Beruflich Dokumente
Kultur Dokumente
Josh Tobin
27th May 2009
1 Complex Functions
2 Analysis in C
In this section we dene the usual concepts of limits and continuity, but with respect to
complex numbers.
Denition (convergence) :
If (z
n
) is a sequence of complex numbers, then we say that (z
n
) converges to z
0
if for all
> 0 there exists N such that
[z
n
z
0
[ < n > N
Denition (convergence to ) :
If (z
n
) is a sequence of complex numbers, then we say that (z
n
) converges to if for all C
there exists N such that
[z
n
[ > C n > N
(z
n
) converges to if (z
n
) converges to .
Example (limit calculation):
Calculate lim
n
z
n
.
Solution.
Let z = re
i
. Then z
n
= r
n
e
in
.
[z
n
[ = [r
n
[ = r
n
lim
n
z
n
= 0 if r < 1, lim
n
z
n
= 1 if r = 1 or lim
n
z
n
= if r > 1.
Lemma:
Let z
n
= x
n
+ iy
n
. Then (z
n
) converges to z
0
= x
0
+ iy
0
i (x
n
) converges to x
0
and y
n
converges to y
0
.
Proof.
(x
n
x
0
)
2
+ (y
n
y
0
)
2
0
0 [x
n
x
0
[ [z
n
z
0
[ 0
[x
n
x
0
[ 0
Similarly we have [y
n
y
0
[ 0.
Suppose (x
n
) and (y
n
) converge to x
0
and y
0
respectively. Then:
[x
n
x
0
[ 0, [y
n
y
0
[ 0
(x
n
x
0
)
2
+ (y
n
y
0
)
2
0
[z
n
z
0
[ 0
z
n
z
0
Lemma:
If z
n
converges to z
0
and w
n
converges to w
0
. Then:
(i) (z
n
+ w
n
) converges to z
0
+ w
0
.
(ii) (z
n
w
n
) converges to z
0
w
0
.
(iii) If w
0
,= 0 then
z
n
w
n
converges to
z
0
w
0
.
Proof.
(i) We have:
[z
n
+ w
n
(z
0
+ w
0
)[ = [(z
n
z
0
) + (w
n
w
0
)[
[z
n
z
0
[ +[w
n
w
0
[ 0
So z
n
+ w
n
z
0
+ w
0
.
(ii) We have:
[z
n
w
n
z
0
w
0
[ = [z
n
w
n
+ z
n
w
0
z
n
w
0
z
0
w
0
[
= [z
n
(w
n
w
0
) + w
0
(z
n
z
0
)[
[z
n
[.[w
n
w
0
[ +[w
0
[.[z
n
z
0
[ 0
So z
n
w
n
z
0
w
0
.
2
(iii) Well show that if w
0
,= 0 then
1
w
n
1
w
0
1
w
0
[ = [
c w
n
cw
n
[
[c w
n
[
[cw
n
[
[w
n
c[
[c[
.
1
[w
n
[
0
So
1
w
n
1
w
0
x
2
+ y
2
which is continuous as a composition.
Denition (uniform convergence) :
A sequence of functions f
n
(x), x A converges uniformly to f(x) if for all > 0 there
exists N such that for all n > N, for all x A:
[f
n
(x) f(x)[ <
Lemma:
f
n
(x) converges uniformly to f(x) i sup
x
[f
n
(x) f(x)[ 0.
Proof.
Assume f
n
(x) converges uniformly to f(x) then for all > 0 there exists N such that
for all n N, for all x, [f
n
(x) f(x)[ < .
sup
x
[f
n
(x) f(x)[
sup
x
[f
n
(x) f(x)[ 0
If sup
x
[f
n
(x) f(x)[ 0 then for all > 0 there exists N such that for all n N.
sup
x
[f
n
(x) f(x)[ <
[f
n
(x) f(x)[ < , for all x.
6
There are possibly some notes missing from sometime between the 17
th
and 23
rd
of February.
Theorem (preimage definition of continuity):
Let f : A C C. Then f is continuous if and only if f
1
(U) is open in A for all open
U C.
Proof.
Assume f is continuous and U is open. Take x f
1
(U). U is open so there exists
> 0 such that the ball of radius around f(x) is contained in U. f is continuous so
there exists a > 0 such that every point in the ball of radius around x maps into
the ball of radius around f(x). Thus f
1
(U) is open.
Assume that the preimage of every open set is open. Take any point x, and any
> 0. Let B be the ball of radius around x. Then f
1
(B) is open. So there exists
> 0 such that the open ball B
of radius around f
1
(x) is contained in B. Thus:
[x x
[ < x
f(x
) B [f(x
) f(x)[ <
That is, f is continuous.
7
Denition (connected) :
A set A is connected if there does not exist non-empty open sets B, C such that A = B C
and B C = .
Theorem (continuous/connected):
If A is connected and f : A C is continuous then f(A) is connected.
Proof.
Assume to the contrary. So f(A) =
B
C with
B,
C non-empty, open and disjoint. Since
f is continuous f
1
(
B) = B, f
1
(
C) = C are non-empty and open in A. Then we have:
B C = f
1
(
B
C) = A
And:
B C = f(
B
C) =
Which contradicts A being connected.
Theorem (continuous/compact):
If K is compact and f : K C is continuous, then f(K) is compact.
Proof.
Choose a sequence w
n
f(K). Then w
n
= f(z
n
). (z
n
) has a convergent subsequence (z
n
k
).
Then (w
n
k
) = (f(z
n
k
)) is a convergent subsequence. Therefore f(K) is compact.
8
Corollary:
If K is compact and f is continuous on K, then f(K) is closed and bounded.
Proof.
This follows immediately from the preceeding two theorems.
Corollary:
If K C is compact, and f : K R is continuous, then f attains a maximum/minimum
at some point z
max
, z
min
.
Proof.
f(K) R is compact, so is closed and bounded. So y
max
= sup f(K), y
min
= inf f(K).
f(K) is closed so y
max
, y
min
f(K). So there exist z
max
, z
min
K such that y
max
= f(z
max
)
and y
min
= f(z
min
).
m=0
z
m
.
Theorem (n
th
term test):
If
n=0
a
n
converges then a
n
0 as n .
Proof.
Proof excluded.
n=0
b
n
converges then
n=0
a
n
converges.
Proof.
Proof excluded.
n=0
n
(z) converges uniformly if the sequence of partial sums f
n
(z) =
n
m=0
m
(z) converges
uniformly.
9
Theorem (Weierstrass comparison test):
Assume that [
n
(z)[ b
n
for all z A and
n=0
b
n
converges. Then
n=0
n
(z) converges
uniformly.
Proof.
Proof excluded.
n=0
a
n
(z z
0
)
n
is a power series centred at z
0
.
Theorem (Abels lemma):
If
n=0
a
n
z
n
is convergent at z = w
0
, then it converges absolutely at every z, [z[ < [w
0
[ and
uniformly on every disc z : [z[ r, r < [w
0
[.
Proof.
Proof exluded.
Corollary:
If R is the radius of convergence of
n=0
a
n
z
n
then
n=0
a
n
z
n
(i) converges absolutely at every z where [z[ < R.
(ii) converges uniformly on every disc [z[ R where r < R.
(iii) diverges at every z where [z[ > R.
Proof.
(i) R = sup[z[ :
n=0
a
n
z
n
converges . [z[ < R so there exists w
0
such that [z[ < [w
0
[
and
n=0
a
n
w
n
0
converges. Then by Abels Lemma,
n=0
a
n
z
n
converges.
(ii) Take [z[ r < R, so there exists w
0
with r < [w
0
[ < R. So
n=0
a
n
w
n
0
converges. By
Abels Lemma, the result follows.
10
(iii) Assume to the contrary. Then
n=0
a
n
z
n
converges, with [z[ > R. This contradicts the
denition of R.
3 Holomorphic Functions
Denition (C-differentiable) :
f(z) is C-dierentiable if lim
zz
0
f(z) f(z
0
)
z z
0
exists.
Example (C-differentiable function):
Is f(z) = z C-dierentiable?
Solution.
f(z) f(z
0
)
z z
0
=
z z
0
z z
0
= 1.
lim
zz
0
f(z) f(z
0
)
z z
0
exists for all z
0
.
So f(z) = z is C-dierentiable.
Example (non-C-differentiable function):
Is f(z) = z C-dierentiable?
Solution.
Fix z
0
, and let z = z
0
+ re
i
. z z
0
re
i
0 r 0. So:
lim
r0
z z
0
z z
0
= lim
r0
z
0
+
re
i
z
0
re
i
= lim
r0
re
i
re
i
= e
2i
So the limit depends on the direction, so the derivative does not exist.
Theorem (properties of C-differentiability):
Assume f, g are C-dierentiable at z
0
. Then:
(i) f + g is C-dierentiable at z
0
.
(ii) f g is C-dierentiable at z
0
.
(iii) fg is C-dierentiable at z
0
.
11
(iv) if g(z
0
) ,= 0 then
f
g
is C-dierentiable at z
0
.
(v) If f is C-dierentiable at z
0
and g is C-dierentiable at f(z
0
), then g f is
C-dierentiable at z
0
and (g f)
(w
0
) = g
(f(z
0
))f
(z
0
).
(vi) If f is C-dierentiable at z
0
, with f(z
0
) = w
0
, f
1
is continuous at w
0
and f
(z
0
) ,= 0,
then f
1
is C-dierentiable at w
0
and (f
1
)
(w
0
) =
1
f
(z
0
)
.
(vii) If f is C-dierentiable at z
0
then f is continuous at z
0
.
Proof.
Proof excluded.
(z
0
) ,= 0, so by the preceeding theorem g is C-dierentiable.
Lemma (Cauchy-Riemann equations lemma 1):
If f = u + iv is C-dierentiable at z
0
, then it satises the Cauchy-Riemann equations:
u
y
= v
x
u
x
= v
y
Or equivalently (at z
0
)
f
y
= if
x
Proof.
We have:
f
(z
0
) = lim
z0
f(z
0
+ z) f(z
0
)
z
Let z = x, x R. Then:
f
(z
0
) = lim
x0
f(x
0
+ x, y
0
) f(x
0
, y
0
)
x
= f
x
(x
0
, y
0
)
Now let z = iy, y R.Then:
f
(z
0
) = lim
y0
f(x
0
, y
0
+ y) f(x
0
, y
0
)
iy
= if
y
(x
0
, y
0
)
Equating these:
if
x
(z
0
) = f
y
(z
0
)
(z
0
) = f
x
(z
0
).
Corollary:
Suppose f
x
, f
y
are continuous on open and satisfy the Cauchy-Riemann equations. Then
f is holomorphic on .
Proof.
This follows from the previous two lemmas.
b] [a, b], a
bicontinuous bijection/homeomorphism with ( a) = a, (
b) = b.
Denition (curve) :
A curve is an equivalence class of paths with respect to paramatrisation.
Denition (integral along path) :
The integral of a continuous function f along a piecewise smooth path is dened by
f(z)dz =
f((t))
(t)dt
Example (path integral):
Find
z
n
where (t) = e
it
, 0 t 2.
Solution.
z
n
=
2
0
(e
it
)
n
(e
it
)
dt = i
2
0
e
it(n+1)
dt = 0
Lemma:
f(z)dz =
b
f((()))
()
(())d
=
b
f((()))[
()[
(())d
=
b
a
f((t))
(t)dt
(f + g)dz =
fdz +
gdz
(ii)
1
+
2
fdz =
1
fdz +
2
fdz
16
Proof.
Proof excluded.
b
a
[
(t)[dt
Lemma (basic estimate):
f(z)dz
max
||
[f[ .L()
Proof.
f(z)dz
b
a
f((t))
(t)dt
b
a
[f((t))[.[
(t)[dt
=
b
a
max
||
[f[.[
(t)[dt
max
||
[f[
b
a
[
(t)[dt
= max
||
[f[ .L()
Denition (anti-derivative) :
F(z) is the anti-derivative of f(z) if F
f(z)dz =
b
a
f((t))
(t)dt
=
b
a
h
(t)dt
= h(b) h(a)
= F((b)) F((a))
Corollary:
If f has an anti-derivative then the path integral along a path depends only on the
end-points of the path.
Proof.
Follows immediately from the lemma.
Corollary:
If is a closed path and f has an anti-derivative, then
f(z)dz = 0.
Proof.
Follows immediately from the lemma.
Theorem (Cauchy-Goursat):
Let f be a holomorphic function in (which is open in C). Let be a closed triangle in
, and be the oriented boundary of . Then
f(z)dz = 0
Proof.
We will prove this by contradiction.
(1) Subdivide triangle:
Divide into four smaller triangles of equal size,
f(z)dz = (
)f(z)dz
Suppose that this integral is non-zero (and we hope to reach a contradiction from
this), so is equal to M > 0, say. Then, for one of the four smaller triangles (which we
will call
1
), we have:
1
f(z)dz
M
4
18
We can repeat this process for
1
giving a new triangle
2
, and again for this new
triangle, etc. This gives us a sequence of triangles
1
2
, with the
property that:
k
f(z)dz
M
4
k
(2) Dene sequence and show that it converges:
Choose points z
n
n
. We will show that (z
n
) is Cauchy.
n m
n
m
Let
z
n
n
, z
m
m
n
z
n
, z
m
n
[z
n
z
m
[ L(
n
) =
L()
2
n
Thus for all > 0 there exists an N such that for all m n N we have:
[z
n
z
m
[
L()
2
n
L()
2
N
Therefore (z
n
) is Cauchy, and so converges to z
0
, say.
(3) Bound M above:
f is C-dierentiable at z
0
, so:
f(z) = f(z
0
) + f
(z
0
)(z z
0
) + R(z)
with
[R(z)[
[z z
0
[
0. So:
n
f(z)dz =
n
(f(z
0
) + f
(z
0
)(z z
0
) + R(z)) df
=
n
f(z
0
)dz + f
(z
0
)
n
(z z
0
)dz +
n
R(z)dz
= 0 + 0 +
n
R(z)dz
(The rst two terms are zero because the integrands have anti-derivatives.) So:
n
R(z)dz
M
4
n
19
(4) Show that the upper bound goes to zero:
We have:
n
R(z)dz
max
n
[R(z)[ L(
n
)
= max
n
[R(z)[
L()
2
n
Which well use in a moment, but for now note that
R(z)
z z
0
R(z)
z z
0
[R(z)[ [z z
0
[
Now, if z
n
, then z, z
0
n
which means that
[z z
0
[ L(
n
) =
L()
2
n
So:
[R(z)[ [z z
0
[
L()
2
n
Thus:
n
R(z)dz
L()
2
n
L()
2
n
M
4
n
L()
2
4
n
M L()
2
This contradicts M > 0.
D
f(z)dz = 0
Proof.
Note that if D is convex we can easily break it into triangles. Then the integral along D
is the same as the integral along the boundaries of the triangles, which are all 0. If D is not
convex, then draw a line horizontally and vertically through each vertex. This will break it
into many smaller convex polygons. By the above argument, the integral along the
boundary of each of these convex polygons is 0. Hence the integral along D is 0.
20
D
f(z)dz = 0
Proof.
My notes on this are a mess, and I missed one of the lectures during the proof. Will
probably follow later.
Denition (residue) :
The residue of a function f at a point a
k
is given by:
Res
a
k
f =
1
2i
(a
k
)
f(z)dz
where > 0 is such that
(a
k
) D a
1
, , an.
Theorem (residue theorem):
Let f be holomorphic in a
1
, , a
n
, and D be a bounded simple region in
containing a
1
, , a
n
. Then:
D
f(z)dz =
n
k=1
2iRes
a
k
f
Proof.
I dont have the notes from that day. Im not even sure if the proof was given.
(a)
g(z)
z a
dz
21
Then we estimate the integral:
(a)
g(z)dz
z a
sup
(a)
g(z)
z a
2
= sup
(a)
[g(z)[ 2
We know that lim
za
g(z) = 0. So sup
(a)
[g(z)[ 0 as 0. Therefore:
(a)
g(z)dz
z a
0
as 0. But the integral is independent of , so
(a)
g(z)dz
z a
= 0 = g(a)
(2) g(a) ,= 0:
Note that
f(z)
g(a)
z a
=
f(z) g(a)
z a
So by part (1):
Res
a
f(z)
g(a)
z a
= 0
Then:
Res
a
= Res
a
g(a)
z a
=
1
2i
(a)
g(a)dz
z a
=
1
2i
2
0
g(a)
1
e
it
ie
it
dt
= g(a)
(a) ,= 0, then:
Res
a
g
h
=
g(a)
h
(a)
Proof.
No proof yet.
22
D
f()d
z
Proof.
No proof yet.
f(x)e
ix
dx = 2i
n
k=1
Res
a
k
(F(z)e
iz
)
Proof.
No proof yet.
5 Power Series
Theorem (power series expansion of holomorphic functions):
Let f be holomorphic on the disc
r
(a). Then f has expansion:
f(z) =
n=0
c
n
(z a)
n
where
c
n
=
1
2i
f()d
( a)
n+1
=
f
(n)
(a)
n!
Proof.
No proof yet.
23
Corollary:
The power series expansion holds in every disc where f is holomorphic. Therefore the
radius of convergence of the power series is the maximum radius of the disc where it
extends to a holomorphic function.
Proof.
This follows immediately from the previous theorem.
n=
c
n
(z a)
n
for r < [z a[ < R, where
c
n
=
1
2i
|za|=
f()d
( a)
n+1
Proof.
No proof yet.
Denition (pole) :
f(z) has a pole at a if f is holomorphic in
r
(a) a for some r and has a Laurent series
expansion
f(z) =
nN
c
n
(z a)
n
for some N > 0 with c
N
,= 0.
Lemma:
For f(z) =
n=
c
n
(z a)
n
in
r
(a) a
Res
a
f = c
1
Proof.
No proof yet.
24
Denition (order of zero) :
If f has Taylor expansion
f(z) =
n=0
c
n
(z a)
n
then the order of zero of f at a is dened as
ord
f
(a) = minnc
n
,= 0
= minnf
n
(a) ,= 0
If f is the 0 function then ord
f
(a) = . Otherwise ord
f
(a) (something isnt quite
right about this - surely f(z) = c ,= 0 also has order ?).
Theorem (identity principle):
Let f, g be holomorphic in (which is a connected open set). Then if f[
A
= g[
A
where
A is an open set with limit points in (and these limit points are non-isolated), then
f = g in all of .
Proof.
No proof yet.
25