4 Cauchys Integral Theorem

Lecture 9
If judged by its applications, Cauchys integral theorem must be a one of the most
important theorems in all of mathematics.
Theorem 4.1 (Cauchys Integral Theorem: Version I). Let D be a domain in C, f :
D Ca holomorphic function, : [a, b] Da piecewise continuously differentiable,
simple, closed path (contour) in D such that the interior of lies in D. Then

f(z)dz = 0.
Theorem 4.2 (Jordan Curve Theorem for a simple closed path). Let : [a, b]
C be a continuous, simple, closed path. Then C \ ([a, b]) is the disjoint union of two
open connected sets, Int() and Ext() where Int() (the interior of ) is bounded
and Ext() is unbounded
The proof of the Jordan Curve Theorem is beyond the scope of this course.
Theorem 4.3 (Elementary version of Cauchys Integral Theorem). Here we as-
sume that the derivative f

is continuous and deduce Cauchys Integral Theorem

Greens Theorem, proved in the Moderations course.
Proof. Put z = x+iy, f(x+iy) = u(x, y) +iv(x, y), x, y, u, v real. We use Greens
theorem from the Moderations syllabus.
Theorem 4.4 (Greens Theorem).

Pdx + Qdy =

Int()

Q
x

P
y

dxdy.
Using Greens Theorem,

f(z)dz =

(u + iv)(x, y)dx + idy

=

udx vdy + i

vdx + udy
=

Int()

v
x

u
y

dxdy +

Int

u
x

v
y

dxdy
= 0
by the Cauchy-Riemann equations.
27
To prove the theorem without assuming that f

is continuous has a number

of different steps. The full proof is not examinable though some of the steps
include ideas used in many places in this course. We include here a sketch of
the proof and will try to make clear as we go along what is worth retaining for
examination purposes.
First we prove a special case where is the boundary of a triangle. This
result is not examinable.
Lemma 4.1. Let be a triangle lying in C and let L be the sum of the lengths of its
three sides. If a, b then |a b| < L.
Proof. Suppose that the vertices are z
1
, z
2
, z
3
. Let R = max{|z
1
z
2
|, |z
1
z
3
|} =
|z
1
z
3
|, say. Then B
R
(z
1
). If a, b B
R
(z
1
) then
|a b| |a z
1
| +|z
1
b| 2R and |z
1
z
2
| +|z
2
z
3
| |z
1
z
3
| = R.
Hence
|z
1
z
2
| +|z
2
z
3
| +|z
3
z
1
| 2R
so L 2R. Therefore |a b| 2R L.
Theorem 4.5 (Goursats Lemma). Let D be a domain in C and f : D C a
holomorphic function. Then for each triangle contained in D,

f(z)dz = 0.
Proof. (Not examinable) Let L be the sum of the lengths of the sides of . Join
the midpoints of the sides of the triangle to form 4 triangles
1
,
2
,
3
,
4
and note that
1
,
2
,
3
,
4
are all similar to same angles and sides half
the length of the corresponding sides in . Let
I =

f(z)dz

k=1

k
f(z)dz

,
because the integrals along the sides inside are each transversed twice, once
in each direction and so cancel out. Therefore
I
4

k=1

k
f(z)dz

.
There exists some

=
k
such that

k
f(z)dz

1
4
I.
28
Join the midpoints of the sides of

to obtain

such that

f(z)dz

1
4

f(z)dz

1
4
2

f(z)dz

=
1
4
2
I.
Continuing in this way we obtain

where

(k)
f(z)dz

1
4
k
I.
By Cantors Intersection Theorem,

k=1

(k)
=
and if a, b
(k)
then |a b|
1
2
k
L. Hence

k=1

(k)
contains a single point, a
say. The function f is holomorphic at a. Put
g(z) =
f(z) f(a)
z a
f

(a),
when z = a and let g(a) = 0. Then g is continuous because g(z) 0 as z a.
Consider
g(z)(z a) = f(z) f(a) f

(a)(z a)
so
f(z) = g(z)(z a) + f(a) + f

(a)(z a).
The following integrals vanish:

(k)
f(a)dz =

(k)
(f(a)z)

dz = 0
because
(k)
is a closed path;

(k)
f

(a)(z a)dz =

(k)

(a)
(z a)
2
2

dz = 0
for the same reason. Therefore,

(k)
f(z)dz =

(k)
g(z)(z a)dz.
Let > 0 and choose > 0 such that |g(z)| < if z B

(a). If
(k)
B

(a),
then

(k)
f(z)dz

(k)
g(z)(z a)dz

L
2
k
L
2
k
,
29
by the standard estimate for integrals and the lemma above. But
|I| =

f(z)dz

4
k

(k)
f(z)dz

4
k
L
2
4
k

and we can take arbitrarily small. Thus |I| = 0, as required.
Lecture 10
Denition 4.1. A set S R
n
is said to be star-shaped if there exists a S such
that for each w S
[a, w] := {(1 t)a + tw : t [0, 1]},
the line segment joining a and w, is contained in S; the set is convex if for any
w
1
, w
2
S, [w
1
, w
2
] S.
Theorem 4.6. Let D be a star-shaped domain in C and f : D C a continuous
function which satises the property

f(z)dz = 0
for any triangle D. (This means the boundary, , and the triangle lie in D.)
Then there exists a holomorphic function F : D C such that F

= f.
(i) Note that if f is holomorphic then, by Goursats Lemma, it satises the
required condition in the theorem.
(ii) Fromthis result it follows that

f(z)dz = 0 for each p.w.c.d. closed path

in D. The function F is called the primitive or pre-derivative of f.)
Proof. (This proof is examinable.) Choose a D such that [a, w] D for all
w D. Dene
F(w) =

[a,w]
f(z)dz,
where this means

[a,w]
f(z)dz =

f(z)dz,
: [0, 1] D, (t) = (1t)a+tw. Let w D and > 0 be such that B

(w) D.
Then
F(w + ) F(w)

=
1

[a,w+]
f(z)dz

[a,w]
f(z)dz.
30
Let T be the triangle with vertices a, w, w + . Then

T
f(z)dz = 0
by assumption, that is,

[a,w]
f(z)dz +

[w,w+]
f(z)dz +

[w+,a]
f(z)dz = 0.
Therefore,

[a,w+]
f(z)dz

[a,w]
f(z)dz =

[w,w+]
f(z)dz.
The function f is continuous and so

[w,w+]
f(z)dz =

[w,w+]
(f(w) + (f(z) f(w))dz
= f(w) +

[w,w+]
(f(w) f(z))dz.
Choose > 0 and 0 < < such that |f(w) f(z)| < if |w z| < . Then if
|| < ,

[w,w+]
(f(z) f(w)dz

||
by the standard estimate for integrals. Therefore,

[w+w+]
f(z)dz f(w)

[w,w+]
(f(z) f(w))dz

.
and it follows that
1

[w,w+]
f(z)dz f(w),
as 0.
By Goursats Lemma, a holomorphic function satises the condition of the
theorem above and we have now proved Cauchys Integral Theorem for star-
shaped domains. It remains to extend it to general domains. For this we must
content ourselves, in this course, with a rough sketch with many details miss-
ing. It is hoped that this will give us sufcient feeling for what is involved
to have condence in the result and to be happy applying it. The content of
the sketch below is not examinable. We have : [a, b] D a p.w.c.d simple
closed path with its interior contained in D and f : D C a holomorphic
function. Then there exist points a = t
1
< t
2
< t
3
< < t
k+1
= b in the
interval [a, b] and
i
> 0 such that ([t
i
, t
i+1
]) B

(
t
i
+t
i+1
2
)

D. Let
i
be
31
the closed path given by |
[t
i
,t
i+1
]
followed by the line segment [(t
i+1
), (t
i
)].
Then
i
is a p.w.c.d. closed path contained in convex ball B

(
t
i
+t
i+1
2
)

and
hence

i
f(z)dz = 0.
Thus

f(z)dz =
k

i=1

|
[t
i
,t
i+1
]
f(z)dz =
k

i=1

[(t
i
),(t
i+1
)]
f(z)dz,
an integral along a polygonal path

. We can chose the t

i
such that the polyg-
onal path is a simple closed path. If the polygon is not convex than we can cut
it into a number of convex polygons each contained in the interior of

. This is
done as follows: followthat path

around and at each t

i
for which the interior
angle exceeds extend the interval [(t
i1
), (t
i
)] until it cuts the polygon or
one of the extended intervals. (Demonstrate this on the whiteboard.) Thus the
polygon given by

is divided into a number of convex polygons each with

its interior in D. For each convex polygon, chose a vertex and join it by a line
segment to the other vertices, thus cutting the convex polygon into triangles.
Finally, we have the polygon

cut into a number of triangles. Integrating f(z)

along each of the triangles we nd with have integrated around

and along
each edge of a triangle not lying along

twice, once in each direction. The

integrals along edges not lying along

all cancel out. The integral of the holo-

morphic function f(z) around each triangle contributes 0 by Goursats lemma.
Finally the integral around

and hence is 0 and the sketch of the proof is

complete.
Example 4.1. Let 0 < R
1
< R
2
and let
D = {z C : R
1
< |z| < R
2
},
an annulus. Let R
1
< r
1
< r
2
< R
2
and
r
i
(t) = r
i
e
it
, t [0, 2]. Let
1
be
a path made up of the line segment [r
1
, r
2
] then anticlockwise along the semicircle

r
2
|
[0,]
, followed by the line segment [r
2
, r
1
] followed by going backwards (clock-
wise) around semicircle
r
1
|
[0,]
see illustration on the board. Let
2
be the path
made up to following the line segment [r
2
, r
1
], then following the path
r
1
|
[,2]
back-
wards (clockwise), followed by the line segment [r
1
, r
2
] followed by the semicircle
(anticlockwise)
r
2
|
[,2]
. Then both
1
and
2
are p.w.c.d simple closed paths with
their interiors in D and so

1
f(z)dz =

2
f(z)dz = 0,
by Cauchys Integral Theorem.
32
Then

1
+

f(z)dz =

r
1
f(z)dz

r
2
f(z)dz
because the integrals along the straight line segments are each transversed twice, once
in each direction and their contributions to the integrals cancel out. Therefore,

r
1
f(z)dz =

r
2
f(z)dz.
This result also follows from the deformation version of Cauchys Integral Theorem
for closed paths.
Lecture 11
Theorem 4.7 (A deformation version of Cauchys Integral Theorem). Let D de
a domain in C, f : D C a holomorphic function,
0
,
1
: [a, b] D p.w.c.d. closed
paths such that
0
can be continuously deformed through p.w.c.d closed paths in D to

1
. Then

0
f(z)dz =

1
f(z)dz/
Denition 4.2. With
0
,
1
as above, the closed path
0
can be continuously deformed
through p.w.c.d closed paths to
1
if there exists a continuous function
H : [0, 1] [a, b] D
such that
(i) H(0, t) =
0
(t), H(1.t) =
1
(t) for all t [a, b];
(ii) H(s, a) = H(s, b) for all s [0, 1];
(iii) for each s [0, 1],
s
(t) := t H(s, t) is a p.w.c.d. closed path in D.
Example 4.2.

|z|=3
1
z
dz =

|z1|=2
1
z
dz,
as there is a deformation between the circles given by
H(s, t) = (1 s)3e
it
+ s(1 + 2e
it
).
Theorem 4.8 (Another deformation version of Cauchys Integral Theorem).
Let D be a domain in C, f : D C a holomorphic function,
0
,
1
: [a, b] D
p.w.c.d. paths such that
0
(a) =
1
(a),
0
(b) =
2
(b) (same endpoints). If
0
can be
33
continuously deformed to
1
within D through p.w.c.d paths
s
, s [0, 1] such that

s
(a) =
0
(a) and
s
(b) =
0
(b) for all s [0, 1] then

0
f(z)dz =

1
f(z)dz.
Here the deformation has the form of a continuous function
H : [0, 1] [a, b] D
such that
s
(t) = H(s, t).
Example 4.3. Let
0
,
1
: [0, 1] C be given by
0
(t) = e
it
,
1
(t) = e
it
. Let
H(s, t) = (1 s)
0
(t) + s
1
(t) =
s
(t). Then

0
z
n
dz =

1
z
n
dz
for all n N. But note that

0
1
z
dz =

1
0
1
e
it
ie
it
dt = i;

1
1
z
dz =

1
0
1
e
it
(1)ie
it
dt = i.
The integrand is not holomorphic at 0 and
0
can not be continuously deformed to
1
within C \ {0}. In the deformation given above,
1
2
(
1
2
) =
1
2
i +
1
2
(i) = 0.
34