0 (t)k dt = 2π.
R
we have γ
1= 0

3.2 Integrating vector fields

As an introductory example consider a particle moving along a curve (path)
C ⊂ R3 . The particle is acted on by a force F~ (~x), ~x ∈ R3 , which is a vector
field F~ : R3 → R3 , ~x 7→ F~ (~x).
What is the total amount of work done as the particles moves along the
curve C? We consider a small displacement d~x at position ~x within the curve
C. Then the work that is done when the particle moves from position ~x to
position ~x + d~x along the curve C is precisely −F~ ·d~x = −hF~ , d~xi.

17
 Figure 14:

Hence, heuristically the total amount of work shall be an integral (the sum
of all these small contributions)
Z
− F~ ·d~x.
C

We make this notion mathematically precise in the following definition.

Definition 3.3 (Tangent line integral) Let C ⊂ Rn be a curve with a C 1

parametrisation γ : [a, b] → Rn . The tangent line integral of a vector field
f : Rn → Rn along C is defined as
Z Z b
f= hf (γ(t), γ 0 (t)i dt. (3.4)
C a

R R R
Alternative notations are C
f · d~s, γ
f ·d~s or γ
f ·T̂ ds.

Remark 3.4 The tangent line integral of a vector field f : Rn → Rn along

the curve C with parametrisation γ : [a, b] → Rn can be written as
Z bD
γ 0 (t) E 0
Z
f= f (γ(t), kγ (t)k dt,
C a kγ 0 (t)k

18
 0
where f (γ(t), kγγ 0 (t)
(t)k
is the projection of f onto the tangent line, see figure
R
15 below. Hence, the tangent line intergal γ f of a vector field f is the scalar
line integral of the component of f along the tangent direction.

Figure 15:

We close with two examples.

Example 3.5 (Tangent line integral) We calculate the work done when
moving a mass m> 0 along a line/curve with parametrisation γ : [0, π] →
t
R2 , t 7→ in the gravity field f : R2 → R2 , (x, y) 7→ f (x, y) =
  − cos t
0
. Here g is a constant, called the earth acceleration. Then the
−mg    
0 π
work to be done moving the mass from to is given by
−1 1
Z Z π D   E
0 1
− f =− , dt
γ 0 −mg sin t
Z π
= mg sin t dt = 2mg.
0

19
 Figure 16:

Example 3.6 (Scalar line integral) Consider the curve C with parametri-
sation γ : [−π, π] → R3 , t 7→ γ(t) = (cos t, sin t, t). The scalar line integral of
u : R3 → R, (x, y, z) 7→ x2 + y 2 along the curve C is
Z Z Z π
2
u = (x + y ) = 2
(cos2 (t) + sin2 (t))kγ 0 (t)k dt,
C C −π
√ √
and as γ 0 (t) = (− sin t, cos t, 1) with kγ 0 (t)k = sin2 t + cos2 t + 1 = 2 we
get Z √
(x2 + y 2 ) = 2π 2.
C


4 Gradient Vector Fields

In this section we will study gradient vector fields. We will prove the Fun-
damental Theorem of vector calculus (FTC) for vector fields. Moreover,
integrals (tangent line integrals) of gradient vector fields along curves can
easily be calculated once the potential (primitive) is known.

20
4.1 FTC for gradient vector fields
Definition 4.1 A vector field f : D → Rn , D ⊂ Rn , is called a gradient
vector field , if there exists a differentiable scalar field Φ : D → R with
f (x) = ∇Φ(x) for all x ∈ D. (4.5)
Φ is called the potential of the vector field f if (4.5) is satisfied. Sometimes
a gradient vector field is also called a vector field of gradient type.

Remark 4.2 (a) If Φ is a potential for the vector field f , then Φ + c is a

potential for f for any c ∈ R as well because ∇(Φ + c)(x) = ∇Φ(x), x ∈
D. That is, the potential is not unique.
(b) Not every vector field is of gradient type, e.g.
f : R3 → R3 , (x, y, z) 7→ f (x, y, z) = (z, z, y).
If Φ : R3 → R would be a potential it would satisfy
∂x Φ(x, y, z) = z ⇒ Φ(x, y, z) = zx + c1 (y, z)
∂z Φ(x, y, z) = y ⇒ Φ(x, y, z) = yz + c2 (x, y),
which has no solution because of xz versus yz.

The most important result about gradient vector fields is the following
theorem.

Theorem 4.3 (FTC for gradient vector fields) Let Φ : Rn → R be a

scalar field (C 1 - differentiable) and let C ⊂ Rn be a curve with a (piece-
wise C 1 - differentiable) parametrisation γ : [a, b] → Rn . Let f : Rn → Rn be
a vector field with ∇Φ = f . Then the tangent line integral of the vector field
is given by
Z
f = Φ(γ(b)) − Φ(γ(a)). (4.6)
C

This is the Fundamental Theorem of calculus for gradient vector fields.

Proof. The map [a, b] → R, t 7→ Φ(γ(t)) has derivative

n
X ∂Φ dγi (t)
(Φ(γ(t))0 = (γ(t))
i=1
∂xi dt
= h∇Φ(γ(t)), γ 0 (t)i = hf (γ(t)), γ 0 (t)i.

21
Thus we get
Z Z b Z b
0
f= hf (γ(t)), γ (t)idt = (Φ(γ(t))0 dt
C a a
= Φ(γ(b)) − Φ(γ(a)),

where we used the continuity of Φ0 and the Fundamental Theorem of Calculus

for functions on the real line.
2

Example 4.4 (Revisit of Example 3.5)  Moving a mass m > 0 in the

0
gravity field f : R2 → R2 , (x, y) 7→ f (x, y) = with potential v : R2 →
−mg
R, (x, y) 7→ Φ(x, y) = −mgy. Then the tangent line integral along any curve
C ⊂ R2 with parametrisation γ : [a, b] → R2 gives the work moving the mass
along C, that is
Z Z
− f = − f = −Φ(γ(b)) + Φ(γ(a)).
C γ

In Example
 3.5  we have a curve C with parametrisation γ : [0, π] → R2 , t 7→
t
γ(t) = , hence the work (tangent line intergal) can be computed
− cos t
directly with the potential
Z
− f = mg(cos π − (−1)) = 2mg.
C

Notation 4.5 (a) A simple curve C ⊂ Rn is a curve which is the image of a

piecewise C 1 -parametrisation γ : [a, b] → R that is one-to-one on the interval
[a, b]. The points γ(a) and γ(b) are the endpoints of the curve.

22
 Figure 17: Simple and not simple curve

(b) Let γ : [a, b] → Rn be a (piecewise C 1 - differentiable) parametrisation.

Then the map
ω : [a, b] → Rn , t 7→ ω(t) = γ(a + b − t)
is the parametrisation of the reversed (inverse direction) path called C −1 .

23
 Figure 18:

Remark 4.6 If f is a gradient vector field with potential Φ and γ is a line

(path or parametrisation), then the tangent line integral depends only on the
endpoints of the line (path or parametrisation) and on the direction of γ.

Figure 19:

Consider v, w ∈ Rn and a parametrised path γ1 : [a1 , b1 ] → Rn from v to w

24
and a parametrised path γ2 : [a2 , b2 ] → Rn from w to v. Taking the direction
into account one gets
Z
f = Φ(γ1 (b1 )) − Φ(γ1 (a1 )) = Φ(w) − Φ(v) = −(Φ(γ(a2 )) − Φ(γ2 (b2 ))
γ1
Z
=− f
γ2

because of γ1 (a1 ) = v = γ2 (b2 ) and γ1 (b1 ) = w = γ2 (a2 ), see figure 19.

Moreover it follows with the same arguments that the tangent line integral
along a closed curve (path or line) is zero, see figure below.

Figure 20: Integral of a closed curve

The gradient vector fields are important in particular in physics.

Physics notation: A C 1 vector field f : D → Rn , D ⊂ Rn , which is defined
on D except possibly for a finite number of points, is said to be conservative
if it has the property that the (tangent) line integral of f along any closed
simple curve C ⊂ Rn is zero:
Z
f = 0.
C closed

Equivalent: The vector field f is conservative if the (tangent) line integral

of f along a curve only depends on the endpoints of the curve C, not on the

25
particular path taken by the curve. That is
Z Z
f= f
C1 C2

for any two curves connecting two points a ∈ Rn and b ∈ Rn in the same
direction, i.e. both curves travel form a to b.
Then one can show that a vector field f is of gradient type if it is conservative.
This provides us with another method to check if a given vector field f
is of gradient type or not. We are left to check if an integral (tangent line)
along a closed curve is zero. However, here one has to ensure that the closed
curve lies in the domain of definition of the vector field.

Example 4.7 Let f : R2 → R2 , (x, y) 7→ f (x, y) = (−y, x) be vector field

and integrate this along the circle line with parametrisation γ : [0, 2π] →
R2 , t 7→ (cos t, sin t). The tangent line along this closed curve is
Z 2π D    Z 2π
− sin t − sin t E
Z
f= , dt = 1 dt = 2π 6= 0.
γ 0 cos t cos t 0

Hence the vector field f cannot be of gradient type. 

4.2 Finding a potential

Corollary 4.8 If f : Rn → Rn is a gradient vector field and Φ : Rn → R a
potential, then for any x ∈ Rn \ {0}
Z
Φ(x) = Φ(0) + f,
γx

where Φ(0) is a constant and where γx is the straight line from the origin 0
to x, that is
γx : [0, 1] → Rn , t 7→ γx (t) = tx.

26
 Figure 21:

We can choose any other reference point x0 ∈ Rn instead the origin. In that
case one needs a path connecting x0 and x. It is important that the point and
the path connecting any point with the reference point are lying in the domain
of definition of Φ and f . If Φ1 and Φ2 are two potentials for the vector field
f , then

Φ1 (x) − Φ2 (x) = Φ1 (0) − Φ2 (0) = constant x ∈ Rn .

Definition 4.9 If f : Rn → Rn is a vector field, a flow line for f is a para-

metric curve (path) γ : [a, b] → Rn such that

γ 0 (t) = f (γ(t)). (4.7)

If one considers the flow of a liquid in a pipe, then the vector field f
yields the velocity vector field of the parametric curve (path) γ. The velocity
vector of the fluid is tangent to a flow line, see figure 22 below.

27
 Figure 22:

If one is given a vector field it is easy to draw the flow line. It is the line
threading its way through the vector field in the plane as shown in figure 23.

Figure 23:

28
Example 4.10 (Finding a potential) Let f : R2 → R2 , (x, y) 7→ f (x, y) =
1
2
(−x, y). We sketch the vector field in figure 24 below.

Figure 24:

This is done by considering the mappings

 
1 x
f (x, 0) = −
2 0
 
1 0
f (0, y) =
2 y
 
1 −x
f (x, x) = .
2 x
Assume that f = ∇Φ for some Φ : R2 → R. Then

1 x2
∂x Φ(x, y) = − x ⇔ Φ(x, y) = − + c1 (y)
2 4
1 y2
∂y Φ(x, y) = y ⇔ Φ(x, y) = + c2 (x),
2 4
y2 x2
i.e. Φ(x, y) = 4
− 4
is a solution (C1 = C2 = 0). 

Example 4.11 Let f : R2 → R2 , (x, y) 7→ f (x, y) = (2y, x + y). Assume

that f is a gradient vector field, that is ∇Φ = f for some Φ : R2 → R. Then

29
we can conclude
∂x Φ(x, y) = 2y ⇔ Φ(x, y) = 2xy + C1 (y)
1
∂y Φ(x, y) = x + y ⇔ Φ(x, y) = xy + y 2 + C2 (x).
2
But it is impossible to find any such functions C1 and C2 because one can
never match 2xy versus xy. Thus f is not a gradient vector field. 

4.3 Radial vector fields

The next example is very important. It contains vector fields which are radial
symmetric.

Example 4.12 (Radial vector fields) For the scalar valued function g : (0, ∞) →
R define the vector field f : Rn → Rn by
x
, for x ∈ Rn \ {0}
(
g(kxk) kxk
f (x) = .
0 , for x = 0

Assume that we can always find a primitive G : (0, ∞) → R with G0 (r) = g(r)
for all r > 0.
We say that f : Rn → Rn is a radial vector field if kf (x)k is constant
within concentric spheres, i.e.

kf (x)k = const for all x ∈ Rn with kxk = c for any c > 0.

Define
q 
n
Φ : R \ {0} → R, x 7→ G(kxk) = G x21 + · · · + x2n .

Then for x ∈ Rn \ {0} and any i ∈ {1, . . . , n}, we compute

2xi xi
∂i Φ(x) = G0 (kxk) p 2 = g(kxk)
2 x1 + · · · + x2n kxk
= fi (x).

Hence, ∇Φ(x) = f (x) for all x 6= 0.

Result: On Rn \ {0} radial vector fields are always gradient vector fields.

We outline two important examples of radial vector fields in physics.

30
(a) Gravitational force field: Put the origin at the centre of the earth or
some other planet having acceleration gplanet and mass M > 0. The gravita-
tional force on some testing body with mass m > 0 is the vector field
gplanet mM
F : R3 \ {0} → R3 , x 7→ F (x) = − x,
kxk3

which has the potential

gplanet mM
Φ : R3 \ {0} → R, x 7→ Φ(x) = .
kxk

The minus sign in the force field ensures that the force is directed to the
centre of the planet.

(b) Coulomb’s law: The force acting on a charge q at position x is due to

a charge Q at the origin with electrostatic constant ε
εqQ
F : R3 \ {0} → R3 , x 7→ F (x) = x,
kxk3

which has the potential

εqQ
Φ : R3 \ {0} → R, x 7→ Φ(x) = .
kxk


Since the potential Φ is constant on the level surface of Φ they are called
equipotential surfaces. Note that the force field is orthogonal to the
equipotential surfaces, compare the example of radial vector fields above.
There, the force field is radial and the equipotential surfaces are concentric
spheres.
We finish this section coming back to the notion of a flow line of a vector
field. We connect it to first order differential equations. Geometrically, a flow
line for a given vector field f is a parametric curve (path) that threads its way
through the domain of the vector field in such a way that the tangent vector
of the parametric curve (path) coincides with the vector field (see figure 23
above). A flow line may be viewed as a solution of a system of differential
equations. Let Γ(x, t) for t ≥ 0 and x ∈ Rn be the position at time t of the
point on the flow line through x (at time t = 0) after time t has elapsed.

31
 Figure 25:

The mapping Γ : Rn × [0, ∞) → Rn with

∂Γ
(x, t) = f (Γ(x, t))
∂t
Γ(x, 0) = x
is also called the flow line of the vector field f . If we denote the differentiation
of Γ with respect to the spatial variable x by Dx (t is fixed) we can interchange
the partial derivative with respect to time t with Dx (under our general
assumptions that all maps are continuous differentiable etc). Then one can
derive the following equation of first variation

∂t Dx Γ(x, t) = Df (Γ(x, t))Dx Γ(x, t).

Here Df (x), x ∈ Rn , is the n × n-matrix with the first partial derivatives of

the vector field at point x, i.e.
 
∂1 f1 (x) · · · ∂n f1 (x)
 ∂1 f2 (x) · · · ∂n f2 (x) 
 
 · · · · · 
Df (x) =  .

 · ··· · 

 · ··· · 
∂1 fn (x) · · · ∂n fn (x)

32
Similarly, Dx Γ(x, t) is a n × n-matrix.

5 Surface Integrals
We want to extend the notion of line integrals to surface integrals in R3 . One
can generalises this to arbitrary k-dimensional surfaces in Rn , 1 < k ≤ n, but
we deal first with the case k = 2 and n = 3.

5.1 Surfaces
We need methods for describing surfaces in R3 . A surface S ⊂ R3 as a subset
of points in R3 can be described by two methods:
1.) Level sets of real-valued functions f : R3 → R. That is,

S = f −1 (c) = {x ∈ R3 : f (x) = c}

for some c ∈ R.
2.) Parametrisation
 
α1 (s, t)
α : Q → R3 , (s, t) 7→ α(s, t) = α2 (s, t)
α3 (s, t)

with some parameter domain Q ⊂ R2 .

Notation 5.1 Let S ⊂ R3 be a surface. A (C 1 -) parametrisation of the

surface S is a map (respectively a C 1 -map) α : Q → R3 with Q ⊂ R2 and
α(Q) = S.

Example 5.2 (a) Cylinder (volume). Let R > 0, H > 0. The parametri-
sation
 
r cos ϕ
α : [0, R] × [0, 2π] × [0, H] → R3 , (r, ϕ, h) 7→ α(r, ϕ, h) =  r sin ϕ 
h

describes a cylinder (see figure 26 below). Note that when the angle variable
varies only in a subset of [0, 2π] one gets not the whole cylinder but one
where a ’piece of cake’ is removed. If one fixes the radius r ∈ (0, R] one gets
the parametrisation of the cylindrical surface of radius r and height H.

33
 Figure 26:

(b) Graph of a function Let h : Q → R1 be a continuous function, Q ⊂ R2 .

The map  
s
α : Q → R2+1 , (s, t) 7→ α(s, t) =  t 
h(s, t)
is a parametrisation of a surface S, the graph of the function h, i.e. S =
α(Q) = graph h (see figure 27).

34
 Figure 27:

(c) Polar coordinates Define Q1 = [−π, π] and

 
1 2 1 cos ϕ
α : Q1 → R , ϕ 7→ α (ϕ) = .
sin ϕ
Then define recursively the following parametric k-dimensional surfaces in
Rn , 1 < k ≤ n. Qk = Qk−1 × [−π/2, π/2] and
 k−1 
k k+1 α (ϕ1 , . . . , ϕk−1 ) cos ϕk
α : Qk → R , (ϕ1 , . . . , ϕk ) 7→ .
sin ϕk

The set S1 = α1 (Q1 ) is the circle line in the plane and the set Sk = αk (Qk )
is a k-dimensional surface. One can prove that
Sk = {x ∈ Rk+1 : kxk = 1}. (5.8)
Sk is called the k-dimensional sphere in Rk+1 . For k = 2 and n = 3 we get
the polar coordinates (earth coordinates)
 
cos ϕ1 cos ϕ2
α2 (ϕ1 , ϕ2 ) =  sin ϕ1 cos ϕ2  ,
sin ϕ2
where the angle ϕ1 ∈ [0, 2π] is the geographical longitude and respectively the
angle ϕ2 ∈ [−π/2, π/2] is the geographical latitude. Note that sometimes one

35
takes the latitude angle Θ = π2 − ϕ2 ∈ [0, 2π] where one takes the angle away
from the north pole. In the latter case one take the angle from the equator
counting negative towards the north pole and positive towards south pole.
Recall that sin ϕ2 = cos Θ respectively cos ϕ2 = sin Θ for ϕ2 ∈ [−π/2, π/2]
and Θ ∈ [0, π].

Figure 28: Polar coordinates in R3

5.2 Integral
For an integration over a surface in R3 we need surface unit normal vectors N b.
The unit normal vectors have norm 1 and they are perpendicular to the plane
at each single point of the surface (more precisely they are perpendicular to
the tangent plane at that point).
If we describe our surface as a level set of a differentiable real-valued function
f : R3 → R we know from Lemma 2.6 that the gradient is orthogonal to the
level surfaces, i.e.
∇f ⊥ S ⇒ N b := ∇f ∈ R3 .
k∇f k
What is the tangent plane at a point to a given surface S ⊂ R3 ? When our
surface S ⊂ R3 is given with a parametrisation α : Q → R3 we see from figure
29 that at each point α(s, t) of the surface S the vectors ∂α
∂s
(s, t) and ∂α
∂t
(s, t)

36
are both tangent to the surface (that is, they are lying in the tangent plane).
Hence, the cross-product
∂α ∂α
(s, t) × (s, t)
∂s ∂t
is a normal to the surface at the point α(s, t).

Notation 5.3 Assume that α : Q → R3 is a C 1 -parametrisation of the sur-

face S ⊂ R3 . If we fix one of the two arguments of the mapping α we get
families of paths/curves lying in the surface S. Fixing t the partial derivative
∂α(s,t)
∂s
=: Ts (α(s, t)) is called the tangent vector in s-direction and fixing
s the partial derivative ∂α(s,t)
∂t
=: Tt (α(s, t)) is called the tangent vector in
t-direction. The tangent vectors at a given point span the tangent plane at
this point of the given surface.
We say that a parametrisation (respectively surface) is regular or smooth
at the point α(s, t) if Ts × Tt 6= 0 at the point α(s, t). Intuitively, a smooth
surface has no ”corners”.

Figure 29: Tangent vectors of the surface

We normalise this vector to get the unit normal for the surface S

b (α(s, t)) = ∂α (s, t) × ∂α (s, t)

N (s, t) ∈ Q. (5.9)
∂s ∂t
37
We will often write N b without the argument. Note that N b : S → R3 is a
vector field whose domain is the surface S, but one can conceive it also as
the map N b : Q → R3 for a given parametrisation.
Reminder: Cross-product x, y ∈ R3
     
x1 y1 x2 y3 − x3 y2
x2  × y2  = x3 y1 − x1 y3 
x3 y3 x1 y2 − x2 y1

Definition 5.4 Let α : Q → R3 , Q ⊂ R2 , be a parametrisation of a surface

S ⊂ R3 . Then the scalar surface integral of a continuous scalar field
u : R3 → R over the surface S is defined as
Z ZZ
∂α ∂α
u= u(α(s, t)) (s, t) × (s, t) dsdt. (5.10)
S Q ∂s ∂t

Remark 5.5 (a) If we integrate 1 over a surface we get the surface area of
S, Z ZZ
∂α ∂α
1= (s, t) × (s, t) dsdt = area (S).
S Q ∂s ∂t
(b) If we take a small rectangle q ⊂ Q in the parameter domain Q whose left
bottom corner is (s, t) ∈ Q and whose side length are ∆s and ∆t respectively,
we can map this rectangle with the parametrisation to R3 . The area of the
image rectangle α(q) can be computed as

∂α ∂α
area (α(q)) = (s, t) × (s, t) ∆s∆t.
∂s ∂t

38
Figure 30: Image of the small rectangle q

If S is a surface in R3 do we have several parametrisations? The answer

is yes, and we will formulate this result for any k-dimensional surface (1 <
k ≤ n) in Rn .
Definition 5.6 Let 1 < k ≤ n and Q1 and Q2 two bounded and closed sets
in Rk (with a smooth boundary).
(a) A bijective map T : Q1 → Q2 is called a C 1 -parameter transfor-
mation if (i) T and T −1 are C 1 -mappings, (ii) det DT t > 0 for all
t ∈ Q1 ⊂ Rk .
(b) Let α1 : Q1 → Rn and α2 : Q2 → Rn be two parametrisations (C 1 -
mappings). α1 and α2 are said to be equivalent if there exists a C 1
parameter transformation T : Q1 → Q2 such that α1 = α2 ◦ T .
The matrix DT t
is the total derivative of the map T : Q1 → Q2 , i.e.
 
∂T1 ∂T2
∂t1 ∂t2
· · · ∂T
∂tk
1

 ∂T2 ∂T2 · · · ∂T2 

 ∂t1 ∂t2 ∂tk 
 · · ··· · 
 
DT t =  .
 · · ··· · 
 · · ··· · 
 
∂Tk ∂Tk
∂t1 ∂t2
· · · ∂T
∂tk
k

39
5.3 Kissing problem
Example 5.7 (a) Spherical cap S: Fix θ ∈ (0, π/2) and define
 
cos ϕ1 cos ϕ2
α : [−π, π] × [θ, π/2] → R3 , (ϕ1 , ϕ2 ) 7→ α(ϕ1 , ϕ2 ) =  sin ϕ1 cos ϕ2 
sin ϕ2

Figure 31: Spherical cap

We want to compute the surface area of the spherical cap defined by the
parametrisation α. We compute the surface normal vectors at (ϕ1 , ϕ2 ):
   
− sin ϕ1 cos ϕ2 − cos ϕ1 sin ϕ2
∂α ∂α
=  cos ϕ1 cos ϕ2  =  − sin ϕ1 sin ϕ2 
∂ϕ1 ∂ϕ2
0 cos ϕ2

and their cross-product

 
cos ϕ1 cos2 ϕ2
∂α ∂α
× = − sin ϕ1 cos2 ϕ2 
∂ϕ1 ∂ϕ2
− sin ϕ2 cos ϕ2

with norm
∂α ∂α p
× = cos2 ϕ2 = cos ϕ2 ,
∂ϕ1 ∂ϕ2

40
where the last equality follows due to θ ∈ (−π/2, π/2). The surface area is
then given as the surface integral of 1,
Z Z π/2 Z π ϕ2 =π/2
area (S) = 1= 1 cos ϕ2 dϕ1 dϕ2 = 2π sin ϕ2
S θ −π ϕ2 =θ

= 2π(1 − sin θ).

Special case: total surface area (θ = −π/2) is 4π.

(c) Newton’s kissing problem How many unit balls can simultaneously
touch a given ball in R3 ? We calculate the shadowed surface area seen in
figure 32.

Figure 32: Kissing problem

We see that 2 sin(π/2 − θ) = 1 which implies π/2 − θ = π/6 givingp θ = π/3.

Each ball shadows the surface area 2π(1 − sin θ) = 2π(1 − 3/4) of the
central ball in figure 32. As the full surface area is 4π we can have at most
4π 4
p = √ ≈ 14.928
2π(1 − 3/4) 2− 3

balls. Hence, the number in question is less than 14. Is the maximum number
now 12, 13 or 14 ?
What about the other dimensions?

41
 Kissing in 1d

In one dimension, the kissing number is obviously 2.

It is easy to see (and to prove) that in two dimensions the kissing number
is 6.

In three dimensions the answer is not so clear. It is easy to arrange 12

spheres so that each touches a central sphere, but there is a lot of space left
over, and it is not obvious that there is no way to pack in a 13th sphere.
(In fact, there is so much extra space that any two of the 12 outer spheres
can exchange places without any of the outer spheres losing contact with
the centre one.) This was the subject of a famous disagreement between
mathematicians Isaac Newton and David Gregory. Newton thought that the
limit was 12, and Gregory that a 13th could fit. The question was not resolved
until 1874; Newton was correct. In four dimensions, it was known for some
time that the answer is either 24 or 25. It is easy to produce a packing of 24
spheres around a central sphere (one can place the spheres at the vertices of
a suitably scaled 24-cell centred at the origin). As in the three-dimensional
case, there is a lot of space left over - even more, in fact, than for n = 3
- so the situation was even less clear. Finally, in 2003, Oleg Musin proved
the kissing number for n = 4 to be 24, using a subtle trick. The kissing
number in n dimensions is unknown for n > 4, except for n = 8(240), and
n = 24(196, 560). The results in these dimensions stem from the existence
of highly symmetrical lattices: the E8 lattice and the Leech lattice. In fact,
the only way to arrange spheres in these dimensions with the above kissing
numbers is to centre them at the minimal vectors in these lattices. There is
no space whatsoever for any additional balls. Rough volume estimates show
that the kissing number in n dimensions grows exponentially. The base of
exponential growth is not known.
The following table lists some known kissing number in various dimen-
sions.

42
Kissing in 2d

43
Kissing in 3d

44
 dimension kissing ] year
1 2
2 6
3 12 (1874)
4 24 (2003)
8 240 (1979)
24 196560 (1979)

6 Divergence of Vector Fields

6.1 Flux across a surface
We consider the flow through a pipe. Let some fluid flowing with velocity ~v
through a pipe. What is the total volume of fluid passing through the pipe
per unit time? This volume flow rate is often called the flux of the fluid
through the pipe or the flux across the surface S that forms the end of the
pipe.

Figure 33: Flow through a pipe

If the velocity ~v is parallel to the walls and if ~v is constant, i.e. k~v k = v0 ,

the flow rate (flux) is given as v0 A, where A is the area of the surface S.

45
 Figure 34: Velocity not parallel

If the velocity ~v is not perpendicular to the small surface dS in figure 34 (de-

scribed by the unit normal vector N b ), only the component of ~v perpendicular
to dS contributes to the flux across the small surface dS.

Definition 6.1 Let S ⊂ R3 be a surface with unit normal vector field N b: S →

3 1 3 2
R defined with a C -parametrisation α : Q → R , Q ⊂ R . Then the flux
of the vector field f : R3 → R3 across the surface S in direction N
b is defined
as the scalar surface integral of hf, N
b i, i.e.
Z ZZ
∂α ∂α
hf, N
bi = f (α(s, t)), N
b (α(s, t) (s, t) × (s, t) dsdt (6.11)
S Q ∂s ∂t

f~·d~s.
R R
Alternative notations: S
f ·N
b ds or
S

Example 6.2 (Flux out of a box) Ω = [ax , bx ] × [ay , by ] × [az , bz ] with

ax , ay , az , bx , by , bz ∈ R and let f : R3 → R3 be a continuous differentiable
vector field with component functions f = (f1 , f2 , f3 ). We calculate the flux
through any side/surface of the box in figure 35.

46
 Figure 35: box Ω

Calculation of the flux through the six surfaces of the box Ω

1.) Flux through the top of the box. A parametrisation is given by

 
x
αtop (x, y) = y 
 for x ∈ [ax , bx ], y ∈ [ay , by ].
bz

The unit normal vector N btop can already be seen in figure 35 but we show
the exact calculation.
     
1 0 0
∂αtop ∂αtop
× =  0 × 1
   =  0 = 1.
∂x ∂y
0 0 1
 
0
Thus, Ntop (x, y, bz ) = 0 with (x, y) ∈ [ax , bx ] × [ay , by ]. Hence, the flux
b 
1
through the top is
Z Z bx Z by
hf, Nbi = f3 (x, y, bz ) dydx.
Stop ax ay

2.) Flux through the bottom. This is the same calculation but with the
opposite direction, hence a minus sign appears for the unit surface normal

47
  
0
Nbbottom (x, y, az ) =  0  with (x, y) ∈ [ax , bx ] × [ay , by ] and the component
−1
function has to be evaluated at az for the third entry.
Z Z bx Z by
hf, N
bi = − f3 (x, y, az ) dydx.
Sbottom ax ay

We sum up both contributions and use the Fundamental Theorem of Calculus

(recall that the vector field f is continuously differentiable) to get
Z Z bx Z by


hf, N i =
b f3 (x, y, bz ) − f3 (x, y, az ) dydx
Stop ∪Sbottom ax ay
Z bx Z by Z bz
∂f3
= (x, y, z) dzdydx.
ax ay az ∂z

3.) and 4.) Similarly we derive the contribution from the left and the right
surface, Sl and Sr , as
Z Z bx Z by Z bz
∂f1
hf, N i =
b (x, y, z) dzdydx.
Sl ∪Sr ax ay az ∂x

5.) and 6.) Finally we get the contribution from the back and the front
Z Z bx Z by Z bz
∂f2
hf, N i =
b (x, y, z) dzdydx.
Sback ∪Sfront ax ay az ∂y

Taking the sum over all contributions of the six surfaces gives the total flux
across the surface

∂Ω := Stop ∪ Sbottom ∪ Sl ∪ Sr ∪ Sback ∪ Sfront

as
Z Z bx Z by Z bz  ∂f ∂f2 ∂f3 
1
hf, N
bi = + + dzdydx. (6.12)
∂Ω ax ay az ∂x ∂y ∂z

6.2 Divergence
In the following we will often write ∂Ω for the surface of some domain Ω ⊂ R3 .
The integrand in (6.12) motivates the following definition.

48
Definition 6.3 Let f : Rn → Rn be a differentiable vector field. The diver-
gence of the vector field f = (f1 , . . . , fn ) is the scalar field

div f : Rn → R
n
X ∂fi
x 7→ div (f )(x) = (x).
i=1
∂xi

~ · f~ or ∇ · f .
Alternative notations are: ∇ · f or ∇

Note the following relations of scalar and vector fields with the corresponding
operations.

scalar field operation vector field

grad
Φ : Rn → R −→ grad Φ : Rn → Rn

div
div f : Rn → R ←− f : Rn → Rn

Definition 6.4 Given a two times differentiable scalar field Φ : Rn → R, the

Laplacian of the scalar field Φ is the scalar field

∆Φ : Rn → R
n
X ∂Φ2
x 7→ ∆Φ(x) = div grad Φ(x) = (x).
i=1
∂x2i

Alternative notation is ∇2 Φ. Note that the Laplacian of scalar field is the

divergence of the gradient field of the scalar field.

Note: n n
X ∂ ∂Φ X ∂ 2 Φ
∆Φ = div (grad Φ) = = 2
.
i=1
∂x i ∂x i i=1
∂x i

We have seen in Example 6.2 that

Z Z
hf, N i =
b div f (x) dx
∂Ω Ω

(for boxes only so far). If Ω is a small box centred at a ∈ R3 , then

Z
div f (x) dx ≈ div f (a)vol(Ω),
Ω

49
where vol(Ω) is the volume of the small box Ω, implies that
R
∂Ω
hf, N
bi
div f (a) ≈ . (6.13)
vol(Ω)

The divergence of a vector field gives the outgoing flux per volume.

Consider the following example.

Example 6.5 (a) f : R3 → R3 , (x, y, z) 7→ f (x, y, z) = (x, 0, 0).

Figure 36: vector field expanding

As seen div f (a) corresponds to the amount of flux of the vector field out
of the small volume divided by the volume. This is a rate of ’expansion’
or ’stretching’ of the vector field (see figure 36 above). The divergence is
div f (x) = 1, x ∈ R3 , and the figure 36 may represent a gas which is expand-
ing.

(b) f : R3 → R3 , (x, y, z) 7→ f (x, y, z) = (−x, 0, 0).

50
 Figure 37: vector field contracting

This vector field is contracting (see figure 37), and its divergence is div f (x) =
−1, x ∈ R3 .

(c) f : R3 → R3 , (x, y, z) 7→ f (x, y, z) = (0, x, 0).

Figure 38: vector field with divergence zero

51
This vector field is neither expanding nor contracting (see figure 38), and its
divergence is div f (x) = 0, x ∈ R3 .


7 Gauss’s Divergence Theorem

Theorem 7.1 (Gauss’s Divergence Theorem) Let Ω ⊂ R3 be a bounded
b . Let f : Ω ∪ ∂Ω → R3 be
region with surface ∂Ω and outward normal N
continuously differentiable. Then
Z Z
div (f )(x) dx = hf, N
b i. (7.14)
Ω ∂Ω

Remark 7.2 (a) The theorem also works for dimension n 6= 3 but one has
to define the surface integral.
R
n = 2 : ∂Ω is a line and ∂Ω hf, N b i is a (tangent) line integral.
R
(b) The left hand side of (7.14),
RRR Ω div (f )(x) dx, is just a volume Integral,
that is an iterated integral Ω div f (x, y, z) dxdydz (Fubini’s theorem,
compare Analysis I).
R R
Alternative notations are: Ω div f dV for n = 3 or Ω div f dA for
n = 2.
Example 7.3 (a) Let Ω = B(0, R) ⊂ Rn a ball of radius R > 0, B(0, R) =
{x ∈ Rn : kxk ≤ R}, and denote by ∂Ω the surface (sphere) of Ω, i.e.
∂Ω = ∂B(0, R) = {x ∈ Rn : kxk = R}. The unit outward normal is
then
N b (x) = x , for all x ∈ ∂Ω.
R
n n
Define f : R → R , x 7→ f (x) = x and check that div f (x) = n for all
x ∈ Rn . Gauss’s Divergence Theorem gives
Z Z Z
x
n dx = hx, i = R.
B(0,R) ∂B(0,R) R ∂B(0,R)

RHS = R area(B(0, R)) = R × surface area of B(0, R)

LHS = n vol(B(0, R)) = n × volume of B(0, R).
n = 2 : vol(B(0, R)) = πR2 , length(∂B(0, R)) = 2πR
4
n = 3 : vol(B(0, R)) = πR3 , area(∂B(0, R)) = 4πR2 .
3

52
 (b) Differentiability in Gauss’s Divergence Theorem is needed as the fol-
lowing example shows.
 
2 2 1 x
f : R → R , (x, y) 7→ f (x, y) = 2 .
x +y y
2

∂ x ∂ y



div f (x, y) = ∂x 2
x +y 2 + 2
∂y x +y 2 = 0. Set Ω = B(0, 1) ⊂ R2
 
2 x
having unit normal N : ∂Ω → R , (x, y) 7→ N (x, y) =
b b . We get
R y
Ω
div f (x, y) dxdy = 0 but
Z Z
hf, N i =
b 1 = length(∂Ω) = 2π.
∂Ω ∂Ω
The theorem does not work due to the singularity at the origin of the
vector field f .

Sketch for the proof of Theorem 7.1. Let Ω ⊂ R3 be a region in R3
and divide it in finitely many small regions Ωi centred at ai ∈ Ω with outward
surfaces ∂Ωi (see figure 39) having unit normal N bi .

Figure 39: partition of Ω into small regions

We apply the approximation of the divergence in (6.13) to every small

region Ωi , that is Z
1
div f (ai ) ≈ hf, N
bi i. (7.15)
vol(Ωi ) ∂Ωi

53
The approximation becomes exact in the limit vol(Ωi ) → 0. Hence, multiply
(7.15) by vol(Ωi ) and sum up to get
X XZ
div f (ai )vol(Ωi ) ≈ hf, N
b i. (7.16)
i i ∂Ωi

Here, the left hand side will become the volume integral in the limit vol(Ωi ) →
0 (Riemann sum). But what about the right hand side in (7.16)? Consider
in figure 40 two small adjacent regions Ω1 and Ω2 .

Figure 40: contributions of the common surface

Along the common surface of ∂Ω1 and ∂Ω2 we get hf, N b1 i + hf, N
b2 i = 0.
All contributions from the interior of Ω to the sum of the right hand side in
(7.16) cancel out, leaving only the surface integral over the exterior surface
b i. 2
R
∂Ω. Hence, in the limit the right hand side of (7.16) becomes ∂Ω hf, N

Example 7.4 Let

 
4x
f : R3 → R3 , (x, y, z) 7→ f (x, y, z) = −2y 2 
z2

and Ω = {(x, y, z) ∈ R3 : x2 + y 2 ≤ 4, z ∈ [0, 3]} the cylinder in figure 41. We

are going to check Gauss’s Divergence Theorem. For that we compute both
the volume integral of the divergence and the surface (flux) integral.

54
 Figure 41: Cylinder

Volume integral: We compute div f (x, y, z) = 4 − 4y + 2z.

ZZZ Z 2 h Z √4−x2  Z 3 
i
(4 − 4y + 2z) dzdydx = √
(4 − 4y + 2z dz dy dx
Ω −2 − 4−x2 0
√
Z 2 hZ 4−x2
i
= √
21 − 12y dy dx
−2 − 4−x2
Z 2 √
= 42 4 − x2 dx
2−
Z 2√
= 42 4 − x2 dx
−2
hx√ x i2
= 42 4 − x2 + 2 arcsin( ) = 84π.
2 2 −2

Surface integral: The surface S of the cylinder Ω has the following three
single parts S = Sbottom ∪ Stop ∪ Scyl.
Sbottom = {(x, y, z)R3 : x2 + y 2 ≤ 4, z = 0}:
   
0 4x
Nb (x, y, z) = − 0 and f (x, y, z) = −2y 2  for (x, y, z) ∈ Sbottom .
1 0

55
 R
As hf (x, y, z), N
b (x, y, z)i = 0 for (x, y, z) ∈ Sbottom we get
Sbottom
hf, N
b i = 0.
Stop = {(x, y, z)R3 : x2 + y 2 ≤ 4, z = 3}:
   
0 4x
b (x, y, z) = 0 and f (x, y, z) = −2y 2  for (x, y, z) ∈ Stop .
N
1 9

A parametrisation is given by
 
s cos(t)
αtop : [0, 2] × [0, 2π] → R3 , (s, t) 7→ αtop (s, t) =  s sin(t) 
3

with    
cos(t) −s sin(t)
∂αtop ∂αtop
(s, t) =  sin(t)  and (s, t) =  s cos(t) 
∂s ∂t
0 0
and cross-product  
0
∂αtop ∂αtop  
× = 0 .
∂s ∂t
s
Hence, the top contributes to the flux
Z Z 2π Z 2
hf, N i =
b 9s dsdt = 36π.
Stop 0 0

Scyl. : The parametrisation

 
2 cos(t)
αcyl. : [0, 2π] × [0, 3] → R3 , (t, z) 7→ αcyl. (t, z) =  2 sin(t) 
z

gives    
−2 sin(t) 0
∂αcyl. ∂αcyl.
(t, z) = 2 cos(t)
  and (t, z) = 0

∂t ∂z
0 1
and  
2 cos(t)
∂αcyl. ∂αcyl. ∂αcyl. ∂αcyl.
× =  2 sin(t)  ⇒ × = 2.
∂t ∂z ∂t ∂z
0

56
  
2 cos t
bcyl. (t, z) = 1 2 cos t, and the cylinder barrel contributes to the
Hence, N 2
0
flux
Z Z 2π Z 3
16 cos2 (t) − 8 sin3 (t) dzdt


hf, N i =
b
Scyl. 0 0
Z 2π Z 3
16 cos2 (t) − 8 sin3 (t) dzdt


=
Z0 2π Z0 3
16 cos2 (t)) + 8 sin(t) cos2 (t) − 8 sin(t) dzdt


=
0 0
Z 2π
16 cos2 (t)) + 8 sin(t) cos2 (t) − 8 sin(t) dt


=3
Z0 2π h1
2 1 i2π
= 48 cos (t) dt = 48 cos(t) sin(t) + t = 48π,
0 2 2 0
where we used that
Z 2π  2π
sin(t) dt = − cos(t) 0 = 0
0
Z 2π 2π
sin(t) cos2 (t) dt = − cos3 (t) 0 = 0.


0

The total flux is

Z Z Z
hf, N i =
b hf, N
bi + hf, N
b i = (36 + 48)π = 84π.
S Stop Scyl.

We have thus proved the Divergence Theorem follows for this example, i.e.
Z ZZZ
hf, N i =
b div f (x, y, z) dzdydx.
S Ω

Example 7.5 (Conservation of mass of a fluid) We consider the flow

of fluid for a domain Ω ⊂ R3 with surface ∂Ω and outward normal N
b , see
figure 42. The function
% : R3 × R+ → R
gives the density %(x, t) of the fluid a point x = (x1 , x2 , x3 ) ∈ R3 and time
t ∈ R+ . The mass of the fluid in Ω at fixed time t is the volume integral

57
RRR
Ω
%(x, t) dx1 dx2 dx3 . The rate of mass flow into the domain Ω is given by
the flux integral Z
− h%~v , N
b i,
∂Ω

where the minus sign signals that N b points outward and where ~v : R3 ×R+ →
R gives the velocity vector ~v (x, t) of the fluid at point x ∈ R3 and time t.
3

Physical law: mass is conserved, that is, the rate of change of mass
in Ω equals the rate at which mass enters Ω.

Figure 42: Flow of fluid in and out the region Ω

This law can be written as

ZZZ Z
d
%(x, t) dx = − h%~v , N
b i.
dt Ω ∂Ω

The right hand side can be written as the volume integral for the divergence of
%~v and the time derivative can be interchanged with the volume integration.
All this gives ZZZ 
∂% 
(x, t) + div (%~v )(x) dx = 0.
Ω ∂t
As the domain is arbitrary the integrand equals identically zero. Hence, we
have derived the conservation law for the mass of a fluid.

58
 Mass conservation law

∂%
+ div (%~v ) = 0.
∂t

8 Integration by Parts
This section introduces integration by parts techniques. We let Ω ⊂ R3 be
some region, i.e., a bounded open subset of R3 , and by ∂Ω we denote the
surface of Ω such that Ω = Ω ∪ ∂Ω. The surface normal is the vector field
Nb : ∂Ω → R3 , N
b (x) = (N
b1 (x), N
b2 (x), N
b3 (x)).

Lemma 8.1 Let the function f : Ω → R be continuously differentiable. Then

Z Z
∂f
(x) dx = fN
bi for i = 1, 2, 3. (8.17)
Ω ∂x i ∂Ω

Proof. Without loss of generality choose i = 1 and define the vector field
∂f
v : R3 → R3 , x 7→ v(x) = (f (x), 0, 0). Then div v(x) = ∂x 1
and v · N
b =
hv, N
bi = fNb1 . The claim follows then by the Divergence Theorem, where we
put v = (0, f, 0) and v = (0, 0, f ) for the other cases. 2

Proposition 8.2 Let the functions g : Ω → R and h : Ω → R be continuously

differentiable. Then
(a) For i = 1, 2, 3,
Integration by parts (IBP)
Z  Z Z
∂g   ∂h 
(x) h(x) dx = ghNi − g(x)
b (x) dx. (8.18)
Ω ∂xi ∂Ω Ω ∂xi

(b) If g : Ω → R is two times continuously differentiable,

Green’s first identity
Z Z
∆g(x) dx = h∇g, N
b i. (8.19)
Ω ∂Ω

59
(c) If f : Ω → R is two times continuously differentiable,
Integration by part (IBP) - vector case
Z Z Z
∆f (x)h(x) dx = bi −
hh∇f, N h∇f (x), ∇h(x)i dx (8.20)
Ω ∂Ω Ω

Proof. (a) Apply Lemma 8.1 to the function f (x) = g(x)h(x), x ∈ Ω, and
use the chain rule
∂f ∂g ∂h
(x) = (x)h(x) + g(x) (x)
∂xi ∂xi ∂xi
for i = 1, 2, 3.
∂g
(b) Use again Lemma 8.1 for f (x) = ∂x i
(x) for i = 1, 2, 3 to obtain
Z 2 Z
∂ g ∂g b
2
(x) dx = Ni .
Ω ∂xi ∂Ω ∂xi

Summing up all the terms gives the identity.

∂f
(c) Apply part (a) to the function g = ∂xi
for i = 1, 2, 3 to get
Z 2 Z Z
∂ g ∂f b ∂f ∂h
2
(x)h(x) dx = h Ni − (x) (x) dx.
Ω ∂xi ∂Ω ∂xi Ω ∂xi ∂xi
Summing up all the terms again gives the identity. 2

Application of Proposition 8.2

Let Ω ⊂ R3 be some piece of material and fix the temperature f (x) ∈ R+ for
every point x ∈ ∂Ω on the surface (boundary) of Ω. In a steady state the
temperature is a scalar field T : Ω ∪ ∂Ω → R+ solving the following boundary
value problem
∆T (x) = 0 for all x ∈ Ω,
(8.21)
T (x) = f (x) for all x ∈ ∂Ω.
The existence of solutions for (8.21) is difficult, but we can use Proposition 8.2
to show the uniqueness of a solution for (8.21). Assume that T : Ω∪∂Ω → R+
and Te : Ω ∪ ∂Ω → R+ are both a solution to (8.21). Define the scalar field
D : Ω ∪ ∂Ω → R+ , x 7→ D(x) = T (x) − Te(x). This solves the following

∆D(x) = ∆T (x) − ∆Te(x) = 0 for all x ∈ Ω

(8.22)
D(x) = T (x) − Te(x) = f (x) − f (x) = 0 for all x ∈ ∂Ω.

60
Proposition 8.2 implies that
Z Z Z
0 = (∆D(x))D(x) dx = − h∇D(x), ∇D(x)i dx + b i.
Dh∇D, N
Ω Ω ∂Ω

Henceforth (because the second integral of the right hand side vanishes)
Z
k∇D(x)k2 dx = 0.
Ω

Hence, we conclude ∇D(x) = 0 for all x ∈ Ω and therefore that D is constant.

But as D(x) = 0 for all x ∈ ∂Ω we get D(x) = 0 for all x ∈ Ω ∪ ∂Ω. Thus
T = Te, and the uniqueness of the solution of (8.21) follows.

9 Green’s theorem and curls in R2

The Divergence Theorem in Section 7 relates the flux out of region to the
volume integral of the divergence of the vector field. In this section we
ask about the flow along the boundary line of a region in R2 . The three
dimensional case will follow in the next section.

9.1 Green’s theorem

Let the vector field f : R2 → R2 , f = (f1 , f2 ), and the rectangle Ω = [a, b] ×
[c, d] ⊂ R2 , a ≤ b, c ≤ d, be given. The boundary ∂Ω of the rectangle is a
curve C in R2 and we denote the unit tangent vector in counter-clockwise
direction (that is, in positive direction of C) by Tb : C → R2 .
We are going to calculate the tangent line integral along the curve C, see
figure 43.

61
 Figure 43: rectangle Ω

We integrate each single side/edge of the rectangle having parametrisations

γR , γL , γT and γB respectively.
Z Z Z dD  E Z d
0
f= f, T =
b f (b, y), dy = f2 (b, y) dy.
γR γR c 1 c

The corresponding left hand side is

Z Z Z 1D  E Z d
0
f= f, T =
b f (a, d + t(c − d)), dt = − f2 (a, y) dy.
γL γL 0 c−d c

We get similar results for the top and the bottom edge. Summing up these
contributions we get the tangent line integral along ∂Ω
Z Z d Z b



hf, Tbi = f2 (b, y) − f2 (a, y) dy + f1 (x, c) − f1 (x, d) dx
∂Ω c a
Z dZ b Z bZ d
∂f2 ∂f1
= (x, y) dxdy − (x, y) dxdy
∂x c ∂y
Zc  a a
∂f2 ∂f1 
= − (x, y) dxdy,
Ω ∂x ∂y
where we used the Fundamental Theorem of calculus in the second line. The
integrand in the last line motivates the following definition of a scalar field.

62
Definition 9.1 Let f : R2 → R2 be a differentiable vector field. The curl of
the vector field f is the scalar field curl f : R2 → R defined by

curl f : R2 → R
∂f2 ∂f1
(x, y) 7→ curl (f )(x, y) = (x, y) − (x, y).
∂x ∂y

R
Remark 9.2 ∂Ω
hf, Tbi is the circulation around ∂Ω.
 
−y
Example 9.3 (a) The vector field v : R2 → R2 , (x, y) 7→ v(x, y) =
x
has positive curl, curl (v)(x, y) = 1 − (−1) = 2 for all (x, y) ∈ R2 .

Figure 44: positive curl

See the sketch of the vector field in figure 44, where the arrows of the
vector field are circulating in positive direction around the origin (that is,
anti-clockwise direction).  
2 2 −x
(b) The vector field v : R → R , (x, y) 7→ v(x, y) = has vanishing
−y
curl, curl (v)(x, y) = 0 for all (x, y) ∈ R2 . See the sketch of the vector field
in figure 45, where the arrows are directed towards the origin.

63
 Figure 45: zero curl

 
0
(c) The vector field v : R2 → R2 , (x, y) 7→ v(x, y) = has positive curl,
x
curl (v)(x, y) = 1 for all (x, y) ∈ R2 . In figure 46 the arrows are pointing
upwards in the positive half plane and downwards in the negative half plane.

64
 Figure 46: positive curl, a small wheel centred at y-axis would
turn


The calculation above for the line integral along the boundary of a rectangle
showing that the integral is the surface integral of the curl of the vector field
can be proved for any region Ω ⊂ R2 . This is the content of the following
theorem.
Theorem 9.4 (Green’s theorem - Stokes’s theorem in R2 ) Let Ω ⊂ R2
be a bounded region and Tb : ∂Ω → R2 be positively oriented unit tangent
vectors for the boundary line ∂Ω of the region Ω. Let Ω = Ω ∪ ∂Ω. If
f : Ω → R2 , f = (f1 , f2 ), is continuously differentiable, then
Z Z
hf, Tbi = curl (f )(x, y) dxdy. (9.23)
∂Ω Ω

Proof. We give a sketch of the proof only, because one can use the proof of
the Divergence Theorem for it. The contributions of the interior cancel out
here as they do for the Divergence Theorem. To apply that theorem directly
define the vector field
 
2 2 f2 (x, y)
v : R → R , (x, y) 7→ v(x, y) = .
−f1 (x, y)

65
Then,
∂f2 ∂f1
div v(x, y) = (x, y) − (x, y) = curl f (x, y),
∂x ∂y
and !
−N D
b2 E
hv, N
b i = f2 N
b1 − f1 N
b2 = f, = hf, Tbi.
Nb1

Note that Tb ⊥ N b . Applying Gauss’s theorem 7.1 we conclude

Z Z Z Z
curl f (x, y) dxdy = div v(x, y) dxdy = hv, N i =
b hf, Tbi.
Ω Ω ∂Ω ∂Ω

2
In the next proposition we show how one can apply Green’s theorem.

Proposition 9.5 Let Ω ⊂ R2 be a bounded region with closed boundary

2 2
 surrounds Ω. Define the vector field f : R → R , (x, y) 7→
line ∂Ω, which
−y
f (x, y) = . Then the area of the region is given by the tangent line
x
integral of f , that is,
Z
1
area (Ω) = hf, Tbi. (9.24)
2 ∂Ω

Proof.
Z ZZ ZZ
hf, Tbi = (1 + 1) dxdy = 2 dxdy = 2area (Ω).
∂Ω Ω Ω

9.2 Application
Proposition 9.5 has a direct application for the construction and the use of
a planimeter. A planimeter is a measuring instrument used to measure the
surface area of an arbitrary two-dimensional shape. The most common use
is to measure the area of a plane shape. There are many different kinds of
planimeters but all operate in a similar way. A pointer on the planimeter is
used to trace around the boundary of the shape. This induces a movement
in another part of the instrument and a reading of this is used to establish
the area of the shape. The precise way in which they are constructed varies,
the main types of mechanical planimeter being polar; linear; and Prytz or

66
”hatchet” planimeters. In the linear and polar planimeter, as one point on
a linkage is traced along the shape’s perimeter, that linkage rolls a wheel
along the drawing. The area of the shape is proportional to the number
of turns through which the measuring wheel rotates when the planimeter is
traced along the complete perimeter of the shape. The concept having been
pioneered by Hermann in 1814, Swiss mathematician Jakob Amsler-Laffon
built the first modern planimeter in 1854, the operation of which can be
justified by appealing to Green’s theorem and in particular to Proposition 9.5.
Let us study briefly the so-called linear planimeter, see figure 47 below. The
measuring arm (elbow) can move up and down along the y-axis. Note that
b is the y-coordinate of that measuring arm. The scalar product of that arm
with a vector field N~ = (Nx , Ny ) is

~ ,N
hEM ~ i = xNx + yNy .

Having N~ (x, y) = (b − y, x) we get hEM ~ ,N

~ i = 0 and that the length of the
p
measuring arm, m := (b − y) + x , is constant. Let Ω be a region in R2
2 2

having boundary ∂Ω. We conclude with Green’s theorem that

Z Z Z
~ , Tbi = ∂N y ∂N x

hN − dxdy = (1 + 1) dxdy = 2area(Ω).
∂Ω Ω ∂x ∂y Ω

Example 9.6 Verify Green’s theorem for the following vector field
 
2 2 xy + y 2
f : R → R , (x, y) 7→ f (x, y) = ,
x2

and the region Ω = {(x, y) ∈ [0, 1] × [0, 1] : x ≥ y ≥ x2 }, see figure 48

67
Figure 47:

68
Prytz

69
Planimeter

Planimeter

70
 Figure 48: region Ω

curl f (x, y) = x − 2y gives

Z Z 1Z x 
curl f (x, y) dxdy = (x − 2y) dy dx
Ω 2
Z0 1 h x ix
= xy − y 2 dx
x2
Z0 1
1
= (x4 − x3 ) dx = − .
0 20

We split the boundary in two pieces ∂Ω = ∂Ω1 ∪ ∂2 (see figure 48) with
parametrisation    
t 1
∂Ω1 : γ1 : [0, 1] → R2 , t 7→ γ1 (t) = 2 , γ10 (t) =
t 2t
   
2 1−t 0 −1
∂Ω2 : γ2 : [0, 1] → R , t 7→ γ2 (t) = , γ2 (t) =
1−t −1
Note that
γ10 (t)
 
1 1
T1 (γ1 (t)) = 0
b =√ .
kγ1 (t)k 1 + 4t 2t
2

Then,

71
 Z Z 1 D E
hf, Tb1 i = f (γ1 (t)), Tb1 (γ1 (t)) kγ10 (t)k dt
∂Ω1 0
Z 1D 3   E
t + t4 1
= 2 , dt
0 t 2t
Z 1
19
t4 + 3t3 dt = .


=
0 20
and
1 D (1 − t)2 + (1 − t)2  −1 E√
Z Z  
1
hf, Tb2 i = ,√ 2 dt
∂Ω2 0 (1 − t)2 2 −1
Z
= −3 (1 − t)2 dt = −1.

1
R
This gives ∂Ω
hf, Tbi = − 20 and thus the proof of Green’s theorem for this
example. 

10 Stokes’s theorem
Stokes’s theorem gives an alternative expression for the surface integral of
the curl of a vector field. This is given by the flow along the boundary.

10.1 Stokes’s theorem

Theorem 10.1 (Stokes’s theorem) Let S ⊂ R3 be a bounded surface,
Nb : S → R3 be unit normal vectors of the surface S and Tb : ∂S → R3 unit
tangent vectors of the boundary ∂S, and let f : S ∪∂S → R3 be a continuously
differentiable vector field. If (N
b , Tb) is positively oriented, then
Z Z
hf, Tbi = hcurl f, N
b i. (10.25)
∂S S

Remark 10.2 (a) Positively oriented means that the right hand rule for the
pair (N
b , Tb) is satisfied. That is, if you walk around the boundary, the surface
should be on your left.
(b) curl

72
f : R3 → R3 , f = (f1 , f2 , f3 ) differentiable, then the curl in R3 is the vector
field
 
∂2 f3 (x, y, z) − ∂3 f2 (x, y, z)
curl f : R3 → R3 , (x, y, z) 7→ curl (f )(x, y, z) = ∂3 f1 (x, y, z) − ∂1 f3 (x, y, z) .
∂1 f2 (x, y, z) − ∂2 f1 (x, y, z)

Alternative notation: curl f = ∇ × f .

If g : R2 → R2 , g = (g1 , g2 ) is differentiable we have a scalar field

curl g(x, y) = ∂x g2 − ∂y g1 (x, y),

and, if we define the vector field

 
g1 (x, y)
f : R3 → R3 , (x, y, z) 7→ f (x, y, z) = g2 (x, y) ,
0
 
0
we get curl f (x, y, z) =  0 . Green’s theorem 9.4 in R2 is just a special
curl g
case of Stokes’s theorem in R3 where S ⊂ R × R × {0} (the x − y-plane).

Proof of Theorem 10.1. Following the last remark we prove the theorem
by ’lifting’ Green’s theorem from R2 to R3 . On the left hand side in figure
49 we have the 2d-world and on the right hand side of figure 49 we have the
3d-world.

73
 Figure 49: lifting from R2 to R3

2d-world: Let Ω ⊂ R2 be a domain (open and bounded subset) with ’smooth’

boundary ∂Ω and denote a parametrisation for the boundary ∂Ω by γ : [a, b] →
R2 .

3d-world: Let α : Ω ∪ ∂Ω → S ∪ ∂S, (s, t) 7→ α(s, t), be a parametrisation

(lifting map from R2 to R3 ) for bounded ’smooth’ surface S. The boundary
∂S can parametrised by
α ◦ γ : [a, b] → R3 , u 7→ α(γ(u)) = α((γ1 (u), γ2 (u))

with unit tangent vector TbS given by the chain rule

1 1  ∂α ∂γ (u) ∂α ∂γ (u) 
1 2
TbS (γ(u)) = 0
α0 (γ(u)) = ∂α ∂γ1 (u) ∂α ∂γ2 (u)
+ .
kα (γ(u))k k ∂s ∂u + k ∂s ∂u ∂t ∂u
∂t ∂u

Note that the norm drops out when we compute the tangent line integral.
Henceforth,
Z Z bD
∂α γ1 ∂α ∂γ2 E
hf, TS i =
b f (α(γ(u))), (γ(u)) (u) + (γ(u)) (u) du
∂S a ∂s ∂u ∂t ∂u
! !
hf, ∂α i hf, ∂α
i
Z bD E Z D E
∂s 0 ∂s
= , γ (u) du = , T
bΩ du.
a hf, ∂α
∂t
i ∂Ω hf, ∂α
∂t
i

74
Similarly we get (after some computation and application of the chain rule)
Z ZZ D
∂α ∂α E
hcurl f, N
bi = curl f, × dsdt
S Ω ∂s ∂t
!
hf, ∂α i
ZZ
∂s
= curl ∂α
dsdt.
Ω hf, ∂t i

The result follows now from Green’s theorem 9.4.

2

Example 10.3 We prove Stokes’s theorem for the following example. Let
the surface S = {(x, y, z) ∈ R3 : z = x2 + y 2 , z ≤ 4} with parametrisation
 
x
α : B(0, 2) → R3 , (x, y) 7→ α(x, y) =  y  ,
x2 + y 2

where B(0, 2) = {(x, y) ∈ R2 : x2 + y 2 ≤ 4} is the ball around the origin

having radius 2. We get the surface unit normal N b as the gradient of the
2 2
function s(x, y, z) = z − x − y (level set), that is,

 
−2x
b (x, y, z) = ∇s(x, y, z) = 1 −2y  ,
N (x, y, z) ∈ S = α(B(0, 2)).
k∇s(x, y, z)k 3
1

The boundary ∂S is parametrised by

 
2 cos t
γ : [0, 2π] → R3 , t 7→ γ(t) =  2 sin t  .
4
 
y
Let the vector field f : R3 → R3 , (x, y, z) →7 f (x, y, z) =  x2  be given.
3z 2
Some calculation gives then
   
Z Z 2π D 2 sin t −2 sin t E
hf, Tbi = 4 cos2 t ,  2 cos t  dt = −4π.
∂S 0 48 0
 
0
curl (f )(x, y, z) =  0  ,
2x − 1

75
and therefore

   
Z Z 0 −2x E
 0  , 1 −2y  3 dxdy
D
hcurl f, N
bi =
S B(0,2) 3
2x − 1 1
Z Z
= 2x dxdy − dxdy = 0 − vol(B(0, 2)) = −4π.
B(0,2) B(0,2)

Thus we have shown Stokes’s theorem for this example. 

Remark 10.4 (a) If f : R3 → R3 is a differentiable gradient vector field

(that is, there is a potential V : R3 → R with ∇V (x) = f (x)), we get
Z Z
hcurl f, N i =
b hf, Tbi = 0
S ∂S

either by FTC for line integrals or by observing

   ∂V   ∂ 2 V ∂2V

∂x ∂x ∂y∂z
− ∂z∂y
    ∂ V  2 ∂2V 

curl grad V (x, y, z) = ∂y  ×  ∂V
∂y 
=  ∂z∂x − ∂x∂z  = 0,
∂z ∂V ∂2V ∂2V
∂z ∂x∂y
− ∂y∂x

where we assume that V : R3 → R is two times continuously differentiable.

(b) If the vector field f : R3 → R3 can be represented as the curl of some

vector field v : R3 → R3 , that is, f = curl v, we get
Z Z
hf, N i =
b div f (x, y, z) dxdydz = 0
∂Ω Ω

by observing div curl v(x, y, z) = 0 for all (x, y, z) ∈ R3 .

Example 10.5 (Faraday’s law) Let the time-dependent electric (E) and
magnetic (H) field be given,

E : R × R3 → R3 ,
H : R × R3 → R3 .

For a bounded surface S ⊂ R3 with closed boundary ∂S = C the voltage

around C is given by the tangent-line integral
Z
hE, Tbi,
C

76
which is time-dependent, and the time-dependent magnetic flux is given by
the flux integral Z
hH, N
b i.
S

Fraday’s law: voltage around C equals the negative rate of change

of magnetic flux through S.

We show that Farday’s law follows from ∇ × E = − ∂H∂t

. By Stokes’s theorem
Z Z
hE, Tbi = h∇ × E, N b i,
C S

which implies that

Z Z Z
∂
− hH, N
bi = h∇ × E, N
bi = hE, Tbi,
∂t S S C

which gives finally the mathematical form of Faraday’s law

Z Z
∂
hE, Tbi = − hH, N
b i.
C ∂t S

We finish this section with a summary on conservative vector fields.

Theorem 10.6 (Conservative vector fields) Let f : R3 → R3 be a C 1 -

vector field. Then the following conditions are equivalent.
R
(i) For any oriented closed simple curve C ⊂ R3 : C hf, Tbi = 0.

(ii) For any two oriented simple closed curves C1 ⊂ R3 and C2 ⊂ R3 having
R R
the same starting and terminal point: C1 hf, Tb1 i = C2 hf, Tb2 i.

(iii) f is the gradient of some C 2 scalar field V : R3 → R.

(iv) ∇ × f (x) = curl (f )(x) = 0 for all x ∈ R3 .

The vector field f is called conservative is one of the conditions (i)-(iv) is

satisfied.

77
10.2 Polar coordinates
As a reminder we summarise briefly important facts about spherical coordi-
nates.
Polar coordinates in R3
 
r cos ϕ cos θ
x : [0, ∞) × [−π, π] × [0, 2π] → R3 , (r, θ, ϕ) 7→ x(r, θ, ϕ) =  r sin ϕ cos θ  .
r sin θ
Here the single coordinates have the following names.
r ∈ [0, ∞) is called the radius,
θ ∈ [−π, π] is called the (geographical) latitude,
ϕ ∈ [0, 2π] is called the (geographical) longitude.
Compare with Example 5.2 part (c) where we defined iteratively polar co-
ordinates for any dimension. A function f : R3 → R can be given in polar
coordinates by writing
f : [0, ∞) × [−π, π] × [0, 2π] → R, (r, θ, ϕ) 7→ f (r, θ, ϕ) = f (x(r, θ, ϕ)).

Basis vectors
 
cos(ϕ) cos(θ)
∂x  er
er = =  sin(ϕ) cos(θ)  , ker k = 1, êr = = er ,

∂r ker k
sin(θ)

−r cos(ϕ) sin(θ) − cos(ϕ) sin(θ)

   
∂x 
eθ = =  −r sin(ϕ) sin(θ)  , keθ k = r, êθ =  − sin(ϕ) sin(θ)  ,
  
∂θ
r cos(θ) cos(θ)

−r sin(ϕ) cos(θ) − sin(ϕ)

   
∂x 
eϕ = =  r cos(ϕ) cos(θ)  , keϕ k = r cos(θ), êϕ =  cos(ϕ)  .
  
∂ϕ
0 0

The system (êr , êϕ , êθ ) is a right hand orthonormal system.

Warning: In some textbooks a slightly modified version is used. Instead of

θ (geographical latitude) one uses Θ ∈ [0, π]. Then π/2 − Θ is the latitude,
i.e. θ = π/2 − Θ.

78
 Any vector field can be expressed in this basis,

V : [0, ∞)×[−π, π]×[0, 2π] → R3 , (r, θ, ϕ) 7→ Vr (r, θ, ϕ)êr +Vθ (r, θ, ϕ)êθ +Vϕ (r, θ, ϕ)êϕ .

Example 10.7 The vector field V : R3 → R3 , x 7→ x can be written as

V : [0, ∞) × [−π, π] × [0, 2π] → R3 , (r, θ, ϕ) 7→ V (r, θ, ϕ) = rêr .

Warning:
∂Vr ∂Vθ ∂Vϕ
div V 6= + + .
∂r ∂θ ∂ϕ
Applying the chain rule one obtains after some calculation the differential
operators.
Differential operators:

∂f 1 ∂f 1 ∂f
grad (f ) = êr + êφ + êθ
∂r r cos θ ∂φ r ∂θ

1 ∂ 2
1 ∂vφ 1 ∂

div (v) = r vr + + cos(θ)vθ
r2 ∂r r cos θ ∂φ r cos θ ∂θ

1  ∂vθ ∂
 1  ∂vr ∂ 
curl (v) = − cos(θ)vφ êr + − rvθ ) êφ
r cos θ ∂φ ∂θ r ∂θ ∂r
1 ∂
1 ∂vr 
+ rvφ − êθ
r ∂r cos θ ∂φ

79
 Part II

Introduction to Complex Analysis

11 Complex Derivatives and Möbius trans-
formations
The second part of the lecture gives an introduction to complex analysis. In
the first section we study complex derivatives. The field of complex numbers
is the set
C = {z = x + iy : x, y ∈ R} with i2 = −1.
Our aim is to understand the calculus for complex-valued functions f : C → C
on the complex numbers. The link to our vector analysis part is given by the
representation of any complex number in real and imaginary part,

f (z) = f (x + iy) = Re(f )(x, y) + iIm(f )(x, y) = u(x, y) + iv(x, y),

where
Re(f ) = u : R2 → R2 , (x, y) 7→ u(x, y) is the real part,
Im(f ) = v : R2 → R2 , (x, y) 7→ v(x, y) is the imaginary part.

There is a bijection between any function f : C → C and two two-

dimensional vector fields u, v : R2 → R2 . This is the connection with the
first part of the course. We will prove some results via the corresponding
results for real vector fields. Reminder:

80
 Figure 50: Graphical representation of a complex number

From figure 50 we get

z = x + iy = Re(z) + iIm(z)
p
= r cos(ϕ) + i sin(ϕ) with r = x2 + y 2 = |z|

and the argument ϕ = arctan( xy ) = arg(z). We call x = Re(z) the real part
p
and y = Im(z) the imaginary part of z = x + iy ∈ C and |z| = x2 + y 2 the
modulus of z = x + iy ∈ C.

Addition: (x1 + iy1 ) + (x2 + iy2 ) = (x1 + x2 ) + i(y1 + y2 ), x1 , x2 , y1 , y2 ∈ R.

Multiplication: Let z = r cos(ϕ) + ir sin(ϕ), w = s cos(ψ) + is sin(ψ) ∈ C,

then z · w = rs cos(ϕ + ψ) + irs sin(ϕ + ψ) ∈ C.

Complex conjugate: The complex conjugate of z = x + iy ∈ C is z =

x − iy ∈ C with the following rules for z, w ∈ C

(i) z = z.

(ii) z + w = z + w.

(iii) zw = zw.

81
 (iv) |z| = |z|.

(v) |z|2 = zz.

Inverse: z ∈ C, z 6= 0, the inverse is defined as

1 x y 1
= 2 − i = z.
z (x + y 2 ) (x2 + y 2 ) |z|2

Moivre-Laplace:

eiϕ = cos(ϕ) + i sin(ϕ) for ϕ ∈ R.

Any complex number z ∈ C can bepwritten in the polar form, i.e. as

z = |z|eiϕ where ϕ = arg(z) and |z| = Re(z)2 + Im(z)2 . Let n ∈ N and
z = |z|eiϕ ∈ C

z n = 1 ⇔ |z|n einϕ = 1
⇔ |z| = 1 and cos(nϕ) + i sin(nϕ) = 1
2kπ
⇔ |z| = 1 and ϕ = k = 0, 1, . . . , n − 1.
n
Hence, the distant roots of z n = 1 are given by
2kπi
zk = e n , k = 0, 1, . . . , n − 1,

and they are called the n-th roots of unity.

Definition 11.1 Let (zn )n∈N be a sequence in C, that is zn ∈ C for all n ∈ N,

and z ∈ C. The sequence (zn )n∈N converges to z as n → ∞ if and only if the
real sequence (|zn − z|)n∈N converges to 0 as n → ∞. Equivalent formulation
with zn = xn + iyn , z = x + iy:

zn → z as n → ∞ ⇔ |zn − z| → 0 as n → ∞
⇔ |xn − x| → 0 as n → ∞ and |yn − y| → 0 as n → ∞.

Note that | · | is used in the first line as the modulus for complex numbers and
in the second line as the modulus for real numbers.

Lemma 11.2 (Rules for converging sequences) Let (zn )n∈N , (wn )n∈N be
complex sequences with zn → z ∈ C and wn → w ∈ C as n → ∞. then

82
 (a) zn + wn → z + w as n → ∞.
(b) zn wn → zw as n → ∞.
zn z
(c) wn
→ w
as n → ∞ if w 6= 0.

Proof. Exercise, use the following lemma. 2

Lemma 11.3 (Inequalities) Let z, w ∈ C.
(a) |Re(z)| ≤ |z|, |Im(z)| ≤ |z|.
(b) |z + w| ≤ |z| + |y| Triangle Inequality.
(c) |z + w| ≥ |z| − |w| .

Proof. Exercise. 2
Notation 11.4 Let ε > 0. The set
B(z, ε) = {z 0 ∈ C : |z 0 − z| < ε}
is the open ball (circle) in C around z ∈ C with radius ε, see figure 51. The
circle line is ∂B(z, ε) = {z 0 ∈ C : |z 0 − z| = ε} and the closed ball is
B(z, ε) = B(z, ε) ∪ ∂B(z, ε) = {z 0 ∈ C : |z 0 − z| ≤ ε}.

Figure 51: ball of radius ε

83
 The circle centre a ∈ C and radius r > 0 is the locus of points at distance
r from a so has equation |z − a| = r. There is another equation: let α, β ∈ C
with α 6= β and let λ ∈ R, with λ > 0 and λ 6= 1. The equation

z−α
= λ, λ > 0, λ 6= 0, (11.26)
z−β

represents a circle. This can be seen using α = α1 + iα2 , β = β1 + iβ2 , and

z = x + iy. The equation (11.26) can be rewritten as |z − α|2 = λ2 |z − β|2 ,
and simplifications lead indeed to the equation
 α1 − λ2 β1 2  α2 − λ2 β2 2
x− + y − = c, for some c > 0.
1 − λ2 1 − λ2
We shall investigate the geometric significance of the points α and β in
(11.26), which are known as inverse points with respect to the circline.
First note that for λ = 1 we have |z − α| = |z − β|. That is, we have a line `.
The points α and β are reflections of each other in `. Let λ 6= 1. There are
exactly two points on the circle which are collinear with α and β. We call
them z1 and z2 respectively and they satisfy

z1 − α = λ(z1 − β) and z2 − α = −λ(z2 − β).

Figure 52: circles from inverse-point representation

84
 These points lie at opposite ends of a diameter and writing the equation
of the circle in the form |z − a| = r we get
1 1
a = (z1 + z2 ) and r = |z1 − z2 |.
2 2
Hence
1 1
α − a = λ(z2 − z1 ) and λ(β − a) = (z2 − z1 ).
2 2
Thus
1
(α − a)(β − a) = (z2 − z1 )(z2 − z1 ) = r2 .
4
The points α and β in C are inverse points with respect to the circle
|z − a| = r if and only if they satisfy (α − a)(β − a) = r2 . Note that we must
always have one of α and β inside the circle and the other outside see figure
52.

The extended complex plane and the Riemann sphere

We next introduce an ingenious device, which will allow us to treat lines

and circles, and half-lines and circular arcs, in a unified way. Consider C as
embedded in Euclidean space R3 by identifying z = x + iy with the point
(x, y, 0) ∈ R3 . Furthermore, define
1 1
Σ = {(x, y, w) ∈ R3 : x2 + y 2 + (w − )2 = }.
2 4
This is a sphere centered at (0, 0, 12 ) having radius 12 , called the Riemann
sphere. It touches the complex plane C at the point S = (0, 0, 0) (the ’south
pole’). Stereographic projection allows us to set up a one-to-one correspon-
dence between the points of C and the points of Σ, excluding N = (0, 0, 1),
the ’north pole’ of Σ. Geometrically, the line from any point z of C to the
north pole N cuts Σ \ {N } in precisely one point z 0 , and, for every point z 0
of Σ \ {N }, the line through N and z 0 meets the plane C in a unique point z.
The irritation of the north pole being ’left out in the cold’ can be removed:
just add to C an extra point ∞ ∈ / C and define the extended complex
plane C to be C ∪ {∞}.
e

85
 Figure 53: Riemann sphere - stereographic projection

Correspondence C
e and Σ:

C 3 z = x + iy = reiθ ←→ z 0 = (x(1 + r2 )−1 , y(1 + r2 )−1 , r2 (1 + r2 )−1 ),

∞ ←→ (0, 0, 1).

We shall add some arithmetic rules for the extended complex plane C:
e

a ± ∞ = ± + a = ∞, a/∞ = 0, for all a ∈ C,

a∞ = ∞a = ∞, a/0 = ∞, for all a ∈ C \ {0},
∞ + ∞ = ∞∞ = ∞.

e than in C, by regarding ∞ as being

We shall consider lines rather in C
adjoined to any line in C. Having the above stereographic correspondence,
circles on Σ which pass through the north pole N project down onto lines in
e Any circle drawn on Σ parallel to the horizontal plane w = 0, necessarily
C.
with centre on the vertical axis x = y = 0, projects down to a circle in
C (with centre 0). Clearly, any circle on Σ which does not pass through N

86
projects onto to a circle in C, and that every circle in C arises in this manner.
It is thus natural to regard lines as ’circles through infinity’, and to adopt
the collective name circline for a circle or straight line in C. e Henceforth,
(11.26) with λ > 0 represents a circline which is a line if λ = 1 and a circle
(in C) otherwise.

Möbius transformations

We introduce a family of mappings of the extended complex plane onto itself

which map circlines to circlines. First some particular mappings.

z 7→ zeiϕ (ϕ ∈ R) anticlockwise rotation through ϕ,

z 7→ Rz (R > 0) stretching by a factor R,
z 7→ z + a (a ∈ C) translation by a,
z 7→ 1/z inversion.

It is easy to see that mappings of the first three types take straight lines to
straight lines and circles to circles. Consider the inversion z 7→ w = 1/z
as a map from C e to C.
e The image of the line Re z = 1 is described in
inverse-point form as |z| = |z − 2| (points equidistant from 0 and 2). We get
that |1/w| = |1/w − 2|, that is, |2w − 1| = 1. As inversion is self-inverse,
we see also that the circle |2z − 1| = 1 maps to the line Re w = 1 under
z 7→ w = 1/z. Henceforth under inversion a line may map to a circle and a
circle may map to a line.

Definition 11.5 A Möbius transformation is a mapping of the form

az + b
z 7→ w = f (z) = , a, b, c, d ∈ C, ad − bc 6= 0.
cz + d

This mapping is viewed as a mapping f from C

e to C,
e when putting f (−d/c) =
∞ and f (∞) = a/c, according to the above mentioned rules in C. e Then
e→C
f: C e is one-to-one and onto, with a well-defined inverse given by

f −1 : C e f −1 (w) = dw − b .
e → C,
a − cw
Proposition 11.6 (Circlines under Möbius transformations) Let C be
a circline with inverse points α and β in C and let f be a Möbius transfor-
mation. Then f maps C to a circline, with inverse points f (α) and f (β).

87
Remark 11.7 In real analysis convergence of a sequence is defined with the
so-called ε − n0 -criterion. For complex valued sequences (zn )n∈N it reads as
follows:

zn → z ∈ C as n → ∞ ⇔ ∀ε > 0 ∃ n0 (ε) ∈ N ∀n ≥ n0 (ε) : zn ∈ B(z, ε).

Definition 11.8 A function f : D → C, D ⊂ C, is continuous at z ∈ D if

for any sequence (zn )n∈N in C with zn → z as n → ∞ one gets that f (zn ) →
f (z) as n → ∞. The function f is continuous on D if it is continuous at
each z ∈ D.

Remark 11.9 The ε − δ-criterion reads as

f : D → C is continuous at z ∈ D
⇔ ∀ε > 0 ∃ δ > 0 : w ∈ D and |z − w| < δ ⇒ |f (z) − f (w)| < ε
⇔ ∀ε > 0 ∃ δ > 0 : w ∈ B(z, δ) ⇒ f (w) ∈ B(f (z), ε).

Note that δ = δ(ε, z).

Definition 11.10 The function f : D → C, D ⊂ C, is said to be differen-

tiable at z ∈ D if the limit (differential quotient)

f (z + h) − f (z)
f 0 (z) := lim
h→0 h

exists, that is, if the limit of the sequence f (z+hhnn)−f (z) exists for every sequence
(hn )n∈N in C with hn → 0 as n → ∞, and this limit is (when it exists) denoted
by f 0 (z).

Lemma 11.11 (Rules) Let f, g : D → C, D ⊂ C, be continuous (differen-

tiable) at z ∈ D. Then f +g, f g, f /g (for g(z) 6= 0 for z ∈ D) are continuous
(differentiable) at z ∈ D.

So far there seems to be no much difference with real analysis. But the
following example sheds some light onto the new features of complex analysis.

Example 11.12 (a) f : C → C, z 7→ f (z) = z is continuous on C and,

because of
f (z + h) − f (z)
= 1, h ∈ C, h 6= 0,
h
differentiable at any z ∈ C with f 0 (z) = 1.

88
(b) f : C → C, z 7→ f (z) = z = x − iy for z = x + iy. As

zn = xn + iyn → z as n → ∞ ⇒ xn → x and yn → y as n → ∞
⇒ z n → z as n → ∞,

we get that f is continuous on C. Pick z ∈ C, for any h ∈ C, h 6= 0, we have

f (z + h) − f (z) z+h−z h
= = .
h h h

If we choose hn = n1 , n ∈ N, we get hn
hn
= 1 for any n ∈ N. But if we choose
hn = i n1 , n
∈ N, we get hn
= −1. Hence, the limit of the differential quotient
hn
does not exist and therefore f is not differentiable at z ∈ C. 

Example 11.12(b) shows that in complex analysis differentiability of complex-

valued function on C is not equivalent to the differentiability of the real and
imaginary part alone.
Warning: Let f = u+iv be a complex valued function on C, z = x+iy ∈ C,

f continuous on C ⇔ u, v continuous on R2 is true!

f differentiable at z ∈ C ⇔ u, v differentiable at (x, y) ∈ R2 is false!

12 Complex Power Series

In the last section we learned that differentiability for complex valued func-
tions on C is not equivalent to the differentiability of the real and imaginary
parts alone. In this section we study another route to complex calculus, the
theory for complex power series. In real analysis you heard about real power
series and that some real functions can be expressed as a power series (Taylor
series).

Definition 12.1 (a) Let cn ∈ C, n ∈ N0 and c ∈ C. The series ∞

P
n=0 cnPcon-
PN ∞
verges if and only if theP
sequence (sN )N ∈N , sN = P n=0 cn , converges. If n=0
converges (in symbols ∞ n=0 cn < ∞) we call c =
∞
n=0 cn = limN →∞ sN the
sum.
(b) Cauchy-criterion: Let (zn )n∈N , zn = xn + iyn , be a complex sequence.

(zn )n∈N is a Cauchy sequence

⇔ ∀ε > 0 ∃ n0 (ε) ∈ N : ∀m, n ≥ n0 (ε) : |zm − zn | < ε
⇔ (xn )n∈N , (yn )n∈N are Cauchy sequences in R.

89
The Cauchy criterion is used in the proof of the following lemma.

Lemma 12.2 Let (fn )n∈N be a sequence of functions fn : D → C, D ⊂ C,

P∞ Mn ≥ 0 for all n ∈ N. If |fn (z)| ≤ Mn for all z ∈ D, n ∈ N, and
and
n=0 Mn < ∞, then the series

∞
X
f (z) := fn (z)
n=0

converges for all z ∈ D. If all fn are continuous functions then so is the

function f defined via the sum of the converging series for any z ∈ D.
PN
Proof. Pick z ∈ D and let sN (z) := n=0 fn (z). Then (w.l.o.g. M > N )
M
X M
X
|sM (z) − sN (z)| = fk (z) ≤ Mk .
k=N +1 k=N +1

The converges of ∞
P
n=0 Mn implies that Mn → 0 as n → ∞. Hence, (|sM (z)−
sN (z)| → 0 as N, M → ∞, and thus f (z) exists (i.e., the sum converges).
For the proof of the continuity part see real analysis. 2

Definition 12.3 (Power series) A power series is defined to be a series of

P∞
the form n=0 cn (z − a)n , a ∈ C, and cn ∈ C for all n ∈ N0 . W.l.o.g. we
often assume a = 0.
P∞
(a) The radius of convergence of the power series n=0 cn (z − a)n is
defined to be
∞
X
cn (z − a)n converges .

R = sup |z − a| : z ∈ C with (12.27)
n=0

We write R = ∞ if ∞ n
P
n=0 cn (z − a) converges for arbitrarily large |z|,
i.e., for all z ∈ C.

(b) The series P ∞ n

P
n=0 cn (z − a) is said to be absolutely convergent if the
real series ∞ n
n=0 |cn (z − a) | is convergent.

Theorem 12.4 (Power series Theorem) Let (cn )n∈N be a sequence in C.

Then there is a R ∈ [0, ∞] such that
P∞ n
(a) n=0 cn z converges for z ∈ C with |z| < R,

90
 P∞ n
(b) n=0 cn z does not converge for z ∈ C with |z| > R.

(c) If n=0 cn (z − a)n converges for z1 ∈ C then ∞

P∞ P n
n=0 |cn ||(z − a)| con-
verges for all z ∈ C with |z − a| < |z1 − a|.

(d) If ∞
P n
P∞ n
n=0 cn (z − a) diverges for z1 ∈ C then n=0 cn (z − a) diverges
for all z ∈ C with |z − a| > |z1 − a|.

(e) On the open ball B(0, R) a function is defined as

∞
X
f : B(0, R) → C, z 7→ f (z) = cn z n .
n=0

The function f is continuous and differentiable on B(0, R) with deriva-

tive given by term-by-term differentiation:
∞
df X
(z) = cn nz n−1 .
dz n=1

Proof. The proof for (a)-(d) is easy. We sketch it for (c): If ∞

P
n=0 cn (z1 −
a)n converges we know that cn (z1 − a)n → 0 as n → ∞. Hence, there is
M > 0 such that |cn (z1 − a)|n ≤ M . We get
z−a n z−a n
|cn (z − a)| = |cn (z1 − a)n | ≤M ,
z1 − a z1 − a
and if the right hand side gives a convergent series (geometric series) we get
the statement for |z − a| < |z1 − a|. The remaining part will be proved later.
The proof is similar to the corresponding one for real analysis. 2

Remark 12.5 (a) The series may or may not converge for |z| = R. This
has to be analysed in detail for any example. Recall that ∂B(0, R) is the circle
line of radius R around the origin and B(0, R) is the closed ball. The set of
complex numbers for which the power series converges includes the open ball
B(0, R) and possibly some points from the boundary line ∂B(0, R).

(b) A similar theorem holds in R.

(c) One can apply the theorem repeatedly:

f is C ∞ on B(0, R) with
dk f X
(z) = cn n(n − 1) · · · (n − k + 1)z n−k .
dz k n=k

91
(d) We have put a = 0 in the theorem because, by shifting,
∞
X
g(z) := cn (z − a)n
n=0

converges for all z ∈ B(a, R) and f (z) = g(z + a), z ∈ B(0, R).

Two methods for the calculation of the radius of convergence

A d’Alembert’s Ratio test:

If
|cn+1 |
→ α ∈ [0, ∞] as n → ∞,
|cn |
P∞ P∞
then
P∞ n=0 |c n | converges if α < 1 (i.e. n=0 cn converges absolutely), and
n=0 |cn | diverges if α > 1. If α = 1 then the test gives no information.

B Cauchy’s n-th root test:

If p
n
lim |cn | = α ∈ [0, ∞),
n→∞
P∞ P∞
then
P∞ n=0 |cn | converges if α < 1 (i.e. n=0 cn converges absolutely), and
n=0 |cn | diverges if α > 1. If α = 1 then the test gives no information.
Proof.
A: Let α < 1. Then there a q ∈ (α, 1). Hence, there is n0 ∈ N such that
cn+1
cn
< q for all n ≥ n0 . W.l.o.g. we can assume that the last statement
holds for all n ∈ N as any addition of finitely many terms does not change
the convergence properties of the series. We get
cn+1 cn c2 cn+1
··· = , ∀n ∈ N.
cn cn−1 c1 c1

Hence, |cn+1 | < |c1 |q n , and ∞

P n
P∞
n=1 |c1 |q < ∞ gives the convergence of n=0 |cn |.
cn+1
If α > 1 we get cn > 1, and hence |cn | < |cn+1 | for all n ∈ N. But this
implies that the necessary condition for convergence, cn → 0 as n → ∞, does
not hold.
p
B: Let α < 1. From the convergence
p of the sequence ( n
|cn |)n∈N we get
for q ∈ (α, 1) a n0 (q) with n
|cn | < q forP∞all n ≥ n0 (q). Hence |cn | <
q n or all n ≥ n0 (q). This implies that q n
converges. Therefore
P∞ Pn=n
∞
0 (q)

n=n0 (q) cn converges absolutely as well as n=0 cn (addition of only finitely

92
many terms). This proves the first (convergence)
p part. If α > 1 we have
a subsequence (cnk )k∈N with limk→∞ nk |cnk | = α such that there is k0 ∈ N
with |cnk | > 1 for all k ≥ k0 . But this contradicts the necessary condition
cn → 0 as n → ∞ for the convergence of the series. Therefore the series
diverges for α > 1. P∞ 2
n
The convergence set of a power series n=0 cn (z −a) is the set of complex
numbers z for which the series converges. This set includes the open disc
|z − a| < R if R is the radius of convergence. It may also contain points of
the boundary, however, this has to check for each example separately.

Example 12.6 (a) For which complex numbers does the series
∞
X ∞
X
n
cn (z − a) = (3 + i)(2i)n (z + i)n
n=0 n=0

converge? Pick z ∈ C and cn = (2i)n (z + i)n . Then

|cn+1 |
= 2|z + i|.
|cn |
If |z + i| < 21 the series converges, and if |z + i| > 12 the series diverges. Hence
the radius of convergence is R = 12 and the series converges for all points in
the open ball B(−i, 21 ) of radius 21 around −i, see figure 54.
Figure 54: ball of convergence (a)

93
 What happens on the boundary of that open ball? For complex numbers
z ∈ C with |z + i| = 12 we get |c|cn+1
n|
|
= 1, and hence (|cn |)n∈N and there-
fore (cn )n∈N does not converge to zero. Thus the series is divergent for any
point on the boundary ∂B(−i, 21 ). The series converges only in the open ball
B(−i, 12 ).

(b) For which complex numbers does the series ∞ zn

P
n=1 n converge?

|cn+1 | n
= |z| → |z| as n → ∞,
|cn | n+1

gives R = 1. We know that on the boundary line ∂B(0, 1) the series converges
for z = −1 but it diverges for z = 1 (harmonic series). The convergence set
is shown in figure 55.
Figure 55: convergence set (b)

What about other boundary points z ∈ ∂B(0, 1) \ {−1, 1} ? This is a

hard problem. 

94
Definition 12.7 For z ∈ C the following power series are defined
∞
X zn
ez := ,
n=0
n!
∞
X (−1)n
cos(z) = z 2n ,
n=0
(2n)!
∞
X 1 2n
cosh(z) := z , (12.28)
n=0
(2n)!
∞
X (−1)n 2n+1
sin(z) := z ,
n=0
(2n + 1)!
∞
X 1
sinh(z) := z (2n+1) .
n=0
(2n + 1)!

Proposition 12.8 The power series on the right hand side of (12.28) have
radius of convergence R = ∞.

Proof. Exercise. The ratio test shows, easily, that all series have infinite
radius of convergence. 2
How do the functions defined in (12.28) behave for z ∈ R? From the
Power Series Theorem 12.4 we know for example that
∞ ∞
d z X z n−1 X z n
e = n = = ez .
dz n=1
n! n=0
n!

The following lemma gives a representation of some functions with the ex-
ponential function.

Lemma 12.9 The following holds for all z ∈ C.

eiz − e−iz
sin(z) = ,
2i
eiz + e−iz
cos(z) = ,
2
ez − e−z
sinh(z) = ,
2
ez + e−z
cosh(z) = .
2

95
Proof. First note that

n n 2in ; if n = 2k is even
i + (−i) =
0 ; if n is odd
Using the power series for the exponential function we get
∞
1 iz −iz

1 X  (iz)n (−iz)n 
e +e = +
2 2 n=0 n! n!
∞
1 X zn n
i + (−i)n


=
2 n=0 n!
∞
X z 2k (−1)k
= = cos(z).
k=0
(2k)!

The remaining relations are proved similar. 2

The following example shows that the complex sine function is unbounded
in contrast to the real sine function.

Example 12.10 The sine function at z ∈ C \ R is

e−y − ey ey − e−y
sin(iy) = =i = i sinh(y), y ∈ R.
2i 2
Choosing y = 10 we get | sin(10i)| = |i sinh(10)| ≈ 22000 in contrast to
| sin(x)| ≤ 1 for x ∈ R. 

Theorem 12.11 (Properties of the exponential function) The exponen-

tial function
∞
z
X zn
exp : C → C, z 7→ exp(z) = e =
n=0
n!
has the properties

(a) e0 = 1,
(b) ez+w = ez ew for all z, w ∈ C,
(c) ez 6= 0 for all z ∈ C.

Proof. (a) Follows from the power series. (b) We first sketch the long
method (Cauchy product rule for series). Put
X z n wm
pl = for all l ∈ N0 .
n,m∈N0 : n+m=l
n! m!

96
Then
∞ n  X∞ ∞ ∞ l   ∞
X z wm  X X 1 X l n l−n X (z + w)l
= pl = z w = .
n=0
n! m=0
m! l=0 l=0
l! n=0
n l=0
l!

Smart method: Fix c ∈ C and consider the function f (z) := ez ec−z , z ∈ C.

The function is differentiable at any z ∈ C with (product rule) f 0 (z) =
ez ec−z − ez ec−z = 0, z ∈ C. Hence (see below Proposition 13.7) there is a
constant K ∈ C such that f (z) = K for all z ∈ C. To find K we put z = c
and obtain K = ec ec−c = ec because of (a). Thus ec = ez ec−z for all z ∈ C.
Putting c = w + z we conclude with (b). (c) Now ez e−z = 1, z ∈ C, gives (c).
2

Periodicity

Lemma 12.12 Let z ∈ C.

(a) ez = 1 ⇔ z = 2kπi for some k ∈ Z.

(b) ez = −1 ⇔ z = (2k + 1)πi for some k ∈ Z.

(c) ez+α = ez for all z ∈ C ⇔ α = 2kπi for some k ∈ Z.

Proof. We prove only (a) as (b) and (c) follow immediately analogously.
Put z = x + iy. Then

ex+iy = ex eiy = ex (cos(y) + i sin(y)) = 1 ⇔ |ex | = 1 and cos(x) + i sin(y) = 1

⇔ x = 0; cos(y) = 1, and sin(y) = 0
⇔ x = 0 and y = 2kπ, k ∈ Z.
2

Argument Any complex number z ∈ C can be written in the polar form

z = |z|eiθ . Let z ∈ C \ {0}, the argument of z is the set

arg(z) := {θ ∈ R : z = |z|eiθ }.

Note that arg(z) is a countably infinite set consisting of all numbers of the
form θ + 2kπ, k ∈ Z.

Example 12.13 Let z, w ∈ C \ {0}.

97
 (a) arg(i) = {(4k + 1) π2 : k ∈ Z}.

(b) arg(zw) = {θ + ϕ : θ ∈ arg(z), ϕ ∈ arg(w)}.

(c) arg( z1 ) = {−θ : θ ∈ arg(z)}.

(d) Let w ∈ C, z ∈ C \ {0}. Put z = ew = eu+iv , u, v ∈ R. Then |z| =

|eu eiv | = eu and arg(z) = {v + 2kπ : k ∈ Z}. Thus

ew = z ⇔ w = log |z| + iθ, θ ∈ arg(z).

Definition 12.14 Let z ∈ C \ {0} and α ∈ C. We use [·]-brackets for a set.

(a) [log z] := {log |z| + iθ : θ ∈ arg(z)}. For w = u + iv one gets

[log ew ] = {log eu + i(v + 2kπ) : k ∈ Z} = {w + 2kπ : k ∈ Z}.

(b) [z α ] := {eα(log |z|+iθ) : θ ∈ arg(z)}.

Example 12.15 (Graphical representation of the n-th roots) Solve z 6 =

−1, that is give the six roots zk , k = 0, 1, . . . , 5, with zk6 = −1. Note that
−1 = eiθ = eiπ , i.e. θ = π. Hence,
 θ + 2kπ
θ + 2kπ

zk = cos + i sin , k = 0, 1, . . . , 5.
6 6
In figure 56 we get the six roots as the corners of the regular six polygon.

98
 Figure 56: the roots of z 6 = −1

13 Holomorphic functions
In this section holomorphic functions will be defined. The starting point is
Example 11.12(b). The complex conjugation is not differentiable at any point
z ∈ C. We want to understand this. Therefore we assume that the function
f : C → C, z 7→ f (z) = u(x, y) + iv(x, y) is differentiable at z = x + iy ∈ C.
We study the following two cases for the limit of the differential quotient.

1.) Let R+ 3 ε > 0 and put h = ε ∈ C, that is h → 0 means ε → 0. Then

u(x + ε, y) − u(x, y) v(x + ε, y) − v(x, y) ∂u ∂v

lim + i lim = (x, y) + i (x, y).
ε→0 ε ε→0 ε ∂x ∂x

2.) Let R+ 3 ε > 0 and put h = εi ∈ C, that is h → 0 means ε → 0. Then

u(x, y + ε) − u(x, y) v(x, y + ε) − v(x, y) ∂u ∂v

lim +i lim = −i (x, y)+ (x, y).
ε→0 iε ε→0 iε ∂y ∂y

99
 The function f is differentiable at z = x + iy and therefore all the limits
of the differential quotients exist and are equal. Thus we get the following
equations

∂u ∂v ∂v ∂u
= and =− . (13.29)
∂x ∂y ∂x ∂y

The two equations (13.29) are called the Cauchy-Riemann equations.

We have thus shown that a function which is differentiable at a point z =
x+iy satisfies the Cauchy-Riemann equations (13.29) at the point (x, y) ∈ R2 .
The converse is true only if the partial derivatives of the real and imaginary
part of the function are continuous in a small neighbourhood of that point.

Theorem 13.1 Let f : D → C, D ⊂ C, with f (x + iy) = u(x, y) + iv(x, y)

for x + iy ∈ D.
(a) If the function f is differentiable at z0 = x0 + iy0 ∈ D, then the partial
derivatives ∂u , ∂u , ∂v and ∂y
∂x ∂y ∂x
∂v
exist at (x0 , y0 ) and satisfy the Cauchy-
Riemann equations (13.29) at the point (x0 , y0 ).
(b) If the partial derivatives ∂u , ∂u , ∂v and ∂y
∂x ∂y ∂x
∂v
exists and are continuous in
an open ball B((x0 , y0 ), ε), ε > 0, around (x0 , y0 ) such that the corre-
sponding open ball in the complex plane is contained in D (B((x0 +
iy0 ), ε) ⊂ D) and the Cauchy-Riemann equations hold at (x0 , y0 ), then
the function f is differentiable at z0 = x0 + iy0 ∈ D with
∂u ∂v ∂v ∂u
f 0 (z0 ) = (x0 , y0 ) + i (x0 , y0 ) = (x0 , y0 ) − i (x0 , y0 ). (13.30)
∂x ∂x ∂y ∂y

Proof. (a) This follows as indicated above. (b) Apply the mean value
theorem of real analysis (exercise). 2

Example 13.2 (a) f : C → C, z 7→ f (z) = z 3 . The real and imaginary parts

are
f (x + iy) = (x + iy)3 = x3 + 3x2 iy + 3x(iy)2 + (iy)3
= (x3 − 3xy 2 ) + i(3x2 y − y 3 ),
with partial derivatives
∂u ∂v
= 3x2 − 3y 2 ; = 3x2 − 3y 2
∂x ∂y
∂u ∂v
= −6xy; = 6xy.
∂y ∂x

100
The Cauchy-Riemann equations are satisfied for all (x, y) ∈ R2 and the
partial derivatives are continuous on R2 , hence the function f is differentiable
on C.

(b) f : C → C, z 7→ f (z) = f (x + iy) = x2 + iy 2 . The partial derivatives of

the real and imaginary part are
∂u ∂v
= 2x; = 2y
∂x ∂y
∂u ∂v
= 0; = 0,
∂y ∂x

they are continuous on R2 but the Cauchy-Riemann equations are only

satisfied on {(x, y) ∈ R2 : x = y}. The function f is differentiable on
{z ∈ C : z = x + ix, x ∈ R}. 

The condition in Theorem 13.1 that there is a small ball in the domain of
definition is crucial. In the following we will assume that the domain for any
function has the property that one can find around any point of that domain
a small ball contained in the domain. Such sets have a special name.

Definition 13.3 (Holomorphic function) (a) Let D ⊂ C. The set D

is called open if for any z ∈ D there is an ε > 0 such that B(z, ε) ⊂ D.

(b) A function f : D → C, D ⊂ C open, is called holomorphic at z0 ∈ D

if f is differentiable at all z ∈ B(z0 , ε) ⊂ D for some ε > 0. The
function f is called holomorphic in D if f is holomorphic at all z ∈ D.

Remark 13.4 Differentiability in real analysis is defined at a single point.

Being holomorphic is not a property of a single point, it depends on a whole
neighbourhood of a single point in the complex plane. This is the main dif-
ference, and the notion of holomorphic functions will prove to be the most
useful one.

Lemma 13.5 Let f, g : D → C, D ⊂ C open, be holomorphic in D. Then

the following holds.

(a) f + g, λf and f g are holomorphic in D, where λ ∈ C.

1
(b) Suppose f (z) 6= 0 for all z ∈ D. Then f
is holomorphic in D with
 1 0 −f 0 (z)
(z) = , z ∈ D.
f f 2 (z)

101
 f
(c) Suppose g(z) 6= 0 for all z ∈ D. Then g
is holomorphic in D with
 f 0 f 0 (z)g(z) − f (z)g 0 (z)
(z) = , z ∈ D.
g g 2 (z)

Proof. Follows analogous to the proofs for real analysis. Exercise. 2

Example 13.6 (a) Let f : C → C be non-constant and holomorphic in C.

Then u = Re(f ) is not holomorphic in C. Assume that u is holomorphic.
We conclude from the Cauchy-Riemann equations that ∂u ∂x
(x, y) = 0 and
∂u 2
∂y
(x, y) = 0 holds for all (x, y) ∈ R . This implies that u is constant in
contrast to f non-constant. Hence, the assumption was wrong and u is not
holomorphic.

(b) Let f : C → C be non-constant and holomorphic in C. Then |f | is

nowhere holomorphic. Denote the real part by u e = Re(|f |) = |f | and the
imaginary part ve = 0. Assume that |f | is holomorphic. Then the Cauchy-
Riemann equations imply that (u, v are the real and imaginary part of f
respectively)
1

e=0= √
∂x u u∂x u + v∂x v
u2 + v 2
1

e=0= √
∂y u u∂y u + v∂y v .
u2 + v 2
This gives
u∂x u + v∂x v = 0 = u∂y u + v∂y v,
and with the Cauchy-Riemann equations for the function f itself we get,
summing the left and right hand side,

0 = u(∂x u + ∂y u) + v(∂x u − ∂y u).

This equation can be satisfied only by u constant which contradicts our

assumption that f is non-constant or by the condition that the terms in the
brackets vanish (other choices like u = −v are not possible). The terms in
the brackets vanish if ∂x u + ∂y u = 0 and ∂x u = ∂y u. This implies 2∂y u = 0
and ∂x u = 0, and thus that u is constant again in contradiction to our
assumption. Thus |f | is nowhere holomorphic.


Proposition 13.7 Let f : D → C, D ⊂ C open, be holomorphic in D. Then

any of the following conditions forces the function f to be constant in D.

102
 (a) f 0 (z) = 0 for all z ∈ D.

(b) |f | is constant in D.

(c) f (z) ∈ R for all z ∈ D.

Our aim is to apply the theorems of vector analysis for the study of
complex valued functions on C. Let f : C → C be a holomorphic function on
C with real and imaginary part u respectively v. Define the two-dimensional
vector field
   
2 2 F1 (x, y) u(x, y)
F : R → R , (x, y) 7→ F (x, y) = = .
F2 (x, y) −v(x, y)

Then
∂u ∂v
div F = − =0
∂x ∂y
F2 ∂F1
curl F = −
∂x ∂y
∂v ∂u
=− − = 0.
∂x ∂y
Hence, the theorems of Gauss and Green are useful here. The missing piece is
the notion of integration for complex valued functions defined on the complex
plane. This is the content of the next section.

14 Complex integration
The complex line integral will be defined, the Fundamental Theorem of Cal-
culus will be proved and examples of non-vanishing integrals along closed
paths are studied. A curve (path) C ⊂ C is a set points in the complex plane
for which one can find a one-dimensional parametrisation (i.e. a parametri-
sation that depends only on one real variable). Let a parametrisation for a
curve C in the complex plane C be given, i.e.

γ : [a, b] → C, t 7→ γ(t) = x(t) + iy(t),

where a ≤ b and where x(t) is the real part Re(γ(t)) and y(t) is the imaginary
part Im(γ(t)) of the curve. Recall that γ is a parametrisation of the curve
C if γ is piecewise C 1 and γ([a, b]) = C. The derivative with respect to the
real parameter t is γ 0 (t) = x0 (t) + iy 0 (t), t ∈ [a, b].

103
Definition 14.1 The line (path) integral of a function f : D → C, D ⊂ C,
along the curve (path) C ⊂ D ⊂ C with parametrisation γ : [a, b] → C is
defined as
Z Z b
f= f (γ(t))γ 0 (t) dt
γ a
Z b Z b
(14.31)
0 0
= Re(f (γ(t))γ (t)) dt + i Im(f (γ(t))γ (t)) dt
a a
R R
Alternative notation: C
f and γ
f (z) dz.

Example 14.2 Let the function f : C → C, z 7→ f (z) = f (x + iy) = x2 + iy 2

and the parametrised path γ : [0, 1] → C, t 7→ γ(t) = t(1 + i) be given. γ
parametrises the straight line from the origin to 1 + i, see figure 57.
Figure 57: path from the origin to 1 + i

As γ 0 (t) = 1 + i for all t ∈ [0, 1] we get the path integral

Z Z 1 Z 1
2 2 2
f= (t + it )(1 + i) dt = 2i t2 dt = i.
γ 0 0 3


The following properties are easily proved (compare part I of the lecture).

104
Lemma 14.3 Let C ⊂ C be a curve in the complex plane with parametri-
sation γ : [a, b] → C (i.e., γ([a, b]) = C) and let f : C → C be a continuous
function.
(a) If γ − : [a, b] → C, t 7→ γ − (t) = γ(a + b − t) is the reversed parametrisa-
tion, then Z Z
f =− f.
γ− γ

(b) Let γ a, eb] → C and suppose that γ

e : [e e = γ ◦ ψ with γ a, eb]) = C and
e([e
that the map ψ : [ea, b] → [a, b] is bijective and has a positive continuous
e
derivative. Then Z Z
f = f.
γ
e γ

In the next example we integrate the complex conjugate along a closed

path. Recall from Example 11.12 that the complex conjugate is not differ-
entiable but continuous.
Example 14.4 Let C = ∂B(i, 2) be the circle line around i with radius 2,
see figure 58.
Figure 58: circle line ∂B(i, 2)

The line integral for the complex conjugate, i.e., for the function f : C →
C, z 7→ f (z) = z, along C is computed as follows. A parametrisation for

105
the circle line is given by γ(t) = i + 2eit for t ∈ [0, 2π] with derivative
γ 0 (t) = 2ieit , t ∈ [0, 2π]. Hence,
Z Z 2π Z 2π
− i + 2e−it 2ieit dt

it

f= it
i + 2e 2ie dt =
γ 0 0
Z 2π it 2π
e
2eit + 4i dt = 2


= + 8πi = 8πi.
0 i t=0

It remains to check that the first term vanishes, that is

Z 2π Z 2π Z 2π
it
e dt = cos(t) dt + i sin(t) dt
0 0 0
2π 2π h i sin(t) + cos(t) i2π
= sin(t) − i cos(t) =
0 0 i 0
eit 2π
= = 0.
i 0


Why does the integral in Example 14.4 does not vanish, although we
have a closed path? We will come back to this later. We now study the most
important integral in complex analysis, the so-called Fundamental Integral.

Example 14.5 (Fundamental Integral) Let a ∈ C and r > 0 and con-

sider the circle line ∂B(a, r) around a with radius r and parametrisation
γ(t) = a + reit , t ∈ [0, 2π]. Then
Z 
n 0 ; n 6= −1
(z − a) = (14.32)
∂B(a,r) 2πi ; n = −1

This can be seen as follows

Z 2π Z 2π
it n it n+1
ei(n+1)t dt


re rie dt = ir
0  0
h i2π
n+1 sin((n+1)t) cos((n+1)t)


 ir n+1
− i n+1
; n 6= −1
0
= h i2π


 i 1 ; n = −1
0

We do have examples of line integrals along closed paths which do not

vanish. The integrand functions of these examples are not holomorphic at

106
points on the curve and/or at points surrounded by the curve. However,
example 14.5 shows that there are cases when the line integral along the
closed circline vanishes although the function is not holomorphic at points
inside the circle line. We therefore expect to have vanishing line integrals
along closed paths as long the integrand functions are holomorphic. The
next theorem is the first step towards the so-called Cauchy-Theorem which
proves this conjecture.

Theorem 14.6 (Fundamental Theorem of calculus in C) Let the func-

tion f : D → C, D ⊂ C, be holomorphic in D with derivative f 0 and C ⊂ D
be a curve (path) with parametrisation γ : [a, b] → D. Then
Z
f 0 = f (γ(b)) − f (γ(a)) (14.33)
γ

Proof.
Z Z b Z b
df df d
= (γ(t))γ 0 (t) dt = (f (γ(t))) dt
γ dz a dz a dt
Z b Z b
d d
= Re(f (γ(t))) dt + i Im(f (γ(t))) dt
a dt a dt
= f (γ(b)) − f (γ(a)),

where we used the chain rule and the FTC for the real functions Re(f (γ(t)))
and Im(f (γ(t))). 2

15 Cauchy’s theorem
In this section the Cauchy Theorem is proved and analysed. The proof
involves to specify an admissible class of paths/curves. Before that we shall
discuss the connection to Green’s and Gauss’s theorem. Let γ : [a, b] →
C, t 7→ γ(t) = x(t) + iy(t) be a parametrisation of a curve (path) C ⊂ C and

107
let a continuous function f : C → C, f = u + iv, be given. We compute
Z Z b
f= (u(γ(t)) + iv(γ(t))(x0 (t) + iy 0 (t)) dt
γ a
Z b b Z
0 0
u(γ(t))y 0 (t) + v(γ(t))x0 (t) dt



= u(γ(t))x (t) − v(γ(t))y (t) dt + i
a a
Z bD   0 E Z bD   0 
u(γ(t)) x (t) u(γ(t)) y (t) E
= , 0 dt + , dt
a −v(γ(t)) y (t) a −v(γ(t)) −x0 (t)
Z Z
= hF, T i + i hF, N
b b i,
γ γ
 0   0 
0 x (t) y (t)
where T = γ (t) =
b 0 and N =
b 0 and F : R2 → R2 , (x, y) 7→
 y
 (t) −x (t)
u(x, y)
F (x, y) = . Note that N
b is orthogonal to Tb. At the end of
−v(x, y)
Section 13 we showed that div F = curl F = 0. Hence, we can easily derive
Cauchy’s theorem from Green’s theorem and Gauss’ s theorem. However, it
will turn out that the topological properties of the curves (paths) are linked
to the calculus of complex valued function defined on domains in C. We only
indicate this issue briefly and defer a detailed study of this relationship to
later courses. Hence, we will not define properly what deformation in the
following definition means. A deformation of a curve (path) is a continuous
transformation in the complex plane.

Definition 15.1 (a) A subset D ⊂ C is connected if it cannot be ex-

pressed as the union of non-empty open sets D1 ⊂ C and D2 ⊂ C with
D1 ∩ D2 = ∅.

(b) A region is a non-empty open connected subset of the complex plane

C.

(c) A region D ⊂ C is called simply connected if every closed curve

(path) in D can be deformed to a single point.

Theorem 15.2 (Cauchy’s theorem) Let the function f : D → C, D ⊂ C

be a simply connected region, be holomorphic on D and C ⊂ D a closed curve
(path) in D with parametrisation γ. Then
Z
f = 0. (15.34)
γ

108
Proof. We sketch the proof ignoring the dependence on the topological
details of the curve C ⊂ D. By Ω we denote the region surrounded by the
curve (path), that is the boundary ∂Ω = C. See figure 59.
Figure 59: region D with Ω ∪ ∂Ω

From the previous calculation we get

Z Z Z
f = hF, T i + i hF, N
b bi
γ
Zγ γ
Z
= curl (F ) + i div (F ) = 0 + 0,
∂Ω Ω

where the last equality follows from Green’s and Gauss’s theorem respectively
using the fact that γ is a closed path. 2

Remark 15.3 The Theorem also holds for more general curves, see figure
60,

109
 Figure 60: γ union of γ1 ∪ γ2

R R R
where the integral γ f = γ1 f + γ2 f = 0 vanishes because Green’s and
Gauss’s theorem can be applied to Ω1 and Ω2 respectively. But the region
the curve (path) is encircling has to be simply connected, i.e., it must be
possible to contract it to a single point. This contraction is not possible in
the example of an annulus in figure 61,

110
 Figure 61: annulus with ∂Ω ⊂ D and Ω 6⊂ D

where γ = ∂Ω but Ω 6⊂ D.

For later purposes we consider the following classes of curves (paths).

Recall the definition of a simple path: a ≤ s < t ≤ b implies that γ(s) 6= γ(t)
unless s = t or s = a and t = b if γ is closed.

Notation 15.4 (a) Let u, v ∈ C. The image C of the parametrisation γ

given by γ : [0, 1] → C, t 7→ γ(t) = (1 − t)u + tv is the line segment
[u, v] traced from u to v.

(b) A circular arc traced (anti-clockwise/clockwise) is the image of the

parametrisation γ : [Θ1 , Θ2 ] → t 7→ γ(t) = a + reit , a ∈ C and r > 0
with Θ1 ≤ Θ2 . The circline ∂B(a, r) around a ∈ C with radius r > 0
is parametrised by γ : [0, 2π] → C, t 7→ γ(t) = a + reit .

(c) A circline path is a path which is the join of finitely many paths of
type (a) and (b), and a contour is a simple closed circline path.

(d) Given a closed curve (path) C ⊂ C with parametrisation γ : [a, b] → C

and the image by γ ∗ = γ([a, b]) = C, the inside I(γ) is the set of points
surrounded by the curve (path) and O(γ) is the complement of I(γ)∪γ ∗ ,
see figure 62.

111
 (e) The curve (path) parametrised by γ : [0, 2π] → C, t 7→ γ(t) = a + reit
is called the positively oriented circle with centre a and radius r,
and the curve (path) with parametrisation γ : [0, 2π] → C, t 7→ γ(t) =
a + re−it is called the negatively oriented circle. Positively oriented
means anti-clockwise in the following.

Figure 62: I(γ) and O(γ)

Example 15.5 For any 0 < R < π

Z
1
dz = 0.
∂B(0,R) ez +1

To see this note that for z = x + iy ∈ C

ez + 1 = 0 ⇔ ez = −1 ⇔ x = 0, y = (2n + 1)π for n ∈ Z.

Hence the function f : C → C, z 7→ f (z) = ez1+1 is holomorphic on and inside

any circle with radius R < π, see figure 63.

112
 Figure 63: ∂B(0, R)

In the next example we analyse again the complex conjugation as in

Example 14.4.

Example 15.6 In Example 14.4 it is shown that

Z
z = 8πi,
γ

where γ is the parametrisation for ∂B(i, 2), a circle line of radius 2 around
i. As 4π is the area of that circle one can guess that the following holds.
Let γ be the parametrisation of any closed curve (path) C. Then
Z
z = 2i(area(I(γ))), (15.35)
γ

where I(γ) is the region enclosed by the curve (path). This can be seen as
follows. Recall that for any a, b ∈ C one can prove that Im(ab) = 2(area of
the triangle spanned by a and b).

113
 Figure 64: z, z + ∆

Consider in figure 64 the triangle spanned by z (vector from the origin to

a point z on γ ∗ ) and the vector z + ∆, where ∆ is a small tangential line
segment attached to the point z on γ ∗ . Then the area of the triangle spanned
by z and z + ∆ is given by 21 Im((z + ∆)z) = 12 Im(∆z). The line integral
is approximated by the sum of all these small contributions, that is, the
integral is obtained as a limit ∆ → 0. One might be inclined to think that
the integral of the complex conjugate z never vanishes for a non-trivial closed
curve (path). But this is not true as we see for the curve in figure 65.

114
 Figure 65: γ = γ1 ∪ γ2

Here γ is the union of γ1 and γ2 . Now γ2 is negatively oriented, that is

the integral along γ2 gets a minus sign. As the two areas of the surrounded
regions of γ1 and γ2 are equal, the two contributions cancel out and the
integral along γ vanishes. 

One can generalise our fundamental integral in Example 14.5 to arbitrary

positively oriented contours. Let γ be a parametrisation of positively oriented
contour with z0 6∈ γ ∗ . Then

Z  0 ; if z0 ∈ O(γ)
(z − z0 )n = 0 ; if z0 ∈ I(γ) and n 6= −1 (15.36)
γ 
2πi ; if z0 ∈ I(γ) and n = −1

The proof follows with the following theorem which we cite here without
proof.

Theorem 15.7 (Deformation) (a) Let γ be a positively oriented contour

and suppose that B(a, r) ⊂ I(γ), a ∈ C, r > 0. Let f be holomorphic
inside (i.e. on I(γ)) and on γ except possibly at a ∈ C. Then

115
 Z Z
f= f.
γ ∂B(a,r)

(b) Suppose that γ, γ

e are positively oriented contours such that γ
e lies inside
∗
e ⊂ I(γ). If f is holomorphic inside and on γ
γ, i.e., γ

Z Z
f= f.
γ γ
e

R
Example 15.8 Integrate ∂B(0,2) f , where f : C → C, z 7→ f (z) = 2(4z 2 −
1)−1 . The function has two singularities inside the circline ∂B(0, 2). We
write
1 1
f (z) = − , z ∈ C.
2z − 1 2z + 1
The Deformation Theorem 15.7 implies that
Z Z Z
1 1
f= −
∂B(0,2) 2z − 1 2z + 1
Z∂B(0,2) Z∂B(0,2)
1 1
= −
∂B( 12 , 14 ) 2z − 1 ∂B(− 12 , 14 ) 2z + 1
1 1
= 2πi − 2πi = 0,
2 2
where we used the circlines ∂B( 12 , 14 ) and ∂B(− 12 , 14 ) such that we avoid one
of the singularities respectively, see figure 66 (other circline (paths) would do
the same).

116
 Figure 66: B( 12 , 41 ) and B(− 21 , 14 )

16 Cauchy’s formulae
Armed with Cauchy’s theorem we can prove a host of striking results about
holomorphic functions. In Subsection 16.2 we prove Liouville’s theorem,
differentiability of any order, and Taylor’s theorem.

16.1 Cauchy’s formulae

Cauchy’s integral formula expresses the value of a holomorphic function at a
point in terms of a ’boundary value integral’ taken over a contour encircling
the point. The ingredients at hand are

• The Deformation Theorem 15.7

• Z
1
dw = 2πi
∂B(a,r) w−a
• f (w) − f (a) → 0 as w → a

117
Theorem 16.1 (Cauchy’s integral formula) Let the function f be holo-
morphic inside and on a positively oriented contour with parametrisation
γ : [a, b] → C. Then
Z
1 f (w)
f (z) = dw for z ∈ I(γ). (16.37)
2πi γ w−z

Proof. Pick z ∈ I(γ). Then there is an ε > 0 with B(z, ε) ⊂ I(γ). Use the
Deformation Theorem 15.7, add and subtract an integrand with numerator
f (z) to obtain
Z Z
f (w) f (w)
dw = dw
γ w−z ∂B(z,ε) w − z
f (w) − f (z)
Z Z
f (z)
= dw + dw
∂B(z,ε) w − z ∂B(z,ε) w−z
=: I + II.

The first term I is the fundamental integral

Z
1
I = f (z) dw = 2πif (z).
∂B(z,ε) w − z

To conclude we need to show that the second term II vanishes for ε ↓ 0.

Z 2π
f (z + εeit ) − f (z) it
|II| = εie dt
0 εeit
≤ 2π sup f (z + εeit ) − f (z) → 0 as ε ↓ 0,
t∈[0,2π]

because the function f is continuous at z. Now the expression on the left

hand side is independent of the radius (Deformation Theorem, the first step
above) and so it must be zero. 2

Remark 16.2 Holomorphic function are special: the values of the function
at the boundary of a closed curve γ determine the value at z ∈ I(γ).

Can we also get integral formulae for the derivatives? Let the function f
be holomorphic inside and on a positively oriented contour parametrised by
γ. Cauchy’s formula (16.37) implies
Z
1 f (w)
f (z) = dw for z ∈ I(γ),
2πi γ w − z

118
so we may differentiate the expression on the right hand side with respect to
z neglecting the integral. We obtain
Z
0 1 f (w)
f (z) = dw for z ∈ I(γ),
2πi γ (w − z)2

if we can prove that integration and differentiation can be interchanged. We

may proceed further for the second derivative (and any higher one as well)
but there is one difficulty. The function f is holomorphic which implies so
far only that f 0 exists. But it is not clear if higher derivatives exist as well. A
way out of this is to apply Cauchy’s integral formula (16.37) to the differential
quotient
f (z + h) − f (z)
.
h
Iteration of that process gives the existence of the derivatives f (n) (z) for
n = 2, 3, . . ., and z ∈ I(γ); all given by differentiation under the integral
sign. This is the content of the following theorem.

Theorem 16.3 (Cauchy’s formulae for derivatives) Let the function f

be holomorphic inside and on a positively oriented contour parametrised by
γ and let z ∈ I(γ). Then f (n) (z) exists for n = 1, 2, 3, . . ., and the derivative
is given as
Z
(n)
n! f (w)
f (z) = dw for z ∈ I(γ). (16.38)
2πi γ (w − z)n+1

Proof. Pick ∈ I(γ). Then there is an ε > 0 such that B(z, 2ε) ⊂ I(γ).
Using formula (16.37) for the differential quotient (for h small enough)

f (z + h) − f (z)
Z
1  1 1 
= f (w) − dw
h 2hπi ∂B(z,2ε) w−z−h w−z
Z
1 f (w)
= dw
2πi ∂B(z,2ε) (w − z − h)(w − z)
Z
1 f (w)
= dw
2πi ∂B(z,2ε) (w − z)2
Z
1  1 1 
+ f (w) − dw
2πi ∂B(z,2ε) (w − z − h)(w − z) (w − z)2
Z Z
1 f (w) 1  hf (w) 
= dw + dw.
2πi ∂B(z,2ε) (w − z)2 2πi ∂B(z,2ε) (w − z − h)(w − z)2

119
We have to show that the second term on the right hand side vanishes as
h → 0. For that choose h so small that |h| < ε and conclude for all w ∈
∂B(z, 2ε) that |w − z − h| ≥ |w − z| − |h| > ε. As f is continuous there is
a M > 0 with |f (w)| ≤ M for w ∈ ∂B(z, 2ε). Hence the second term on
the right hand side is bounded by |h|M 2ε
ε(2ε)2
which converges to zero as h → 0.
The higher derivatives are proved via induction using formula (16.37) for the
difference f (k+1) (z + h) − f (k) (z). 2

The last theorem guarantees that a holomorphic function is infinitely often

differentiable. As in real analysis one is interested to represent such a function
as a power series. If we can interchange differentiation and integration we
get an integral formula for the coefficients of that power series. This will be
the content of the next theorem in Subsection 16.2.

16.2 Taylor’s and Liouville’s theorem

Theorem 16.4 (Taylor’s theorem) Let the function f be holomorphic on
B(a, R), a ∈ C, R > 0. Then there exist unique constants cn ∈ C, n ∈ N0 ,
such that ∞
X
f (z) = cn (z − a)n for z ∈ B(a, R), (16.39)
n=0

and the coefficients are given by

Z
1 f (w) f (n) (a)
cn = dw = , (16.40)
2πi γ (w − a)n+1 n!

where γ is the parametrisation of a circle line ∂B(a, r), 0 < r < R, or of any
positively oriented contour in B(a, R) enclosing the point a.

Proof. Pick z ∈ B(a, R) and choose r > 0 such that |z − a| < r < R and
let γ be the parametrisation of the circline ∂B(a, r). The formula (16.37)
gives Z
1 f (w)
f (z) = dw. (16.41)
2πi ∂B(a,r) (w − z)
Because of |z−a| < |w−a| = r for all w ∈ ∂B(a, r) we have |z−a|/|w−a| < 1
and can expand (in a geometric series)
1 1 1
=  .
w−z w − a 1 − z−a
w−a

120
Inserting this into the integrand in (16.41) and interchanging summation
and integration (to be justified later in Theorem 18.7) we conclude with
the existence of the coefficients and their representation by
P∞(16.40). The
uniqueness follows from the following. Assume that f (z) = n=0 cn (z − a)n
for all z ∈ B(a, r) for some 0 < r < R. Provided summation and integration
can be interchanged we have for any n ∈ N0
Z Z ∞
f (z) X 
dz = ck (z − a)k (z − a)−n−1 dz
∂B(a,r) (z − a)n+1 ∂B(a,r) k=0
∞
X Z
= ck (z − a)k−n−1 dz = 2πicn
k=0 ∂B(a,r)

because of the fundamental integral in example 14.5. 2

The following theorem is a striking result from Cauchy’s formula.

Theorem 16.5 (Liouville’s theorem) Let the function f be holomorphic

and bounded on C. Then f is a constant function.

Proof. Suppose |f (w)| ≤ M for all w ∈ C. Pick a, b ∈ C and choose

R > 2 max{|a|, |b|}. Then |w − a| ≥ 21 R and |w − b| ≥ 12 R whenever |w| = R,
i.e., whenever w ∈ ∂B(0, R). Cauchy’s formula (16.37) gives
Z
1  1 1 
f (a) − f (b) = f (w) − dw
2πi ∂B(0,R) w−a w−b
a − b Z f (w)
= dw.
2πi ∂B(0,R) (w − a)(w − b)

Hence
1 |a − b|2πRM
|f (a) − f (b)| ≤ → 0 as R → ∞,
2π ( 12 R)2
implying f (a) = f (b). 2

The following example shows how Cauchy’s formulae can be used to evaluate
integrals.

Example 16.6 (a)

Z  ew  2πi  d2 z 
dw = e = πi.
∂B(0,1) w3 2! dz 2 z=0

121
(b)

2πi  d3  1 
Z
1 2πi
dw = =− .
∂B(−1,3) (w − 4)(w + 1)4 3! dz 3 z − 4 z=−1 54

(c) Let γ be a parametrisation of a positively oriented contour.


Z  0 , if i ∈ O(γ)
sin(w)  
dw = 2πi sin(i) , if i ∈ I(γ) ,
w−i
γ ? , if i ∈ γ ∗


The problem when singularities are on the curve is studied in the next section.


17 Real Integrals
This section will demonstrate the usefulness of Cauchy’s Theorem for evalu-
ating real integrals. We focus on the following example. What is the integral
Z ∞
sin(x) 
dx ?
−∞ x

Real analysis tells us that limx→0 sin(x)

x
= 1, i.e., the integrand is well-defined
for all x ∈ R. In the following we prove that
Z ∞  sin(x) 
dx = π. (17.42)
−∞ x

Proof of (17.42). The idea is to consider the complex function

eiz
f : C → C, z 7→ f (z) = ,
z
which is holomorphic in the punctuated complex plane C \ {0}.
Pick some R > 0, and choose ε > 0 with ε < R. Consider the closed curve
(path) γ in figure 67. It is a union of the curves (paths) γ1 , γ2 , γ3 and γ4 .

122
 Figure 67: path γ

γ1 is the half-circle of radius R > 0 from (R, 0) to (−R, 0), γ2 is the straight
line from (−R, 0) to (−ε, 0) and γ4 is the straight line from (ε, 0) to (R, 0).
To avoid the singularity at the origin the path γ3 is the small half-circle of
radius ε from (−ε, 0) to (ε, 0). This path is negatively oriented (clockwise).
iz
The function f (z) = ez is inside γ and on γ ∗ holomorphic. Hence, Cauchy’s
theorem gives Z Z
f= f = 0. (17.43)
γ ∪4i=1 γi

The contributions from the segments on the real axis are evaluated in the
limit ε → 0.
Z Z Z −ε ix Z R ix
e e
f+ f= dx + dx
γ2 γ4 −R x ε x
Z −ε Z R
cos(x) + i sin(x) cos(x) + i sin(x)
= dx + dx
−R x ε x
Z −ε Z R
i sin(x) i sin(x)
= dx + dx
−R x ε x
Z R
sin(x)
−→ i dx as ε → 0.
−R x

123
From (17.43) we will conclude
Z Z  Z R Z Z
sin(x) 
lim f+ f =i dx = − lim f+ f . (17.44)
ε→0 γ2 γ4 −R x ε→0 γ3 γ1

To prove (17.42) we divide (17.44) by i and take the limit R → ∞. For that
we compute the integrals on the right hand side of (17.44).

(1) Adding and subtracting an integrand we get for the first integral on the
right hand side of (17.44)
Z iw Z iw
e −1 e −1
Z Z Z
1 1 1
f= dw + dw = − dw + dw
γ3 γ3 w γ3 w 2 ∂B(0,ε) w γ3 w
Z iw
e −1
= −iπ + dw =: −iπ + II,
γ3 w

where the minus sign is due to the negative orientation of γ3 and the 12 is
there because we only have the half-circle. It remains to show that the second
term II on the right hand side of the last equation vanishes as ε → 0.

eiw − 1
|II| ≤ hlength(γ3 )i max ≤ M πε → 0 as ε → 0,
w∈∂B(0,ε) w
iw
because e w−1 is continuous on ∂B(0, ε) (attains its maximum value because
the circline is closed and bounded). Hence,
Z
lim f = −iπ. (17.45)
ε→0 γ3

Thus we need to show in (2) that the second integral on the right hand side
of (17.44) vanishes in the limit R → ∞. The other term does not depend on
R.

(2) Let γ1 : [0, π] → C, t 7→ γ1 (t) = Reit be the parametrisation for the

half-circle. We estimate
Z iz Z π iR(cos(t)+i sin(t)) Z π
e e it
dz = it
iRe dt ≤ eiR(cos(t)+i sin(t)) dt
γ1 z 0 Re 0
Z π Z π/2 Z π/2
2tR
−R sin(t) −R sin(t)
= e dt = 2 e dt ≤ 2 e− π dt
0 0 0
h π − 2Rt it=π/2 π
1 − e−R ≤ π/R → 0 as R → ∞.


=2 − e π =
2R t=0 R

124
 t 2t
The inequality follows from the estimation sin(t) ≥ π/2 = π
for all t ∈
[0, π/2], see figure 68. A proof can be found in Lemma 17.1.
Figure 68: estimation sin(t) on [0, π/2]

2t
This implies e−R sin(t) ≤ e−R π for all t ∈ [0, π/2]. We conclude from (17.44)
and (17.45)
Z R Z Z
sin(t) 
lim i dx = − lim lim f+ f = iπ
R→∞ −R x R→∞ ε→0 γ3 γ1

to get (17.42). 2
Lemma 17.1 (Jordan’s inequality)
2 sin(x) π
≤ ≤1 for 0 < x ≤ .
π x 2
Proof. It suffices to show that the function sin(x)
x
decreases for x ∈ (0, π2 ],
i.e., to show that
d  sin(x)  x cos(x) − sin(x) π
= 2
≤ 0, x ∈ (0, ].
dx x x 2
This follows easily from [x cos(x) − sin(x)]|x=0 = 0 and from
d

x cos(x) − sin(x) = −x sin(x) ≤ 0.
dx
2

125
18 Power series for Holomorphic functions
This section completes our study of power series and its proofs are based on
Cauchy’s theorem and Cauchy’s formulae.

18.1 Power series representation

Theorem 18.1 Let the power series ∞ n
P
n=0 cn z have radius of convergence
R > 0 and define on the open ball B(0, R) the function f : B(0, R) → C by
∞
X
f (z) := cn z n , z ∈ B(0, R).
n=0

Then the following statements are true:

P∞ n−1
(a) n=1 ncn z has radius of convergence R.
P∞
(b) f is holomorphic on B(0, R), and f 0 (z) = n=1 ncn z n−1 for all z ∈
B(0, R).
(c) f has derivatives of all orders n ∈ N in the open ball B(0, R) and
f (n) (0) = n!cn , n ∈ N0 .

Proof. We only sketch the P ideas, parts of the proof follow analogously to
the real case. Assume that ∞ c
n=0 n z n
converges for |z| < R. Pick 0 < ρ < R
|z|
n
and let |z| < ρ and assume z 6= 0. Now ρ < 1 implies that ∞
|z| P
n=0 n ρ
converges (Ratio test for example). Hence there is a M > 0 such that
|z|
n
n ≤M for all n ∈ N0 .
ρ
Thus
M
|ncn z n−1 | ≤ |cn ρn | for all n ∈ N0 ,
|z|
gives (a) because the series ∞ n
P
n=0 |cn ρ | converges. To conclude with (b) we
have to show that
∞ 
f (z + h) − f (z) X (z + h)n − z n 
(∗) := − g(z) = − nz n−1 → 0 as h → 0,
h n=1
h

where g(z) := ∞ n−1

P
n=1 ncn z . We shall use the binomial expansion
n  
n
X n n−k k
(z + h) = z k , n ∈ N, z, h ∈ C.
k=0
k

126
For z, z + h ∈ B(0, R) we must prove that
n  
(z + h)n − z n n−1
X n k−2 n−k
− nz =h h z
h k=2
k
n−2
X n!
=h hr z n−2−r
r=0
(n − (r + 2))!(r + 2)!

(with r = k − 2) vanishes in the limit h → 0. With the infinite version of the

triangle inequality the estimation and again with the binomial expansion the
following estimate
∞ n−2
X  X n! 
|(∗)| = cn h hr z n−2−r
n=1 r=0
(n − r − 2)!(r + 2)!
∞ n−2
X X (n − 2)! 
≤ |h| n(n − 1)|cn | |h|r |z|n−2−r
n=1 r=0
(n − 2 − r)!r!
∞
X
= |h| n(n − 1)|cn |(|z| + |h|)n−2 .
n=1

Fix z and pick ρ with |z| < P ρ < R, so that |z| + |h| < ρ whenever |h| <
ρ − |z|. Using part (a) twice, ∞ n=2 n(n − 1)|cn |ρ
n−2
converges to a constant
0
independent of h. We conclude that f (z) does exist and equals g(z). The
proof of (c) follows wit (16.38). 2
P∞
Example 18.2 We know that the geometric series n=0 z n has radius of
convergence R = 1. Theorem 18.1 implies that
∞
−2 d X
(1 − z) = (1 − z)−1 = nz n−1 , |z| < 1.
dz n=1

Via induction one can easily see that

1 X n
= z n−k , |z| < 1, k ∈ N.
(1 − z)k+1 n≥k k

Definition 18.3 A function f is called analytic if it can be expanded into

a power series around every point.

127
Theorem 18.1 gives

f holomorphic ⇔ f analytic.

This does not hold in R:

f once differentiable on R 6⇒ f twice differentiable on R
f infinitely many times differentiable on R 6⇒ f has a power series expansion
This can be seen in the following example.
Example 18.4 The real-valued function

0 ; if x ≤ 0
f : R → R, x 7→ f (x) = − x1
e ; if x > 0
dk f
is infinitely many times differentiable, f ∈ C ∞ , and dxk
(0) = 0 for all k ∈ N0 .
But the corresponding power series
∞  k
X d f  xk
k
(0) =0
k=0
x k!

in contrast to f (x) 6= 0 for x > 0. 

18.2 Power series representation - further results

Theorem 18.5 (Uniqueness Theorem) Let the functions f, g be holomor-
phic on B(a, R), a ∈ C, R > 0, and (zn )n∈N be a sequence in B(a, R) \ {a}
with zn → a as n → ∞ and f (zn ) = g(zn ) for all n ∈ N. Then f = g on
B(a, R).

Proof. The function h = f − g is holomorphic on B(a, R). Thus

h(z) = c0 + c1 (z − a) + c2 (z − a)2 + · · ·
for z ∈ B(a, R). From continuity and the assumptions we get h(a) =
limn→∞ h(zn ) = 0 which implies that the coefficient c0 = 0. The limit of
h(zn )
= c1 + c2 (zn − a) + · · ·
(zn − a)
exists (h is holomorphic) and equals c1 = limn→∞ h(z n)
zn −a
= 0. Induction gives
cn = 0 for all n ∈ N0 , and from the power series representation of h, Theo-
rem 18.1, we conclude h = 0 on B(a, R), i.e., f = g on B(a, R). 2
The next examples show the power of this theorem.

128
Example 18.6 (a) Let f (z) = sin2 (z) + cos2 (z) and g(z) = 1 defined for
all z ∈ C, both functions are holomorphic on B(0, R) for all R > 0. As
f (x) = g(x) for all x ∈ R we conclude with Theorem 18.5 that f (z) = g(z)
for all z ∈ C.

(b) Define f ( n1 ) = sin( n1 ) for any n ∈ N, that is the function f is defined

only on the countable set { n1 : n ∈ N}. By Thoerem 18.5 we conclude that
f (z) = sin(z) for all z ∈ C. 

We have already used earlier the possibility to interchange summation

and integration. The next theorem provide a proof for this.

Theorem 18.7 (Interchange theorem) Let γ be a parametrisation for a

curve (path) in
P∞ C and let the functions U, uk , k ∈ N0 , be continuous on γ ∗ .
Assume that k=0 uk (z) converges to U (z) for all z ∈ γ ∗ and that there are
∗
P∞ Mk , k ∈ N0 , such that |uk (z)| ≤ Mk for all z ∈ γ and k ∈ N0 and
constants
that k=0 Mk converges. Then
∞ Z
X ∞
Z X Z
uk (z) dz = uk (z) dz = U (z) dz.
k=0 γ γ k=0 γ

Proof. The partial sum UN (z) := N ∗

P
k=0 uk (z) and U are continuous on γ .
Finite sums can be interchanged with integration
Z N Z
X Z  
U (z) dz − uk (z) dz = U (z) − UN (z) dz
γ k=0 γ γ
n o
≤ sup U (z) − UN (z) hlength(γ)i
z∈γ ∗
∞
n X o
≤ sup |uk (z)| hlength(γ)i
z∈γ ∗
k=N +1
∞
 X 
≤ Mk hlength(γ)i → 0 as N → ∞,
k=N +1
P∞
because k=0 Mk converges. 2

The next theorem is the Fundamental Theorem of Algebra which is derived

from Liouville’s theorem 16.5.

Theorem 18.8 (Fundamental Theorem of Algebra) Every non-constant

polynomial p on C has a root, i.e., p(z) = 0 for some z ∈ C.

129
Proof. Let p(z) = cn z n + cn−1 z n−1 + · · · + c1 z + c0 with c1 , . . . , cn ∈ C and
1
cn 6= 0. Assume that p has no root. Then f : C → C, z 7→ f (z) = p(z) is
holomorphic on C. As |p(z)| → ∞ as |z| → ∞, there exists R > 0 such that
1
|f (z)| = | p(z) | < 1 for |z| > R. Hence, f is bounded and therefore constant
due to Liouville’s theorem 16.5. This gives the required contradiction. 2

Remark 18.9 Let p a polynomial of order n. If p(a) = 0 for a ∈ C then

p(z) = (z − a)q(z). Iterating this (induction) gives: there are z1 , . . . , zn ∈ C
and a constant c ∈ C such that

p(z) = c(z − z1 )(z − z2 ) · · · (z − zn ), z ∈ C.

Every n-th order polynominal has n roots.

19 Laurent series and Cauchy’s Residue For-

mula
In the first part the we will extend Cauchy’s theorem to a larger class of curves
(paths). Then we study power expansions of functions around singularities.
These power expansions include also negative powers, the so-called Laurent
series.

19.1 Index
Definition 19.1 (Index) Let γ be a parametrisation of a closed path, and
consider the complement Ω := C \ γ ∗ (recall that γ ∗ = γ([a, b])). The index
of the curve (path) is the function Indγ : Ω → C defined by
Z
1 dw
Indγ (z) = , z ∈ Ω. (19.46)
2πi γ w − z

Without proof we cite the following theorem about the values of the index
function.

Theorem 19.2 The index function Indγ is an integer-valued function on

Ω and is constant in each component of Ω and is zero in the unbounded
component of Ω, see an example in figure 69.

130
 Figure 69: components of Ω of a path γ

Example 19.3 (Some index functions) The examples in figure 70 show

an example with index Indγ (z0 ) = 1, Indγ (z0 ) = −1 and Indγ (z0 ) = 2
respectively.

131
Figure 70: three index functions

Example 19.4 Consider the parametrisation γ : [0, 4π] → C, γ(t) = (1+i)+

eit of a circline around (1 + i) with radius 1, see figure 71 around z0 = 1 + i.
The index is
Z Z 4π
1 dw 1
Indγ ((1 + i)) = = e−it ieit dt = 2.
2πi γ w − 1 − i 2πi 0

Hence, we do know that

Z 
2i 
dz = −8π.
γ z−1−i

Is there a connection with the index? This is indeed the case, and we shall
show it later. 

132
 Figure 71: example with Indγ ((1 + i)) = 2

We want to enlarge the class of curves (paths) for Cauchy’s theorem. We

will do this step by step. The following is the first version which extends to
closed curves (paths) without specifying the orientation.

Cauchy Theorem (preliminary version)

Let f : D → C, D ⊂ C a region, be holomorphic and let γ be a parametrisa-
tion of a closed curve (path) in D. If all points of the complement C \ D are
lying in the outside O(γ) of the curve (path) Cauchy’s theorem holds, i.e.,
Z
f (z) dz = 0.
γ

We shall explore the meaning of the condition that the complement of the
domain of definition lies outside of the curve (path). We do this with a couple
of examples. In figure 72 we have two circle lines around the origin, ∂B(0, R)
and ∂B(0, r), r < R , one circle line is lying inside the other one. Consider
the path γ as the union of γ1 , the outer circle line in positive direction, the
straight connection of the outer circle with the inner circle γ2 , the inner circle
line γ3 with negative orientation and the straight line back (on the same line
as before but in different direction) γ4 .

133
 Figure 72: paths γ1 , γ2 , γ3 , γ4

Let f be a holomorphic function on the annulus of the two circle lines in

figure 72. Then from Theorem 15.7 we get that
Z Z
f (z) dz = f (z) dz,
γ1 γ3

but inside of the inner circle ∂B(0, r) we have points which do not belong to
the domain of definition and do not belong to the outside of the whole path
γ. A way out of this is the following. Pick ε > 0 and open both circle lines
by a small piece of length ε to get the key-hole path in figure 73.

134
 Figure 73: key-hole path

The curves (paths) in figure 74 and 75 are encircling both two singularity
points. We color the two components around the two singularities by red (the
upper one) and by blue (the lower one). Let γ be the closed paths enclosing
(surrounding) the two points respectively- one (figure 74) puts the points on
the outside and one (figure 75) encircles the upper point for example in one
movement, see the figure 74 and 75 respectively. Is there a difference in the
methods of encircling? In which components is the index Indγ 6= 0 ? If we
consider the curve (path) as a fence keeping some animals we would see no
difference at all. But there is a difference.

135
 Figure 74: enclosing two singularities - method (a)

In figure 74 both the red and blue component have zero index but in figure
75 the blue component has a non-zero index. This shows that for the curve
(path) γ in figure 74 the two singularities are lying outside γ.

136
 Figure 75: enclosing two singularities - method (b)

This discussion shall motivate the following final version of Cauchy’s theorem.

Theorem 19.5 (Cauchy’s theorem - final version) Let D ⊂ C be a re-

gion and the function f : D → C, D ⊂ C, be holomorphic and γ be a closed
curve (path) in D which does not surround a point in the complement C \ D,
that is, Indγ (z) = 0 for all z ∈ C \ D. Then
Z
f = 0.
γ

Theorem 19.6 (Cauchy’s formulae - index version) Let D ⊂ C be a

region and the function f : D → C, D ⊂ C, be holomorphic and γ be a closed
curve (path) in D which does not surround a point in the complement C \ D,
that is, Indγ (z) = 0 for all z ∈ C \ D. Then
Z
n! f (w)
Indγ (z)f (n)
(z) = dw, for all z ∈ D \ γ ∗ .
2πi γ (w − z)n+1

137
19.2 Laurent series
We study now power expansions with negative powers. We start with the
following Binomial expansions derived from the geometric series.

Binomial expansions
For |z| < 1 the geometric series gives the expansion of (1 − z)−1 with positive
powers
∞
1 X
= zn.
1−z n=0

What about |z| > 1? Because of

1
|z| > 1 ⇔ <1
z
an expansion of (1 − z1 )−1 is possible.
∞ −1
1 1 1 X
−n−1
X
=−   =− z =− zm for |z| > 1.
1−z z 1− 1 n=0 m=−∞
z

In the same way we can expand (a − z)−1 as a series in positive powers of z

if |z| < |a| and as a series in negative powers if |z| > |a|.

19.7 A series ∞
P
Definition
P∞ P∞an , an ∈ C, converges (to s = s1 + s2 ) if
n=−∞
n=0 an converges (to s1 ) and n=1 a−n converges (to s2 ). A power series
with negative and positive powers is called a Laurent series.

Suppose the function f is holomorphic in the punctuated ball B(a, R) \

{a}, and at the point a something nasty happens (singularity). If the function
is not holomorphic in a we cannot hope for a power series P∞ expansion. But
what about to expand the function f as a Laurent series n=−∞ cn (z − a)n
for 0 < |z − a| < R?

Theorem 19.8 (Laurent’s theorem) Let an annulus A,

A = {z ∈ C : R < |z − a| < S; 0 < R < S ≤ ∞, a ∈ C},

be given and let the function f be holomorphic on A. Then

∞
X
f (z) = cn (z − a)n for z ∈ A, (19.47)
n=−∞

138
where the coefficients are given by
Z
1 f (w)
cn = dw, n ∈ Z, (19.48)
2πi γ (w − a)n+1
with γ ∗ = ∂B(a, r) for any r with R < r < S.

Proof. We sketch only the main ideas. W.l.o.g. we set a = 0.

Figure 76: γ
e and γ
b in A

Pick z ∈ A and choose R < P < |z| < Q < S and consider the two closed
paths γ
e and γ e is the closed path enclosing z ∈ A and
b in figure 76, where γ
z ∈ O(bγ ). From (16.37) we get
Z
1 f (w)
f (z) = dw,
2πi γe w − z
and from Cauchy’s theorem 19.5 we have
Z
1 f (w)
0= dw.
2πi γb w − z
The integrals along the line segments connecting the two circlines ∂B(0, Q)
and ∂B(0, P ) in figure 75 cancel out each other and the remaining gives
Z Z
1 f (w) 1 f (w)
f (z) = dw − dw. (19.49)
2πi ∂B(0,Q) w − z 2πi ∂B(0,P ) w − z

139
We can expand the integrands (w − z)−1 in (19.49) in positive powers for
the first integral and in negative powers for the second integral in (19.49)
respectively (see Binomial expansions).
∞
1 1 1 1 X  z n
=   = for all w ∈ ∂B(0, Q)
w−z w 1− z w n=0 w
w

because | wz | < 1 for w ∈ ∂B(0, Q).

∞
1 1 1 1 X  w m
=−   =− for all w ∈ ∂B(0, P )
w−z z 1− w z m=0 z
z

because | wz | < 1 for w ∈ ∂B(0, P ). Inserting theses expansions into the

integrands in (19.49) and using Theorem 18.7 we get
∞  Z ∞ Z
X 1 f (w)  n X  1 
f (z) = n+1
z + f (w)w m
dw z −m−1 .
n=0
2πi ∂B(0,Q) w m=0
2πi ∂B(0,P )

Putting n = −m − 1 and an application of the Deformation Theorem 15.7

gives the proof. 2

It is often very difficult to calculate the coefficients of a Laurent expansion.

An easier way is to find Binomial expansions directly as the following example
shows.

Example 19.9 (Laurent expansions of a function) The function

1
f (z) =
z(1 − z)2

is holomorphic in A1 = {z ∈ C : 0 < |z − 1| < 1} and A2 = {z ∈ C : |z − 1| >

1}. Hence we have two Laurent expansions. As
1 1
f (z) = 2
 ,
(z − 1) 1 + (z − 1)

we conclude that the Laurent expansion for 0 < |z − 1| < 1 is

∞
X
f (z) = (−1)n (z − 1)n .
n=−2

140
For the annulus A2 we write
1 1
f (z) =  ,
(z − 1)3 1 + 1
z−1

and hence the Laurent expansion for |z − 1| > 1 is

−3
X
f (z) = (−1)n+1 (z − 1)n .
n=−∞

The notion singularity becomes clear in the following.

Notation 19.10 Let f : D → C, D ⊂ C, be a function.

(a) A point a ∈ D is a regular point of the function f if f is holomorphic

at a.

(b) A point a ∈ D is a singularity of f if a is not a regular point but a

limit point of regular points, e.g., a is singularity of
1
z−a
because a is limit of the regular points zn = a + n1 .

(c) A singularity of f is said to be isolated if f is holomorphic in the open

punctuated ball B(a, R) \ {a} for some R > 0.

Theorem 19.8 indicates that the isolated singularities are the candidates
for Laurent expansions. We need the following classification of isolated sin-
gularities.

Classification of isolated singularities Suppose that the function f has

an isolated singularity at a, hence f is holomorphic in some annulus

A = {z ∈ C : 0 < |z − a| < r, r > 0}.

Theorem 19.8 gives

−1
X ∞
X
n
f (z) = cn (z − a) + cn (z − a)n , z ∈ A. (19.50)
n=−∞ n=0

141
The first term on the right hand side of (19.50) is called the principal part
of the Laurent expansion. The isolated singularity a is said to be

• a removable singularity if cn = 0 for all n < 0.

• a pole of order m (m ≥ 1) if c−m 6= 0 and cn = 0 for all n < −m.
• an essential singularity if there does not exist m such that cn = 0 for
n < −m.

The example 1/(z − 1)2 has at z = 1 a pole of order 2, also called double
pole.

Definition 19.11 (Residue) Let f : D → C, D ⊂ C, be a function which

has a pole at a ∈ D. The residue of the function f at a is the unique
coefficient c−1 of the term (z − a)−1 of the Laurent expansion of f about a
and it is denoted by res{f ; a}.

We are now able to answer our questions about when integrals of functions
along closed paths vanish or not. If we surround poles of the integrand the
integral does not vanish, because there is a reminder which is given by the
residue of that pole. This is the content of the last theorem.

Theorem 19.12 (Cauchy’s residue theorem) Let the function f be holo-

morphic inside and on a positively oriented contour γ except at a finite num-
ber of poles a1 , . . . , aN ∈ I(γ). Then

Z N
X
f (z) dz = 2πi res{f ; ak }. (19.51)
γ k=1

Proof. The N poles give exactly N principal parts

−1
X
n
fk (z) = c(k)
n (z − ak ) k = 1, . . . , N.
n=−∞

For any k = 1, . . . , N there is mk ∈ N with c(k)

−mk 6= 0 and cn = 0 for all
(k)

n < −mk . Integration of each principal part gives (see fundamental integral)
mk
c(k)
Z X −n
fk (z) dz = = 2πic(k)
−1 = 2πi res{f ; ak }.
γ n=1
(z − ak )n

142
 PN
The function g : I(γ) → C, given by g(z) = f (z)− k=1 fk (z), is holomorphic
and Theorem 19.5, i.e.,
Z Z N Z
X
0= g(z) dz = f (z) dz − fk (z) dz,
γ γ k=1 γ

and we conclude with the proof. 2

In order to take advantage of the last theorem we provide some simple
rules for calculating the residue of a function.

Lemma 19.13 Let B(a, R) be the open ball around a ∈ C with radius R > 0,
and assume that the function f : B(a, R) → C is holomorphic on B(a, R) \
{a}.

(a) If a is simple pole (i.e. a pole of order 1), then

res{f ; a} = lim (z − a)f (z).

z→a

Furthermore, if the function g : B(a, R) → C is holomorphic and g(a) 6=

g(z)
0 such that f (z) = z−a for all z ∈ B(a, R) \ {a}, then

res{f ; a} = g(a).

(b) Let h, k : B(a, R) → C be holomorphic on the ball B(a, R) such that

f (z) = h(z)
k(z)
or all z ∈ B(a, R) \ {a}. Furthermore, assume that h(a) 6=
0, k(a) = 0, and k 0 (a) 6= 0, then

h(a)
res{f ; a} = .
k 0 (a)

g(z)
(c) Let f have a pole at a of order m > 1 and let f (z) = (z−a) m,z ∈

B(a, R) \ {a}, with g holomorphic on B(a, R) and g(a) 6= 0. Then

1
res{f ; a} = g (m−1) (a).
(m − 1)!

Proof. (a) This follows immediately from the the Laurent series expansion
of the function f .
(b)
h(z) z−a h(a)
lim (z − a) = h(a) lim = 0 .
z→a k(z) z→a k(z) − k(a) k (a)

143
(c) Cauchy’s formulae give

(m − 1)! (m − 1)!
Z Z
(m−1) g(z)
g (a) = m
dz = f (z) dz
2πi ∂B(a,R/2) (z − a) 2πi ∂B(a,R/2)

= (m − 1)! res{f ; a}.

2

144