Vector Analysis Notes

Matthew Hutton
Autumn 2006 Lectures
1
Contents
0 Introduction 4
0.1 What is vector analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
0.2 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
0.2.1 Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
0.2.2 Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
0.2.3 Inner Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
0.2.4 Partial derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1 Lecture 1 The real thing 5
1.1 Gradients and Directional derivatives . . . . . . . . . . . . . . . . . . . . 5
1.2 Directional Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.1 Generally in one dimension . . . . . . . . . . . . . . . . . . . . . 7
1.2.2 Generally in n dimensions . . . . . . . . . . . . . . . . . . . . . . 7
2 Visualisation of a function f : R
n
R
n
7
2.1 Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3 Line Integrals 11
3.1 Integrating by scalar elds . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Integrating vector elds . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4 Gradient Vector Fields 12
5 Surface Integrals 16
6 Divergence of Vector Fields 20
7 Gauss Divergence Theorem 23
8 Integration by Parts 28
8.1 Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
8.2 Uniqueness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
9 Greens Theorem 29
10 Stokes Theorem (curls in R
3
) 33
11 Spherical Coordinates 36
12 Complex Dierentation 36
12.1 Basic properties of Complex Numbers . . . . . . . . . . . . . . . . . . . . 37
12.2 Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
12.3 Continuity & Dierentiation . . . . . . . . . . . . . . . . . . . . . . . . . 38
13 Complex power series 39
2
14 Holomorphic Functions 42
15 Complex Integration 44
16 Cauchys theorem 46
17 Cauchy Integral Formula 48
18 Real Integrals 54
19 Power Series for holomorphic functions 56
20 Real Sums 60
3
Lecture 0:3/10/06
0 Introduction
0.1 What is vector analysis
In analysis dierentiation and integration were mostly considered on R or on rectangles
(between points a and b). However a function on a circle is as valid as on a straight line.
Vector analysis generalises these results onto curves, surfaces and volumes in R
n
Example 0.1. The normal way to calculate an integral is to nd an anti-derivative of
the function and use the fundamental theorem of calculus (FTC)
f(x) + F
(1)
(x)
_
b
a
f(x)dx =
_
b
a
F(x)dx = F(b) F(a) (0.1)
The value of
_
b
a
f(x)dx can be computed by looking at the boundary points, a and b.
This can be generalised to R
n
, by Gauss Theorem. Gauss Theorem says that we can
nd the area with just the boundary lines.
0.2 Notation
There are many dierent notations you may use, especially in Physics/Engeneering
1
0.2.1 Vectors
x R
n
, x = (x
1
, x
2
, . . . , x
n
)
alternatives x, x, x, x I normally use x, physicists generally use r = (x, y, z) where
r =
_
x
2
+ y
2
+ z
2
, however this is no good with n dimensions as you soon run out of
letters!
0.2.2 Functions
f : R
m
R
n
with component functions f
1
, . . . , f
n
R
m
R, alternate ways of showing
functions are

f , f, . . .
0.2.3 Inner Product
x, y) =
n

i=1
x
i
y
i
for x, y R
n 2
(0.2)
alternate (xy), x y, x
T
y
1
The biggest challenge is getting LaTeX to write them all... ;)
2
In this case this is actually the scalar (dot) product
4
Figure 1: Graph showing a peak
0.2.4 Partial derivatives
for f : R
n
R, x R
n
d
dx
f(x) = lim
h0
f(x + he
i
) + f(x)
h
alternates, f(x),
f
x
i
(x), d
x
i
, . . .
f
t
(t, x) =

t
f(t, x)
f
x
(t, x) =

x
f(t, x)
1 Lecture 1 The real thing
1.1 Gradients and Directional derivatives
How does a function f : R
n
R change when we move from a point x R
n
in some
direction y R
n
? This can be seen in gure 1 We can reduce the problem to one
dimension. Consider : R R, f(x+y) the change of f at a point x in direction
of y equals the change of at point = 0 and thus is (0).
Denition 1.1 (The directional derivative). f : R
n
R at x R
n
in the direction of
y R
n
D
y
f(x) = lim
0
f(x + y)

(1.1)
5
Figure 2: Graph showing how the directional derivative varies
Example 1.1.
f : R
2
R
2
, f(x) = x
2
1
+ x
2
2
f(x + y) = (x
1
+ y
1
)
2
+ (x
1
+ y
2
)
2
= x
2
1
+ x
2
2
+ 2x
1
y
1
+ 2x
2
y
2
+
2
y
2
1
+
2
y
2
2
D
y
f(x) = 2x
1
y
1
+ 2x
2
y
2
= 2x, y)
This is shown in gure 2.
3
In this case this gives:

() =
n

i=1

i
f(x
i
+ y
i
)
D
y
f(x) =

(0) =
n

i=1

i
f(x)y
i
where
i
f(x) is the gradient part (of f).
1.2 Directional Derivatives
D
y
f(x) = lim
0
f(x + y) f(x)

Do you really need to calculate this for every D

y
f(x) for all directions y?
3
Its probably worth looking over section 1.2 rst as thats the order we did it in lectures.
6
We have to calculate the derivative of : x + y f(x + y)
R
g
R
n

f
R
In general this is shown in the next two subsections.
1.2.1 Generally in one dimension
_
f(g()
_

= f

_
g()
_
g

()
1.2.2 Generally in n dimensions
_
f(g()
_

=
n

i=1

i
f

_
g()
_
g

i
()
Denition 1.2. For R
n
R the vector f = (
i
f, . . . ,
n
f) is called the gradient of f.
Alternative notations include gradf, . . . f
Denition 1.3 (Cauchy-Schwarz inequality). If x, y C then
[x, y)[
2
x, x) y, y) [1] (1.2)
Another form of this, which we use here is:
[x, y)[ |x| |y| (1.3)
Remark 1.1. From the Cauchy Schwarz inequality (equation 1.3) we get:
[D
y
f(x)[ = [f(x), y)[ |f(x)||y| (1.4)
Where [[y[[ = 1 we get:
[[f(x)[[ [[D
y
f(x)[[ [[f(x)[[
And for: y =
f(x)
||f(x)||
we get
[[D
y
f(x)[ =
_
f(x),
f(x)
|f(x)|
_
This then implies that y is maximal if y points in direction of the gradient.
2 Visualisation of a function f : R
n
R
n
Graphs of scalar elds.
Denition 2.1. f : D R where D R
m
Example 2.1. m = n = 1 as in Analysis I, f(x) = sin(x), this is shown in gure 3
7
Figure 3: Graph of f(x) = sin x
Figure 4: Graph of the function: z =
sin
(
x
2
+3y
2
)
0.1+r
2
+ (x
2
+ 5y
2
)
exp
(
1r
2
)
2
, r =
_
x
2
+ y
2
8
Figure 5: This graph is the original Grapher example, it is included because it looks cool,
and not a boring uniform orange, if it confuses you ignore it, it adds nothing to gure
4 in terms of vector analysis. In the image the colour gradient represents the height (z
value, on x, y, z axis), but could be use to represent some extra criteria.
9
Example 2.2. For m=2,n=1 we can draw something as shown in gure 4
4
For m 3, n = 1 it is dicult, colour coding could be used to represent a variable,
in a similar way to how gure 5 does for the z direction. Applications of this include
temperatures distributions on 3D bodies or pressure in a liquid.
f
1
(c) = x R
n
[ f(x) = 0
5
f(x, y) = x
2
+ y
2
2.1 Curves
These can be described in two ways:
1. Implicitly giving the graph c R
n
2. Explicitly as parametric curves : R R
n
Example 2.3.
1 = x
2
+ y
2
(2.1)
Then equation 2.1 can be written parametrically as:
(t) =
_
cos t
sin t
_
(2.2)
For a curve : R R
n
where t is the position at time t.

(t) = lim
h0
(t + h) (t)
h
If ,= 0 then

(t) is a tangent vector and the tangent line is given by (t) +

(t)
Denition 2.2. A vector x R
n
is orthogonal to a curve : R R
n
at the point (t)
if x, (t)) = 0 i.e. if it is orthogonal to the tangent line.
Lemma 2.1. Let f : R
n
R be a scalar function, a R
n
. Then f(a) : x
R
n
[f(x) f(a) this means the gradient is orthogonal to the tangent lines.
Proof. Let : R x [ f(x) = f(a) be a curve with (0) = a.
= f((t),

(t))

t=0
= f(a),

(t))
=
m

i=1

i
(t)
i
f

((t))
Then
0 =
d
dt
f((t))

t=0
4
Sorry for the excessively complex example. Blame Apple for making such a cool graph.[/ apple/maths
geek]
5
It seems something is missing here
10
3 Line Integrals
We want to take integrals along a curve c R
n
, there are two methods of integrating
line integrals.
3.1 Integrating by scalar elds
Denition 3.1 (Length of curve). Let : [a, b] R
n
be a curve, then the scalar line of
u : R
n
R is:
_

u =
_
b
a
u
_
(t)
_
[[

(t)[[dt
other forms of line integrals
_

uds the
6
has no useful role.
3.2 Integrating vector elds
Denition 3.2. Ley be a curve and f a vector eld the tangent line integral of f along
[a, b] R
n
a vector eld. Then the tangent line integral of f along is given by:
_

f =
_
b
a

f
_
(t)
_
,

(t)
_
dt
alternative notations are:
_

f

T

ds,
_

f

ds,
_
C
f

ds
Lecture 5: 16/10/06
Example 3.1. Length of the circle line, let
(t) =
_
cos t
sin t
_
t [0, 2]

(t) =
_
sin t
cos t
_
t [0, 2]
[[

(t)[[ =
_
(sin t)
2
+ (cos t)
2
=

1 = 1
_
C
1 =
_
2
0
1 1dt = 2
This answer is the same as you get from the old fashioned method of nding the circum-
ference using circumference = 2r
Remark 3.1. The tangent line integral can be written as:
_
C
f =
_
b
a
_
f
_
(t)
_
,

(t)
[[

(t)[[
_
[[

(t)[[dt
with an inner product.
7
_
C
f is the scalar line integral of the component of f along the
tangent line.
6
a piece of LaTeX is missing here check original notes
7
doesnt make much sense to me now, maybe I was distracted ;)
11
Example 3.2. Work done when moving a mass along the line cos x
(t) =
_
t
cos t
_
f(x, y) =
_
0
mg
_

f =
_

0
__
0
mg
_
,
_
1
sin t
__
dt =
_

0
mg sin tdt =
_
mg cos t
_

0
= mg+mg = 2mg
(3.1)
This example is continued as example 4.1 in the next chapter.
4 Gradient Vector Fields
Denition 4.1. A gradient vector eld is a vector eld f : R
n
R
n
with f = V for
some V : R
n
R, V is called the potential of f.
Remark 4.1. 1. V is not unique as V = (V + C), c R, this means the
potential is not unique.
2. Not every vector eld is a gradient vector eld
Theorem 4.1 (Fundamental Theorem of Calculus for gradient vector eld). Let V : R
R? be a scalar eld f = V ? and : [a, b] R
n
? a curve then
_

f = V
_
(b)
_
V
_
(a)
_
Proof. The ??? V
_
(t)
_
has R R derivative.
_
V
_
(t)
_
_

=
n

i=1

i
V

_
(t)
_

(t) =
_
V
_
(t)
_
,

(t)
_
(4.1)
Therefore
8
_

f =
_
b
a
_
f
_
(t)
_
,

(t)
_
dt =
_
b
a
_
V
_
(t)
_
_

dt = V
_
(b)
_
V
_
(a)
_
(4.2)
As we now know:
_

f = V
_
(b)
_
V
_
(a)
_
8
More destractedness I think ;)
12
Figure 6: Diagram showing routes between two points a and b, and a loop about a point
a
Example 4.1.
f =
_
0
mg
_
can be written as V

and V = mgy for every curve, (t) = (x(t), y(t)) we get:

f = V ((b)) + V ((a)) = mgy(b) mgy(a)

but in this case y(b) = 1, y(a) = 1 = mg +mg = 2mg
9
f is a vector eld. is a curve.

f which is the integral of f over the curve .

f = V
_
(b)
_
V
_
(a)
_
Remark 4.2. If f
10
is a vector eld and is a line then
_

f does not depend on the

path but only the two end points.
Remark 4.3. If f = V then gure 6 implies that
_

1
f = V (b) V (a) and
_

2
f = V (a) V (b)
this means that
_

1
f =
_

2
f.
Denition 4.2. A loop is where the end point is the start point as shown in gure 6.
Example 4.2.
f(x, y) =
_
y
x
_
(t) =
_
cos t
sin t
_
t [0, 2] =
_

f =
_
2
0
__
sin t
cos t
_
,
_
sin t
cos t
__
dt
_
2
0
sin
2
t + cos
2
tdt =
_
2
0
1 dt = 2
this means that f is not a gradient vector eld.
9
2mg is what we got before
10
not 100% on whether this is f
13
Figure 7: Graph of a vector eld
V (x) = V (0) +
_

x
f
where
x
is the straight line from 0 to x for every potential V . If V
1
and V
2
are two
dierent potentials, then:
V
1
(x) V
2
(x) = V
1
(0) +

f V
2
(0)

f
Is then constant.
Example 4.3 (Finding a potential). Let f(x, y) =
1
2
_
x
y
_
(this vector eld is shown
in gure 7) Suppose f = V then this implies that
V
x
=
1
2
x V =
x
2
4
+ C
1
V
y
=
1
2
y V =
y
2
4
+ C
2
These two equations imply that:
V =
y
2
4

x
2
4
+ C
so V (x, y) =
y
2
4

x
2
4
is a valid solution.
14
Figure 8: Graph of a vector eld
Example 4.4. Let f(x, y) =
_
2y
x + y
_
(this vector eld is shown in gure 8) Suppose
f = V then
V
x
= 2y V = 2xy + C
1
(y)
V
y
= x + y V = xy +
1
2
y
2
+ C
2
(x)
This has no solution which implies that f cannot be a gradient eld.
Denition 4.3. f : R R is called a radial vector eld if
f(x) =
_
g(|x|)
x
x
if x ,= 0
0 x = 0
19/10/06
_

f =
_

V = V ((b)) V ((a))
(if f is a gradient)
Example 4.5. Radial vector elds: Let:
f(x) =
_
g(|x|)
x
x
if x ,= 0
0 x = 0
, where g : (0, ) R We always nd (0, ) R with

= g. Let v(x) = (|x|) =

_
_
x
2
1
+ + x
2
n
_
We then get:

x
v(x) = (|x|) =
1
2
_
x
2
1
+ + x
2
n
2x
i
=

(|x|)
x
i
|x|
= g
x
i
|x|
15
= f
i
(x) V = f
Thus we now know that radial vector elds are always gradients.
5 Surface Integrals
There are two methods of describing a surface in R
3
1. Level set of f : R
3
R
2. Parameterisation r : A R
3
where A R
3
r(s, t) =
_
_
r
1
(s, t)
r
2
(s, t)
r
3
(s, t)
_
_
Find the surface cs normal vectors for level sets:
f c

N =
f
|f|
for parameterisations
11
A plane is dened by two vectors and a point. Or a point and
a vector orthogonal to the plane. At r(s, t) the vectors
r
s
and
r
t
are tangent vectors of
the surface:
c N =
r
s

r
t
is therefore normal to c this implies that

N =
r
s

r
t
_
_
r
s

r
t
_
_
12
_
_
x
1
x
2
x
3
_
_

_
_
y
1
y
2
y
3
_
_
=
_
_
x
2
y
3
x
3
y
2
x
3
y
1
x
1
y
3
x
1
y
2
x
2
y
1
_
_
Denition 5.1. Let r : A R
3
, A R
2
be a parameterisation of some surface C. Then
the scalar surface integral of f : R
3
R is:
_
r
f =
__
A
f
_
r(s, t)
_
_
_
_
_
r
s

r
t
_
_
_
_
dsdt
Remark 5.1.
_
r
1 is the surface area of C.
Remark 5.2. Alternative notations
_
C
f
1
,
_
C
fds,
_
C
fdA (some diagrams Im not copy-
ing)
_
r
f

s,t
f
_
r(s, t)
_
_
_
_
_
r
s

r
t
_
_
_
_
dsdt
11
(see paper notes)
12
next bit unclear
16
Figure 9: Diagram showing how the area of a parallelogram is found.
Figure 10: Diagram showing a sphere.
23/10/06
_

f =
__
A
f
_
r(s, t)
_
_
_
_
_
r
s

r
t
_
_
_
_
dsdt (5.1)
Properties of x, let z = xy then:
1. z x and z y
2. |z| is the area of the parallelogram spanned by x and y (shown if gure ??.
3. The orientation of z is given by the right hand rule.
13
Example 5.1 (Spherical Cap). First we parameterise it.
r(s, t) =
_
_
cos s cos t
cos s sin t
sin s
_
_
r
s
=
_
_
sin s cos t
sin s sin t
cos s
_
_
13
http://en.wikipedia.org/wiki/Right_hand_rule
17
Figure 11: Diagram showing a y = cos x for x [, ]
So therefore
_
_
_
_
r
s

r
t
_
_
_
_
=
_
cos
4
s + cos
2
s sin
2
s = cos s
_
cos
2
s + sin
2
s = cos s (5.2)
Therefore the area of the cap with radius 1, is:
_

1 =
_

_
2
0
1 cos sdsdt
=
_
2
0
2 cos sds
= [2 sin s]

2
0
= 2 2 sin
= 2
_
1 sin
_
If =

2
, sin = 1 So therefore
2(1 1) = 4 (5.3)
Which is the surface area of a sphere of radius 1.
Example 5.2 (Newtons kissing problem). An example of this in one and two dimesions
is shown in gure 12 For R
3
how many simultaneously touching balls can touch a given
ball?
14
To nd an upper bound by calculating the area shadowed surface area taken
by each ball, this is shown in gure 13.
1
2
= sin , as we can dene

2
=
14
J Leech proved that the correct number is 12, in Math Gazette 40 (1956) p22/23
18
Figure 12: Diagram showing Newtons kissing problem in one and two dimensions.
Figure 13: Diagram showing the shadowed area in Newtons kissing problem
19
1
2
= cos =

3
Shadowed surface has area 2(1 sin ) as we found in example 5.1.
2
_
1 sin

3
_
= 2
_
1

3
2
_
= 2

2

2
=
_
2

3
_

This therefore gives an upper bound of

4
&

_
2

3
_
&

= 14.93 (2dp)
This gives between 12 and 14 balls. Below are the results for various values of n
n=1 2
n=2 6
n=3 12
n=4 24*
n=8 240
n=24 196560
For n = 4 the answer is probably 24 but a strict upper bound of 25 is
as far as we know for sure. [2]
15
24/10/06
_

V = V ((b)) V ((a))
6 Divergence of Vector Fields
Denition 6.1. Let S R
3
be a surface with unit normals

N : S R
3
then the ux
of f : R
3
R
3
across S in the direction of

N is
_
S
f,

N). The inner product tells you
how much of f is pointing in the direction

N and this is shown in gure 14. Alternative
notations include
_
S
f

N and
_
S
fd

s .
Example 6.1 (Flux out of a box). If we take the box = [a
x
, b
x
] [a
y
, b
y
] [a
z
, b
z
], this
is shown in gure 15
The ux through the top
Using the parameterisation
r
t
= (x, y) =
_
_
x
y
t
z
_
_
, x [a
x
, b
x
], y [a
y
, b
y
]
f,

N) =
_
b
x
a
x
_
b
y
a
y
f
3
_
_
_
_
r
x

r
y
_
_
_
_
dydx
15
For this citation the ability to open .ps and .gz les required, i.e. Linux/Mac OS X, if you use
Windows youre basically screwed, you can use a command line tool to open the .gz but then you need
a copy of ghostscript or Acrobat Professional (well for the cost of that you might as well buy a Mac ;)
), drop me an email and I can send you a copy.
20
Figure 14: Diagram showing how the inner product works
Figure 15: Flux Box
21
_
_
_
_
r
x

r
x

r
y
_
_
_
_
=
_
_
_
_
_
_
_
_
1
0
0
_
_

_
_
0
1
0
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
0
0
1
_
_
_
_
_
_
_
_
= 1 (6.1)
_
s
t
f,

N) =
_
b
x
a
x
_
b
y
a
y
f
3
(x, y, t
z
)dydx
The ux through the bottom

N =
_
_
0
0
1
_
_
f,

N) = f
3
b
2
changed a
2
_
S
bottom
f,

N) =
_
b
x
a
x
_
b
y
a
y
f
3
(x, y, t
z
)dydx
Top+Bottom
_
b
x
a
x
_
b
y
a
y
f
3
(x, y, b
2
) f
3
(x, y, a
2
)dydx
=
_
b
x
a
x
_
b
y
a
y
_
b
z
a
z
f
3
z
(x, y, z)dz
. .
by FTC
dydx (6.2)
Similarly
_
S
left
S
right
f,

N) =
_
b
x
a
x
_
b
y
a
y
_
b
z
a
z
f
1
x
(x, y, z)dzdydx (6.3)
_
S
front
S
back
f,

N) =
_
b
x
a
x
_
b
y
a
y
_
b
z
a
z
f
2
y
dxdydz (6.4)
Taking the sum of equations 6.2, 6.3 and 6.4 we then get the total ux which is:
_

f,

N) =
_
b
x
a
x
_
b
y
a
y
_
b
z
a
z
f
1
x
+
f
2
y
f
3
z
dzdydx (6.5)
Denition 6.2 (Divergence). The divergence of a vector eld f : R
n
R
n
is divf :
R
n
R
divf =
n

i=0
f
i
x
Alternative notations: V ,

V and V
Remark 6.1. Scalar eld to vector eld
V : R
n
R gradV : R
n
R
n
f : R
n
R
n
f : R
n
R
22
Figure 16: Diagram showing f(x)
Denition 6.3. Let V : R
n
R be a scalar eld. The Laplacian of V is V = div gradV
(Scalar eld)
gradV =
_
V
x
1
V
x
n
_
div gradV =
n

i=1
V
i
k
i
Alternative notations include: V , V ,
2
V .
26/10/06
divf(x) =

f
x
(x)
= [a
1
, b
1
] [a
2
, b
2
] [a
3
, b
3
]
_

f,

N) =
_
divf(x)dx (6.6)
Remark 6.2. If is a small box around a R
3
then = [a
1
, a
1
+] [a
2
, a
2
+
] [a
3
, a
3
+ ] Then
_

div(x)dx divf() V
a
()
divf(a)

=
_

f,

N)
V
0
()
Thus the divergence gives the outward ux per unit volume.
Remembering that divf(x) = f
7 Gauss Divergence Theorem
Denition 7.1 (Divergence). The divergence of a C
1
vector eld f R
3
is:
divV = V =
3

i=1
f
i
x
i
(7.1)
23
Figure 17: Positive Divergence
Figure 18: Zero Divergence
Figure 19: Negative Divergence
24
Theorem 7.1 (Divergence Theorem). Let R
3
be a bounded with a surface and
outward unit normals

N. Let V : R
3
be continuously dierentiable, then:
_

V,

N) =
_

divV (x)dx (7.2)

Remark 7.1. The theorem also works for n ,= 3 but one has to dene the surface integral,
n = 2 is a line and
_

is a (scalar) line integral.

Remark 7.2. divf(x)dx is just an iterated integral, alternative notation is:
_

divfdV (n =
3),
_

divfdA (n = 2).
Example 7.1. Box in R
3
already done in example 6.1
Example 7.2. Ball of radius R in R
n
= B(0, R) R
n
= Shell of Sphere,

N(x) =
x
R
Let:
f(x) = x divf(x) =
x
1
x
1
+ +
x
n
x
n
= 1 + + 1 = n

_
B(0,R)
ndx =
_
B(0, R)
_
x,
x
R
_
_
B(0,R)
R n Volume of Ball = R Surface Area of Ball
For n = 2 then VolB(0, R) = R
2
This means that the length of B(0, R) = 2R, for
n = 3 the volume of the ball is
4
3
R
3
And the surface area is
3
R
times that.
30/10/06
Sketch Proof of Divergence theorem.
1. As we covered in example 6.1, it is already true for boxes.
2. Therefore we can use the fact that it holds for boxes to suppose the theorem holds
for
1
,
2
as shown in gure 20, what about the region
1

2
? Then:
_

_
V,

N
_
=
_

1
_
V,

N
_
+
_

2
_
V,

N
_
Since the contribution form the shared boundary cancels as they are in opposite
directions, this is equal to:
=
_

1
V +
_

2
V =
_

2
V
16
16
Im using nabla instead of div here as its easier
25
Figure 20: Diagram of 2 boxes for sketch proof of divergence theorem
3. For simple regions (in R
2
)
is simple =
_
(x, y)

f(y) x g(y) y [a, b]

_
=
_
(x, y)

f(x) y g(x) x [c, d]
_
So a circle with a hole in the middle of it (like a wheel with a missing axle) isnt
simple but if you cut it in half into a semi circle then it would be simple.
17
.
4. If we show is simple then:
_

V
1
x
dA =
_
b
a
_
g(y)
f(y)
V
x
1
(x, y)dxdy =
_
b
a
V
1
(f(y), y)
_
V
1
_
f(y), y)
_
dy (7.3)
Flux of
_
V
1
0
_
through the boundary, the region is shown in gure 21 The ux

k
() =
_
g(y)
y
_
for y [a, b]

k
() =
_
g

(y)
1
_
Generally we know that
_
x
y
_

_
y
x
_

N =
1
|
R
|
_
g

(y)
1
_
=
1
_
1 +
_
g

(y)
_
2
_
g

(y)
1
_
_

k
__
V
1
0
_
,

N
_
17
In metric spaces language it is simple if it is topologically equivalent to a square
26
Figure 21: Diagram showing the region with two boundary functions and two ux
functions.
1
|

(y)|
__
V
1
0
_
,
_
1
g

(y)
__
_
b
a
V
1
1
$
$
$
$$
|

R
(y)|
$
$
$
$$
|

R
(y)|dy
=
_
b
a
V
1
_
g(y), y
_
dy
Similarly for the LHS:
_

L
__
V
1
0
_
,

N
_
=
_
b
a
V
1
_
f(y), y
_
dy
As
_

T
__
V
1
0
_
,

N
_
=
_

B
__
V
1
0
_
,

N
_
= 0

_
b
a
V
1
_
g(y), y
_
V
1
_
f(y), y
_
dy
=
_

_
V
1
0
_
dA
This then shows that the theorem holds for V =
_
V
1
0
_
. If we then use

f and g
instead of f and g, and if we interchange x and y we get that:
_

__
0
V
2
_
,

N
_
=
_

_
0
V
2
_
dA
Adding these results together completes the proof.
27
8 Integration by Parts
If the following hold:
R,
18
=
Proposition 8.1. Then for f : R where f is C
1
. Then:
_

f
x
=
_

f

N
19
(8.1)
Proof. Let v = (f, 0, 0) then:
v =
f
x
1
v,

N) = f

N
1
If the divergence thoerem is applied to the special vector eld v given in equation 8.1,
then using (0, f, 0) and (0, 0, f) we can also show it for i = 2, 3.
Remark 8.1. The full divergence theorem (theorem 7.1) can be derived from equation
8.1
Proposition 8.2. Integration by parts:
_
b
a
uv

dx =
_
uv

b
a

_
b
a
u

vdx (8.2)
Then applying equation 8.1 to f = gh we get:
f
x
i
=
g
x
i
h + g
h
x
i

g
x
i
=
_

g
h
x
i
+
_

gh

N
i
Proposition 8.3. Let g : R be twice continuously dierentiable, i.e. apply proposi-
tion 8.1to f =
g
x
i

2
g
x
2
i
=
_

y
x
i

N
i
Summing up over i then gives
_

g =
_

g,

N) (8.3)
Proposition 8.4. Applying proposition 8.2 with g =
f
x
i
and sum over i we get:
_

fh =
_

f, h) +
_

gf,

N)
gradf, h) gradf, gradh) (8.4)
18
This means the closure of as in metric spaces
19
in the lectures described as

N
i
for i = 1, 2, 3.
28
(a, c) (b, c)
(b, d) (a, d)

T
Figure 22: A rectangle
8.1 Application
Temperature distribution in steady state. R
3
a piece of material. Fix temperature
f(x) x . In steady state the temperature solves:
_
T(x) = 0 x
T(x) = f(x) x
The existence of a solution s of this is a dicult question.
8.2 Uniqueness
Assume T and

T are solutions, then D = T

T solves
D = T

T = 0 0 = 0 x
D(x) = T(x)

T(x) = f(x) f(x) = 0 x
Using proposition 8.4 we get:
_

D
. .
=0
=
_

D, D) +
_

DD,

N)
_

|D|
2
= 0
D(x) = 0 x
D is constant and D = 0 at the boundary, therefore T =

T so T is a unqiue solution.
9 Greens Theorem
20
20
This section is done by Matthew Pusey
29
Consider a function f : R
2
R
2
and a rectangle = [a, b] [c, d] with boundary unit
tangent vectors

T in the anticlockwise (positive) direction. This is shown in Figure 22.
On the right side,
R
. which is the line segment from (b, c) to (b, d):
_

R
_
f,

T
_
=
_
d
c
_
f(b, y),
_
0
1
__
dy =
_
d
c
f
2
(b, y)dy
The remaining sides,
L
,
B
,
T
are similar (but take care with signs):
_

L
_
f,

T
_
=
_
d
c
f
2
(a, y)dy
_

B
_
f,

T
_
=
_
b
a
f
1
(x, c)dx
_

T
_
f,

T
_
=
_
b
a
f
1
(x, d)dx
Summing these gives:
_

_
f,

T
_
=
_
d
c
f
2
(b, y) f
2
(a, y)dy +
_
b
a
f
1
(x, c) f
1
(x, d)dx
The Fundamental Theorem of Calculus then gives:
=
_
d
c
_
b
a
f
2
x
(x, y)dxdy
_
b
a
_
d
c
f
1
y
(x, y)dydx
=
_
d
c
_
b
a
f
2
x

f
1
y
dydx
Denition 9.1 (2-D curl). For f : R
2
R
2
, the curl of f is given by:
curl f(x) =
f
2
x

f
1
y
Remark 9.1.
_

_
f,

T
_
is the circulation around . By talking small boxes we nd that
curl f(x) is the circulation of f around x.
Sometimes you can see what the curl is:
Example 9.1.
v(x, y) =
_
y
x
_
This is shown in gure 23.
curl v(x, y) = 1 (1) = 2
30
Figure 23: Diagram showing a vector eld.
Figure 24: Diagram showing a vector eld.
31
Figure 25: Diagram showing a vector eld.
Example 9.2.
v(x, y) =
_
x
y
_
This is shown in gure 24.
curl v(x, y) = 0 0 = 0
Example 9.3.
v(x, y) =
_
0
x
_
This is shown in gure 25.
curl v(x, y) = 1 0 = 1
Theorem 9.1 (Greens Theorem or Stokes Theorem in R
2
). Let be a bounded region
in R
2
, and

T be positively oriented tangent vectors for . If f :

R
2
is continously
dierentiable then:
_

_
f,

T
_
=
_

curl f
Proof. Dene g =
_
f
2
f
1
_
.
By the Divergence Theorem:
_

_
g,

N
_
=
_

div g
By considering the relationship between

N and

T, and the denition of div g this
becomes:
_

_
f,

T
_
=
_

curl f
32
10 Stokes Theorem (curls in R
3
)
6/11/06
Theorem 10.1. Let o R
3
be a bounded surface with

N as unit normals of o, and

T unit tangent vectors to the boundary line o. Let f : o o R

3
be continuously
dierentiable. If
_

N,

T
_
is positively oriented, then:
_
S
_
curlf,

N
_

_
S
f,

T) = 0
Remark 10.1. Positive oriented means that if

N points upwards, the tangent vectors
are anti-clockwise,and if

N points downwards, the tangent vectors are clockwise. (i.e. the
right hand rule applies).
Remark 10.2. In R
3
, curl = f
_
_

2
f
3

3
f
2

3
f
1

1
f
3

1
f
2

2
f
1
_
_
Remark 10.3. In R
2
curlg =
1
g
2

2
g
1
Example 10.1. For g : R
2
R
2
and f : R
3
R
3
f(x, y, z) =
_
_
g
1
(x, y)
g
2
(x, y)
0
_
_
curlf =
_
_
0 0
0 0
curl g
_
_
= curl g
For R R 0 and

R
2
(the x-y plane), we then get
_

curl g =
_

_
curl f,
_
_
0
0
1
_
_
_
And so by Stokes theorem
=
_

_
f,

T
_
=
_

_
g,

T
_
Example 10.2. o = (x, y, z)

z = x
2
+ y
2
, z 4 This is then of the form
r(x, y) =
_
_
x
y
x
2
+ y
2
_
_
33
Figure 26: Graph showing a hemispherical bowl, the top should be smooth at the point
of the peaks, but grapher cant draw it correctly.
N(x, y, z) = (2 x
2
y
2
) =
_
_
2x
2y
1
_
_
= o
Is then parameterised by:
(t) =
_
_
2 cos t
2 sin t
4
_
_
, t [0, 2]
A diagram is shown in gure 26 Now let
f :=
_
_
y
x
2
3z
2
_
_
_
S
_
f,

T, =
_
_
2
0
_
_
_
2 sin t
4 cos
2
t
3 16
_
_

_
_
2 sin t
2 cos t
0
_
_
_
= 4
curl f =
_
_
0 0
0 0
2x 1
_
_
_
S
curl f,

N) =
__
_
_
0
0
2x 1
_
_

_
_
2x
2y
1
_
_
dsdt =
__
2x 1dsdt
__
B(0,2)
2x 1dxdy
__
B(0,2)
2xdxdy
__
B(0,2)
1dxdy
34
Figure 27: Graph showing region being lifted to R
3
0 4 = 4 (as we got before!)
_
S
_
curl f,

N
_
=
_
S
_
f,

T
_
Proof of theorem 10.1. We lift Greens theorem from R
2
to R
3
, this is shown in gure
27. Now we parameterise by : [a, b] R
2
. o is parameterised by r
_
(u)
_
=
u [a, b]. Tangent T
S
_
r
_
(u)
_
_

r
s

1
u
+
r
t

2
u

_
S
f,

T)
=
_
b
a
_
f
_
(u)
_
,
r
s

1
u
+
r
t

2
u
_
=
_
b
a
__
f,
r
s
_

f,
r
t
_
_
,

(u)
_
du
_

__
f,
r
s
_

f,
r
t
_
_
,

T

_
(10.1)
Similarly
_
S
_
curl f,

N
_
35
=
_
curl f
_
r(s, t)
_
,
r
s

r
t
_
dsdt
Omitting the middle (Chain rule + hard work)
=
__

curl
_
f,
r
s
_

f,
r
t
_
_
(10.2)
By Greens theorem equations 10.1 and 10.2 are equal. When showing the equality of
10.2 we have to keep track of many (about 48) terms, in later courses we nd dierential
forms useful for this.
Remark 10.4. If f : R
3
R
3
is a gradient then:
_
S
_
curl f,

N
_
=
..
Stokes
_

_
f,

T
_
= 0
_

_
f,

T
_
= V (b) V (a)
by the FTC for gradient vector elds.
curl grad v =
_
_

z
_
_

_
_
v
x
v
y
v
z
_
_
=
_

2
v
yz


2
v
zy
_
= 0
Curl of gradient vector elds is always zero. Similarly f : R
3
R
3
with f = curl v we
get
_

_
f,

N
_
=
_

div f, div curl v = = 0

11 Spherical Coordinates
Skipped, will be a PDF of most of this topic included later.
12 Complex Dierentation
21
The aim of this section is to understand calculus for functions f : C C, and its link
to vector analysis.
Denition 12.1 (Complex Numbers). The complex numbers are dened by:
C =
_
x + iy

x, y R, i
2
= 1
_
Clearly, there is a one-to-one correspondence between C C functions and R
2
R
2
functions:
f : C C (u, v) : R
2
R
2

Where f(x + iy) = u(x, y) + iv(x, y).

36
Re
Im
z
r
x
y

1
Figure 28: Planar representation of a real number z
12.1 Basic properties of Complex Numbers
z = x + iy
z = r cos + ir sin
r =
_
x
2
+ y
2
= [z[
= arctan
_
y
x
_
= arg(z)
Addition
(x
1
+ iy
1
) + (x
2
+ iy
2
) = (x
1
+ x
2
) + i(y
1
+ y
2
)
Multiplication
(x
1
+ iy
1
)(x
2
+ iy
2
) = (x
1
x
2
y1
y
2) + i(y
1
x
2
+ x
1
y
2
)
The meaning of this is clearer in polars, for example if z = r cos + ir sin and w =
s cos + is sin then:
zw = rs cos( + ) + irs sin( + )
Complex conjugation
z = x + iy z = x iy
12.2 Limits
Denition 12.2 (Limits). For z
n
, z C:
z
n
z [z
n
z[ 0
21
The next two sections are done by Matthew Pusey
37
By the denition of [z[:
z
n
z
_
(x
n
x)
2
+ (y
n
y)
2
0 x
n
x and y
n
y
Lemma 12.1. If z
n
z and w
n
w then:
z
n
+ w
n
z + w
z
n
w
n
zw
z
n
z
w

z
w
if w ,= 0
Sketch proof. Exactly as for R, but need to avoid inequalities in C, which make no sense.
Still have that:
[zw[ = [z[[w[
[z + w[ [z[ +[w[
And can dene an open ball:
B(z, ) = z

C : [z z

[
So that:
z
n
z > 0N N such that n N, z
n
B(z, )
12.3 Continuity & Dierentiation
Denition 12.3 (Continuity). A function f : D C for D C is continuous at z D
if: B(z, ) D for some > 0 and:
z
n
z = f(z
n
) f(z)
Note: This must hold for all sequences (z
n
) with z
n
z. We say f is continuous on D
if f is continuous at every z D.
Remark 12.1. f is continuous at z > 0, > 0 such that:
[z
n
z[ < = [f(z
n
) f(z)[ <
Denition 12.4 (Dierentiation). A function f : D C with D C is dierentiable
at z D if
f

(z) = lim
h0
f(z + h) f(z)
h
exists. Note that this means
f

(z) = lim
n
f(z + h
n
) f(z)
h
n
exists for any h
z
0.
38
Example 12.1.
f(z) = z
Its clear that
z
n
z = f(z
n
) f(z)
So f is continuous.
f(z + h) f(z)
h
=
z + h z
h
= 1 1 = f

(z) 1
Example 12.2.
f(z) = z
z
n
z = x
n
x, y
n
y
= x
n
x, y
n
y
= f(z
n
) f(z)
So f is continuous. But it is not dierentiable, since the limit
lim
h0
f(z + h) f(z)
h
= lim
h0

h
h
does not exist. For example, with h
n
=
1
n
it tends to 1 but with h
n
=
i
n
it tends to -1.
Lemma 12.2. Let f, g : D C, with D C, be continuous (and dierentiable) at z.
Then f + g, fg and
f
g
(g ,= 0) are also continuous (and dierentiable).
Let f(x + iy) = u(x, y) + iv(x, y). Then certainly:
f continuous uv continuous
But:
f dierentiable uv dierentiable
does not hold in general.
13 Complex power series
Denition 13.1.

n=0
c
n
converges (to c) if:
S
N
=
N

n=0
c
n
converges (to c).
As in R, we have the Cauchy criterion for a sequence (z
n
):
> 0, N N such that m, n N, [z
m
z
n
[ <
Loosely speaking, a sequence is Cauchy if [z
m
z
n
[ 0 as m, n . It is easy
to show that if (z
n
) is Cauchy its real and imaginary parts are Cauchy, so the sequence
converges in C since the parts must converge in R.
39
Lemma 13.1. If [f
n
(z)[ M
n
z D and

n=0
M
n
< then
f(z) =

n=0
f
n
(z)
converges for all z D. Also, if all the f
n
are continuous then so is f.
Proof. Let S
N
=

N
k=0
f
k
(z). Then, assuming without loss of generality that m > n:
[S
m
(z) S
n
(z)[ =

k=n+1
f
k
(z)

k=n+1
M
k

k=n+1
M
k

0 as n
Theorem 13.2 (Power series). Let (c
n
) be a sequence in C, and dene:
f(z) =

n=0
c
n
z
n
Then there exists some R [0, ] such that f(z) converges if [z[ < R, and doesnt
converge if [z[ > R.
Notes:
1. f may or may not converge when [z[ = R.
2. A similar theorem holds in R and the proof carries over.
3. The theorem can be applied repeatedly.
4. f is C

on B(0, R), with:

k
z
k
f(z) =

n=k
n(n 1) (n k + 1)c
n
z
nk
5. If f(z) converges on B(0, R) then g(z) =

n=0
c
n
(z a)
n
converges on B(a, R).
To calculate R we can use the ratio test.
Lemma 13.3 (Ratio test). If
[z
n+1
[
[z
n
[
L [0, ]
then

n=0
z
n
converges if L < 1 and diverges if L > 1.
40
Sketch proof. Observe that

k=n+2
z
k

k=n+1
[z
k
[
and use the result on R + Cauchy
Example 13.1.
f(z) =

n=0
(3 + i)(2i)
n
. .
c
n
(z + i)
n
[z
n+1
[
[z
n
[
=

(2i)
n+1
(z + i)
n+i
(2i)
n
(z + i)
n

= [2i(z + i)[ = [2i[[z + i[

= 2[z + i[ 2[z + i[ = L
So when [z + i[ <
1
2
, f(z) converges, and when [z + i[ >
1
2
, f(z) doesnt converge. This
gives R =
1
2
.
Example 13.2.
f(z) =

n=1
z
n
n
[z
n+1
[
[z
n
[
=

z
n+1
n
(n + 1)z
n

=
n
n + 1
[z[ [z[ = L
This gives R = 1.
What about when [z[ = 1? f(1) diverges, f(1) converges to log 2. In general this is
a hard problem.
Denition 13.2 (Common power series).
e
z
= exp(z) =

n=0
z
n
n!
cos(z) =

n=0
1
(2n)!
z
2n
(1)
n
cosh(z) =

n=0
1
(2n)!
z
2n
sin(z) =

n=0
1
(2n + 1)!
z
2n+1
(1)
n
sinh(z) =

n=0
1
(2n + 1)!
z
2n+1
Lemma 13.4.
sin(z) =
e
iz
e
iz
2i
cos(z) =
e
iz
+ e
iz
2
sinh(z) =
e
z
e
z
2
cosh(z) =
e
z
+ e
z
2
41
Proof. Use the power series. For example, to prove the one for cos(z):
e
iz
+ e
iz
2
=
1
2

n=0
_
(iz)
n
n!
+
(iz)
n
n!
_
=
1
2

n=0
z
n
((i)
n
+ (i)
n
)
n!
The numerator here is 2i
n
if n is even, and 0 is n is odd. Therefore it equals:

k=0
z
2
k
2k!
(1)
k
= cos(z)
Example 13.3.
sin(iy) =
e
y
e
y
2i
= i
e
y
e
y
2
= i sinh(y)
14 Holomorphic Functions
Let f(x + iy) = u(x, y) + iv(x, y), for h = + i 0
lim
0
u(x + , y) u(x, y)

+ i lim
0
v(x + , y) v(x, y)

=
u
x
+ i
v
x
For h = 0 + i.
f

(z) = (i) lim

0
u(x, y + ) u(x, y)
(i)
+

i lim
0
v(x, y + ) v(x, y)

i
f is only dierentiable if
u
x
+ i
v
x
= i
u
y
+
v
y
Denition 14.1 (The Cauchy Riemann equations). A complex function f(x + iy) =
u(x, y) + iv(x, y) is dierentiable if and only if:
u
x
=
v
y
v
x
=
u
y
Theorem 14.1. Consider f : D C, D C with:
f(x + iy) = u(x, y) + iv(x, y)
1. If f is dierentiable at (x
0
, y
0
) then
u
x
,
u
y
,
v
x
,
v
y
exist at (x
0
, y
0
) and the Cauchy
Riemann equations (denition 14.1) hold at (x
0
, y
0
).
42
2. If
u
x
,
u
y
,
v
x
,
v
y
and are continuous in a small Ball around (x
0
, y
0
) then f is
dierentiable at (x
0
, y
0
) with z
0
= x
0
+ iy
0
with
f

(z
0
) =
u
x
+ i
v
x
=
v
y
i
u
y
Example 14.1 (Example to show the dierence between points 1 and 2). f(z) = z
3
must be dierentiable everywhere.
f(x + iy) = (x + iy)
3
= x
3
+ 3x(iy)
2
+ 3x
2
(iy) + (iy)
3
= x
3
3xy
2
. .
u(x,y)
+i (3yx
2
y
3
)
. .
v(x,y)
u
x
= 3x
2
3y
2
(14.1)
v
y
= 3x
2
3y
2
(14.2)
As you can see equations 14.1 and 14.2 are the same.
u
y
= 6xy (14.3)
plv
x
= 6xy (14.4)
As you can see equations 14.3 and 14.4 are the same.
21/11/06
Example 14.2 (Hard example).
f(x + iy) = x
2
+ iy
2
u
x
= 2x,
v
y
= 2y, (these are in general not equal so the function isnt dierentiable.
v
x
= 0,
u
y
= 0 (these are equal and have to be equal for dierentiability.)
Therefore f is only dierentiable if x = y.
Denition 14.2. If a function f : D C where D C is holomorphic at z
0
D if
f is dierentiable for all z B(z
0
, ) for some > 0. f is holomorphic on D if it is
holomorphic z D.
Remark 14.1. The aim is to apply Vector Analysis to this problem.
43
If f(x + iy) = u + iv is holomorphic then we dene:
f(x, y) =
_
u(x, y)
v(x, y)
_
For R
2
R
2
, then
F(x, y) =
u
x

v
y
=
..
by Cauchy Riemann
0
and
curl F(x, y) =
u
y

v
x
= 0
F
2
x

F
y
22
15 Complex Integration
Theorem 15.1. Consider a parameterised curve [a, b] C
(t) = x(t) + iy(t)

(t) = x

(t) + iy

(t)
Remark 15.1. The Divergence theorem and Greens theorem might be useful here.
Denition 15.1. For F : D C:
_

f =
_
b
a
f
_
(t)
_

(t)dt
=
_
b
a
Re
_
f
_
(t)
_

(t)
_
dt + i
_
b
a
Im
_
f
_
(t)
_

(t)
_
dt
Example 15.1.
f(x + iy) = x
2
+ iy
(t) = t(1 + i) for t [0, 1]
_

f =
_
1
0
t
2
+ it
2
(1 + i)dt
=
_
1
0

t
2
+ it
2
+ it
2

t
2
dt
_
1
0
2it
2
dt
= 2i
_
1
0
t
2
dt
= 2i
_
t
3
3
_
1
0
=
2i
3
22
I dont understand this.
44
Remark 15.2. If and parameterise the same path in the same direction, then if:
_

f =
_

f. If the direction is reversed then
_

f =
_

f.
Example 15.2. Integrate f(z) = z around B(i, 2). Since e
it
= cos t +i sin t, we can use
(t) = 2e
it
+ i to parameterise B(i, 2).
_

f =
_
_
2e
it
+ i
_
2ie
it
dt
_
2
0
(
&
&
i + 2e
it
+

i)2ie
it
dt
=
_
2
0
2e
it
+ 4idt
=
_
2e
it
i
_
2
0
+ 8i
= 8i
_

f =
_
b
a
f
_
(t)
_

(t)dt
_
2
0
e
it
dt =
_
e
it
i
_
t=2
t=0
23
Theorem 15.2 (Fundamental Theorem of Calculus for Complex Integrals). Let f : D
C be holomorphic for D C, : [a, b] C then
_

= f
_
(b)
_
f
_
(a)
_
Proof.
_

f
z
=
_
b
a
f
z
_
(t)
_

t
(t)dt
_
b
a

t
_
f
_
(t)
_
_
dt = f
_
(b)
_
f
_
(a)
_
This last statement is true by applying the real Fundamental theorem of calculus to
Re
_
f((t)
_
and Im
_
f
_
(t)
_
_
Remark 15.3. If f

= 0 and f is over a connected region

24
, then f is constant.
23
This doesnt make much sense tbh
24
As in metric spaces
45
16 Cauchys theorem
For : [a, b] C
(t) = x(t) + iy(t)
f(x + iy) = u(x, y) + iv(x, y)
We get that
_

f =
_
b
a
u
_
x(t), y(t)
_
+ iv
_
x(t), y(t)
__
x

(t) + iy

(t)
_
dt
=
_
b
a
(ux

vy

)dt + i
_
b
a
(uy

+ vx

)dt
=
_
b
a
_
u
v
_

_
x

_
dt + i
_
b
a
_
u
v
_

_
y

_
dt
=
_

_
F,

T
_
+ i
_
?????
_
F,

N
_
where F(x, y) =
_
u(x, y)
v(x, y)
_
Now by using Greens theorem (9.1) and the divergence
theorem (7.1)
Theorem 16.1 (Cauchy). Let a function f : D C, D C be holomorphic and D
a region with a boundary of . If is a parameterisation of then:
_

f = 0
25
Proof.
_

f
..
C
=
_

_
F,

T
_
. .
R
2
+i
_

_
F,

N
_
. .
R
2
_

_
F,

T
_
=
..
Green

curl F =
..
Cauchy-Riemann
0
_
?????
_
F,

N
_
=
..
Divergence

div F =
..
Cauchy-Riemann
0
Remark 16.1. The theorem holds for more general curves so as the curve in gure 29
and in this case
_

f =
_

1
f +
_

2
f = 0
However the curve has to be simple, i.e. it must be possible to contract the curve to a
point, so for example it wouldnt apply to the curve in gure 30, as D
25
This is the main result of the Complex Analysis part if the course
46
Figure 29: A more general curve to which Cauchys theorem applies.
Figure 30: A region and domain which it doesnt apply
47
Figure 31: Complex circle of radius .
Example 16.1.
f(z) =
1
z
, (t) = Re
it
t [0, 2]

f =
_
2
0
1

Re
it
i

Re
it
dt = i
_
2
0
1dt = 2i ,= 0
27/11/06
Proposition 16.2. We have seen that:
_
B(0,R)
1
1 + e
z
dz = 0 Real numbers
Proof.
1 + e
z
= 0 e
z
= 1
This is shown in gure 31. Now we know that [e
z
[ = e
x
and arg(e
z
) = y and e
z
= 1
which implies that x = 0, y = (2n+1), n Z. This means that for a ball of radius less
than , i.e R [0, ) then the value of the integral is zero. This means f is holomorphic
on any complex circle with R < .
17 Cauchy Integral Formula
Let be the boundary of a connected region C with positive orientation.
Remark 17.1 (Idea 1).
26
We can deform without changing the integral
_

f. As we
26
These are really ideas but I using remarks instead
48
Figure 32: Diagram showing the region bounded by a curve , and a second curve
splitting it into two pieces.
can see in gure 32 if there are a few points in the region bounded by gamma which arent
holomorphic we can split the region into two pieces without changing the integral with a
split o region, bounded by the curve . If f is holomprohic on and between and ,
then if is the part of and bounding this new region then
_

f = 0
Also
_

f
_

f =
_
f
_
f
27
_
simple loop
f = 0
Remark 17.2 (Idea 2). If f is holomorphic on and inside except for a nite number
of points z
1
, . . . , z
n
, this is shown in gure 33 and leads to what is shown in 34.
_

f =
n

i=1
_
B(z
i
,)
f
> 0 small enough so that the balls dont overlap.
27
check original notes
49
Figure 33: Diagram showing the integral inside a curve
Figure 34: Integrals around the non homomorphic points.
50
Remark 17.3 (Idea 3). If:
f(z) =

n=0
a
n
z
n
converges on B(0, R), R > 0, then:
_
B(0,)
f(z)
z
dz =
_
B(0,)
a
0
z
+ a
1
+ a
2
z +
. .
Holomorphic
dz
= a
0
2i
..
As in example
+ 0
..
Cauchy
Similarly
_
B(0,)
f(z)
z
n
dz =
_
a
0
z
n
+
. .
zero as primitive by FTC
+
a
n1
z
+ a
n
+ a
n+1
z +
. .
Holomorphic

z
z
1k
= (1 k)z
k
(except for k = 1)

_
a
0
z
n
+ +
a
n2
z
2
dz =
..
by the FTC
0

_
B(0,R)
f(z)
z
n
= 0
..
FTC
+a
n1
2i + 0
..
Cauchy
28/11/06
Theorem 17.1 (Cauchy Integral Formula). Let be the boundary of a connected region.
Let f be holomorphic on and inside .
_

f(z)
z z
0
= 2if(z
0
) z
0
inside
Remark 17.4. Holomorphic functions are special, the values of f along completely
determine the values of f inside .
Lemma 17.2.

_
b
a
x(t) + iy(t)dt

_
b
a
[x(t) + y(t)[dt
Proof. If:
_
b
a
x(t) + iy(t)dt R

_
b
a
x(t) + iy(t)

dt =

_
b
a
x(t)+

dt
..
by Analysis 3
_
b
a
[x(t)[dt
_
b
a
[x(t) + iy(t)[dt
51
Generally
_
b
a
x(t) + iy(t)dt = re
i
forr, R, [0, 2]
r = e
i
_
b
a
x(t) + iy(t)dt =
_
b
a
e
i
_
x(t) + iy(t)
_
dt
=

_
b
a
e
i
_
x(t) + iy(t)
_
dt

_
b
a

e
i

x(t) + iy(t)

dt =
_
b
a
1

x(t) + iy(t)

dt
Remark 17.5. If [f(z) M on a curve then:

_
b
a
_
f
_
(t)
_

(t)
_

_
b
a
< M[

(t)[dt = M
_
b
a
[

(t)[dt
Therefore this means that M is the length of .
Proof of theorem 17.1. By deforming
_

f(z)
z z
0
=
_
B(z
0
,)
f(z)
z z
0
dz
=
_
B(z
0
,)
f(z
0
)
z + z
0
dz
. .
(a)
+
_
B(z
0
,)
f(z) f(z
0
)
z z
0
. .
(b)
dz (17.1)
Then part a of equation 17.1 is equal to:
(a) = f(z
0
)
_
B(0,)
1
z
dz = f(z
0
)2i
Then the absolute value of part b of equation 17.1 is:
[(b)[ max

f(z) f(z
0
)
z z
0

2
_
[f

(z
0
)[ + 1
_
2 (17.2)
then for small epsilon the right hand side of equation 17.2 tends to zero as 0, therefore
part b of equation 17.1 tends to zero.
28
Example 17.1.
_
sin z
z i
dz =
_
_
_
0 If i is outside
2i sin i If i is inside
? If i is on the curve though we are lost.
This is shown in gure 35
29
28
Not convinced complete.
29
Check complete.
52
Figure 35: Diagram showing the three possible cases for the curve
53
Figure 36: Diagram showing the integral of split into four pieces.
18 Real Integrals
Aim to nd the integral of:
_
+

sin x
x
dx (18.1)
As
sin x
x
1 as x 0 (by LHopital as
cos x
1
1) So the function is dened everywhere.
A diagram of the curve which we will use to integrate equation 18.1 is shown in gure
36. The strategy for solving this integral is to Integrate f along by Cauchys theorem
(16.1). As we known that f is holomorphic in the region enclosed by in gure 36 then
we know that:
_

1
f +
_

3
f +
_

2
f +
_

4
f
. .
what we want
= 0 (18.2)
30/11/06 To solve equation 18.1 we nd f over
1
and
3
.
f(z) =
e
iz
z
Equation 18.2 oviously then leads to:
_

2
f +
_

4
f =
_

R
e
ix
x
dx +
_
R

e
ix
x
dx
54
=
_

R
cos x + i sin x
x
dx +
_
R

cos x + i sin x
x
dx
As since cos is an even function, and x is odd.
cos(x)
x
=
cos x
x

0
i
_
R
R
sin x
x
dx
From Cauchys theorem as we showed in equation 18.2
_

1
f +
_

2
f + int

3
f +
_

4
f = 0
i
_
R
R
sin x
x
dx = lim
0
__

1
f +
_

3
f
_
So:
_

3
f =
_

3
1
z
dz
. .
(a)
+
_

3
e
iz
1
z
dz
. .
(b)
(18.3)
Now (a) = i by question 2.2 of sheet 4.
[(b)[ max

e
iz
1
z

Length of
3
C where C is a constant. Which tends to zero as 0.
e
iz
1
z
= i
e
iz
e
0
iz
ie

(0)
30
This
_

3
f i as 0.
Now we claim that
_

1
0 as R This is because
1
(t) = Re
it
, t [0, ]
= R(cos t + i sin t)

_

0
e
iR(cos t+i sin t)

Re
it
i

Re
it

_

0

e
iR(cos t+i sin t)

dt =
_

0
e
Rsin t
dt (18.4)
So as sin t is positive over 0, and it is symmetric along

2
. Equation 18.4 then becomes:
= 2
_
2
0
e
Rsin t
dt (18.5)
As on
_
0,

2

sin t
t

2
So therefore equation 18.5 then is less than
(18.5) 2
_
2
0
e
Rt
/2
= 2
_
2
0
e
2Rt

30
not totally sure why last point holds
55
=

2R

2
_
e

2R

e
2R0

_
=

R
_
e
R
1
_
=

R
_
1 e
R
_
0 as R By Cauchys Theorem as R
i
_
+

sin x
x
dx
. .

2
,
4
+ 0
..

1
i
..

3
= 0
19 Power Series for holomorphic functions
Theorem 19.1. Suppose f : D C and D C is holomorphic, then:
f =

r=0
a
r
z
r
where

r=0
a
r
is a convergent power series on any ball B(a, R) D
Remark 19.1. If f is holomorphic implies that f is a power series so f is C

Theorem 19.2.
_

f(z)
z
dz = 2f(0)
This is assuming that f : D C is holomorphic, B(a, R) D (An image of this is
shown in gure 37). Together they imply that f is equal to a power series on B(a, R)
Proof. We can assume (without loss of generality) that a = 0 by shifting everything to
the origin. Then for every 0 < r < R
f(0) =
1
2i
_
B(0,r)
f(z)
z
dz z
0
B(0, R) (19.1)
If z
0
B(0, r) then:

z
0
z

< 1
1
z z
0
=
1
z
1
1
z
0
z
=
1
z

n=0
_
z
0
z
_
n
f(z
0
) =
1
2i
_
B(0,R)
f(z)
1
z

n=0
_
z
0
z
_
n
=

n=0
_
1
2i
_
B(0,r)
f(z)
1
z
n+1
dz
_
z
n
0
Thus f equals a power series on B(0, r) for all r < R, this implies the radius of convergence
is greater than or equal to R.
56
Figure 37: A region containing a ball
Remark 19.2. Let f be holomorphic and g D. By the theorem, f(z) =

n=0
c
n
(za)
n
f

(z) =

n=1
nc
n
(z a)
n1

k
z
k
f(z) =

n=k
n(n 1) (n k)c
n
(z a)
nk

k
z
k
f(a) = k!c
k
(0
0
+ 0
1
+ )
Thus we get:
f(z) =

n=0

n
z
n
f(a0
(z a)
n
n!
Remark 19.3 (Taylors formula).
f(x) = f(a) + f

(a)(x n) +
1
2
f

(a)(x a)
2
+ (19.2)
Corollary 19.3. If is a simple loop, f is holomorphic on and inside then:
_

f(z)
(z a)
n+1
dz = 2ic
n
=
2i
n!

n
z
n
f(a)
57
Figure 38: As you can see the black function is C
1
, the blue function is the dierential
of it.
f(z)
z
n+1
=
c
0
z
n+1
+
. .
=0 by FTC
+
c
n
z
+ c
n+1
+
. .
=0 by Cauchy
(19.3)
For n = 0, this is the Cauchy Integral formula (CIF).
Denition 19.1. A function is called Analytic if it can be expanded into a power se-
ries everywhere, around every point. We have just seen that analytic holomorphic in
complex analysis.
Lecture 28
31
In C f is C
1
f is C

f is C

f can be expanded as a power

series. In R if f is C
1
it doesnt imply that f is C

. e.g:
f(x) :=
_
x
3
x > 0
x
2
is C
1
as you can see in gure 38 In R if f is C

this doesnt imply that f can be expanded

as a power series.
31
As (well it seems like usual), I have borrowed Jack Heals notes for this lecture, I was very very tired
so I missed it
58
Figure 39: f(x) = e

1
x
, x > 0 and f(x) = 0, x 0.
Figure 40: A diagram shown the sequence z
i
tending to a point z

.
Example 19.1.
f(x) =
_
0 x 0
e

1
x
x > 0
f is drawn in gure 39. There is no power series for this function because

n
f
x
n
(0) = 0
32
Proposition 19.4. Let (z
n
) D be a sequence, such that lim
n
z
n
= z

D If f and
g are holomorphic, and f(z
n
) = g(z
n
) for every n then f = g. This is shown in gure
40.
Proof. Let h = f g, since h is holomorphic, one has:
h(z) = a
0
+ a
1
(z z

) + a
2
(z z

)
2
+
Take limits as n , this implies that a
0
= 0. If you then divide by z z

then:
h(z)
z z

= a
1
+ a
2
(z z

) +
32
As x 0, f(x) 0 faster than any polynomial
59
If you evaluate at z = z
n
and take limits as n we get a
1
= 0. By induction, you
then get a
k
= 0 k.
Remark 19.4. If (z
n
) doesnt converge the above doesnt hold. e.g. f(x) = sin x, g(x) = 0
[(z
n
) = kk Z] e.g. f(x) = sin
2
x+cos
2
x g(x) = 1 One has sin
2
x+cos
2
x = 1 x C.
Theorem 19.5. Let f : C C be holomorphic, if their exists M such that [f(z)[ < M
z C then f is constant.
Proof. We have:
f(z) =

n=0
a
n
z
n
with
a
n
=
1
2i
_
B(0,R)
f(z)
z
n+1
dz
[a
n
[ =
1
2

_
B(0,R)
f(z)
z
n+1
dz

1
2
_
B(0,R)

f(z)
z
n+1

dz

1
2
_
B(0,R)
M
R
n+1
dz
[a
n
[
1
2
M
R
n+

1
2

R =
M
R
n
This holds for every R > 0, therefore a
n
= 0 for n 1. This means that
f(z) =

n=0
a
n
z
n
= a
0
(i.e. f(z) is constant.)
Theorem 19.6. Every non-constant polynomial P on C has at least one root. ( z C
s.t P(z) = 0)
Proof. Suppose f has no root. Then f(z) =
1
P(z)
is holomorphic in C. P = a
0
+ a
1
+
+ a
n
z
n
There exists an R > 0 and C > 0 s.t. [P(t)[ C[z[
n
for [z[ > R. Therefore
[f(z)[
1
CR
n
for [z[ > R. Since P has no root, f has no C pole (i.e. 1/0 is undened)[3],
this means that their exists M s.t [f(z)[ < M for [z[ < R, by the previous theorem f
must be constant, which implies that P is also constant. If P is a polynomial of degree
n and P(a) = 0, then P(z) = (z a)

P(z) where

P is a polynomial of degree n 1.
20 Real Sums
The aim is to nd the solution of:

k=
(20.1)
60
Figure 41: A diagram of the box
N
61
Idea: calculate
_

N
f(z)
cos(z)
sin(z)
dz
f is holomorphic except at z
1
, z
2
, . . . , z
M
.
_

N
f(z)
cos(z)
sin(z)
dz (20.2)
=
..
CIF
M

k=1
_
B(z
N
,)
+
N

k=N
2
&
i
f(k)
&

This allows us to compute

N
k=N
f(k).
Remark 20.1 (Details).
sin(z) =
e
iz
e
iz
2i
= 0
e
iz
e
iz
z Z
Using:
sin(z) = z
z
3
3!
+
z
5
5!

As we know that sin(a + b) = sin a cos b + cos a sin b
sin(z) = sin
_
(z k) + k
_
= sin
_
(z k)
_
cos k + cos
_
(z k)
_
sin(k)
. .
=0
= (z k)
_
1

2
(z k)
2
3!
+
_
. .
=g(z)z as zk
cos(k)

_
B(k,)
f(z)
cos(z)
sin(z)
=
_
1
z k
f(z)
cos
_
k
_
g(z) cos(pik)
=
..
CIF
2if(k)
cos(k)
cos(k)
= 2if(k)
Example 20.1.
f(z) =
1
z
2
+ 1
=
1
(z i)(z + i)
At i :
_
B(i,)
f(z)
cos(z)
sin(z)
=
..
CIF
2i
1
i + i
cos(i)
sin(i)
=
cos(i)
sin(i)
At i : 2i
1
i i
cos(i)
sin(i)
=
cos(i)
sin(i)
,
cos(z)
sin(z)
has poles at
1
(zk)
for z Z. The integral along the box
N
(as shown if gure
41) as N .

N
f(z)
cos(z)
sin(z)
dz

max
z
N
[ [ (Length of
N
)
C
N
2
4(2N + 1) 0
62
With N
0 = 2
cos(i)
sin(i)
+ 2i

k=
1
k
2
+ 1

k=
1
k
2
+ 1
=

i
cos(i)
sin(i)

k=
1
k
2
+ 1
=
cosh
sinh
= tanh() 3.15
References
[1] Cauchy Schwarz inequality on Wikipedia
http://en.wikipedia.org/wiki/Cauchy-Schwarz_inequality
[2] Kissing Number Problem
www.lix.polytechnique.fr/
~
liberti/kissing-ctw.ps.gz
[3] Pole (Complex Analysis)
http://en.wikipedia.org/wiki/Pole_(complex_analysis)
63