xy =
x
1 =
_
(1)(1) =
1 = i
2
= 1.
Furthermore, by denition
z z.
Proposition 1.3. [[ is the Euclidean norm on R
2
. In particular, the following hold:
(i) [z[ 0 with [z[= 0 if and only if z = 0;
(ii) [z + w[ [z[+[w[ for z, w C.
Moreover, we have [zw[= [z[[w[ for z, w C and z
1
=
z
z
2
for z C 0.
Proof. Noting that [x + iy[=
_
x
2
+ y
2
, we see that [[ is the Euclidean norm, which
entails (i) and (ii). Letting z, w C, we see that
[zw[
2
= zwzw = (z z)(w w) = [z[
2
[w[
2
.
Also, since [z[
2
= z z, we have 1 = z
z
z
2
for z ,= 0 and thus z
1
=
z
z
2
.
There is a remarkable similarity between the complex multiplication rule
(x, y) (u, v) = (xu yv, xv + yu)
8 CHAPTER 1. THE COMPLEX NUMBERS
and the trigonometric angle sum formulae. Notice that
(cos , sin ) (cos , sin ) = (cos cos sin sin , cos sin + sin cos )
= (cos( + ), sin( + )).
That is, multiplication of 2 complex numbers on the unit circle x
2
+y
2
= 1 corresponds
to addition of their angles of inclination to the x axis. In particular, the mapping
f(z) = z
2
doubles the angle of z = (x, y) and f(z) = z
n
multiplies the angle of z
by n. These statements hold even if z lies on a circle of radius r ,= 1:
(r cos , r sin )
n
= r
n
(cos n, sin n);
this is known as deMoivres Theorem.
Chapter 2
Complex Dierentiation
Denition. Let D C, and let z
0
be an interior point of D, i.e. there exists > 0
such that B
(z
0
) := z C : [z z
0
[< D. A function f : D C is called
complex dierentiable at z
0
if
f
(z
0
) := lim
zz
0
f(z) f(z
0
)
z z
0
exists.
Proposition 2.1. Let D C, let z
0
int D, and let f : D C be complex dieren
tiable at z
0
. Then f is continuous at z
0
.
Proof. Since lim
zz
0
f(z) f(z
0
)
z z
0
exists, we have
0 = lim
zz
0
(z z
0
) lim
zz
0
f(z) f(z
0
)
z z
0
= lim
zz
0
(z z
0
)
f(z) f(z
0
)
z z
0
= lim
zz
0
(f(z) f(z
0
)),
so that f(z
0
) = lim
zz
0
f(z).
Proposition 2.2. Let D C, and let f, g : D C be complex dierentiable at
z
0
int D. Then the following functions are complex dierentiable at z
0
: f + g, fg,
and, if g(z
0
) ,= 0,
f
g
. Moreover, we have:
(f + g)
(z
0
) = f
(z
0
) + g
(z
0
),
(fg)
(z
0
) = f
(z
0
)g(z
0
) + f(z
0
)g
(z
0
),
and
_
f
g
_
(z
0
) =
f
(z
0
)g(z
0
) f(z
0
)g
(z
0
)
g(z
0
)
2
.
9
10 CHAPTER 2. COMPLEX DIFFERENTIATION
Proof. As over R.
Proposition 2.3. Let D, E C, let g : D C and f : E C be such that
g(D) E, and let z
0
int D be such that w
0
:= g(z
0
) int E. Further, suppose
that g is complex dierentiable at z
0
and f is complex dierentiable at w
0
. Then f g
is complex dierentiable at z
0
with
(f g)
(z
0
) = f
(g(z
0
))g
(z
0
).
Proof. As over R.
Examples.
1. All constant functions are (on all of C) complex dierentiable, as is z z on
C. Consequently, all complex polynomials are complex dierentiable on all of
C, and rational functions are complex dierentiable wherever they are dened.
2. Let
f : C C, z z,
and let z
0
= x
0
+iy
0
C. Assume that f is complex dierentiable at z
0
. Then
we have
f
(z
0
) = lim
zz
0
z z
0
z z
0
= lim
yy
0
(x
0
iy) (x
0
iy
0
)
(x
0
+ iy) (x
0
+ iy
0
)
= lim
yy
0
i(y
0
y)
i(y y
0
)
= 1
as well as
f
(z
0
) = lim
zz
0
z z
0
z z
0
= lim
xx
0
(x iy
0
) (x
0
iy
0
)
(x + iy
0
) (x
0
+ iy
0
)
= lim
xx
0
x x
0
x x
0
= 1,
which is impossible. Hence, f is not complex dierentiable at any z
0
C. (On
the other hand, f is continuously partially dierentiableas a function of two
real variableson all of C.)
Lemma 2.1. The following are equivalent for an Rlinear map T : C C:
11
(i) there exists c C such that T(z) = cz for all z C;
(ii) T is Clinear;
(iii) T(i) = iT(1);
(iv) the real 22 matrix representing T with respect to the standard basis of R
2
may
be written as
A =
_
a b
b a
_
for some real a, b R.
Proof. (i) = (ii) = (iii) is obvious.
(iii) = (i): Set c := T(1). For z = x + iy C, this means that
T(x + iy) = T(x) + T(iy)
= xT(1) + yT(i)
= xT(1) + iyT(1)
= zT(1)
= cz.
(iv) (iii): Let a, b, c, d R be such that
A =
_
a b
c d
_
.
represents T with respect to the standard basis of R
2
. Note that
T(1) =
_
a b
c d
_ _
1
0
_
=
_
a
c
_
= a + ic,
and
T(i) =
_
a b
c d
_ _
0
1
_
=
_
b
d
_
= b + id.
Since
iT(1) = c + ia,
we see that
T(i) = iT(1) c = b and d = a.
Theorem 2.1 (CauchyRiemann Equations). Let D C be open, and let z
0
D.
Let f : D C and denote u := Re f, v := Imf. Then the following are equivalent:
(i) f is complex dierentiable at z
0
;
12 CHAPTER 2. COMPLEX DIFFERENTIATION
(ii) f is totally dierentiable at z
0
(in the sense of multivariable calculus), and the
CauchyRiemann dierential equations
u
x
(z
0
) =
v
y
(z
0
) and
u
y
(z
0
) =
v
x
(z
0
)
hold.
Proof. (i) = (ii): Dene
T : C C, z f
(z
0
)z,
and note that
[f(z) f(z
0
) T(z z
0
)[
[z z
0
[
=
f(z) f(z
0
)
z z
0
f
(z
0
)
0
as z z
0
. Therefore, f is totally dierentiable at z
0
. From multivariable calculus, it
follows that the matrix representation of T with respect to the standard basis of R is
the Jacobian of f, i.e.
J
f
(z
0
) =
_
u
x
(z
0
) u
y
(z
0
)
v
x
(z
0
) v
y
(z
0
)
_
.
Since T is Clinear, Lemma 2.1 yields that
u
x
(z
0
) = v
y
(z
0
) and u
y
(z
0
) = v
x
(z
0
).
(ii) = (i): Since f is totally dierentiable at z
0
, we have a unique Rlinear map
T : C C such that
lim
zz
0
[f(z) f(z
0
) T(z z
0
)[
[z z
0
[
= 0.
As we know from multivariable calculus, T is represented by J
f
(z
0
) with respect to the
standard basis of R
2
. Since the CauchyRiemann dierential equations are supposed
to hold, J
f
(z
0
) is of the form described in Lemma 2.1(iv). By Lemma 2.1, there thus
exists c C such that T(z) = cz for all z C. It follows that
f(z) f(z
0
)
z z
0
c
=
[f(z) f(z
0
) c(z z
0
)[
[z z
0
[
0
as z z
0
. Hence, f is complex dierentiable at z
0
.
Remark. In the situation of Theorem 2.1, we have
f
(z
0
)1 =
_
u
x
(z
0
) u
y
(z
0
)
v
x
(z
0
) v
y
(z
0
)
_ _
1
0
_
= u
x
(z
0
) + iv
x
(z
0
)
as well as
f
(z
0
)i =
_
u
x
(z
0
) u
y
(z
0
)
v
x
(z
0
) v
y
(z
0
)
_ _
0
1
_
= u
y
(z
0
) + iv
y
(z
0
),
so that
f
(z
0
) = u
x
(z
0
) + iv
x
(z
0
) = v
y
(z
0
) iu
y
(z
0
).
13
Example. Let
f : C C, z [z[
2
.
Then f is totally dierentiable, with
u
x
= 2x, u
y
= 2y, v
x
= v
y
= 0,
noting that v = 0. The CauchyRiemann equations
u
x
(z
0
) = v
y
(z
0
) and u
y
(z
0
) = v
x
(z
0
)
thus hold if and only if z
0
= 0. By Theorem 2.1, this means that f is complex
dierentiable at z
0
if and only if z
0
= 0.
Corollary 2.1.1. Let D C be open and connected, and let f : D C be complex
dierentiable. Then f is constant on D if and only if f
0.
Proof. Suppose that f
n=0
a
n
(zz
0
)
n
with z, z
0
, a
0
, a
1
, a
2
, . . . C. The point z
0
is called the point of expansion for the series.
Examples.
1. For m N, we have
m
n=0
z
n
=
1 z
m+1
1 z
if z ,= 1. For [z[< 1, we obtain (letting m )
n=0
z
n
=
1
1 z
.
2. For z C, dene
exp(z) :=
n=0
z
n
n!
.
Let z ,= 0, and note that
z
n+1
(n + 1)!
z
n
n!
=
[z[
n + 1
0
as n . As the ratio test holds for series with summands in C as well as for
series over R, we conclude that exp(z) converges absolutely.
Let z, w C, and note that the Cauchy product formula for series over R also
14
15
holds over C. We obtain:
exp(z) exp(w) =
_
j=0
z
j
j!
__
k=0
w
k
k!
_
=
n=0
n
k=0
z
nk
(n k)!
w
k
k!
by the Cauchy product formula, letting n = j + k,
=
n=0
1
n!
n
k=0
_
n
k
_
z
nk
w
k
=
n
n=0
(z + w)
n
n!
= exp(z + w).
We call exp : C C the exponential function. The above property suggests
using the shorthand e
z
for exp(z). An interactive threedimensional graph of
exp(z) is shown in Figure 3.1.
3. The sine and cosine functions on C are dened as
sin(z) :=
n=0
(1)
n
z
2n+1
(2n + 1)!
and
cos(z) :=
n=0
(1)
n
z
2n
(2n)!
for z C. As for exp(z), we see that both sin(z) and cos(z) converge absolutely
for all z C. Moreover, we have for z C:
e
iz
=
n=0
(iz)
n
n!
=
n=0
(iz)
2n
(2n)!
+
n=0
(iz)
2n+1
(2n + 1)!
=
n=0
(1)
n
z
2n
(2n)!
+ i
n=0
(1)
n
z
2n+1
(2n + 1)!
= cos(z) + i sin(z).
Interactive threedimensional graphs of the complex cosine and sine functions
are shown in Figures 3.2, and 3.3.
16 CHAPTER 3. POWER SERIES
Figure 3.1: Surface plot of exp(z) in the complex plane, using an RGB color wheel
to represent the phase. Red indicates real positive values.
Theorem 3.1 (Radius of Convergence). Let
n=0
a
n
(z z
0
)
n
be a complex power
series. Then there exists a unique R [0, ] with the following properties:
n=0
a
n
(z z
0
)
n
converges absolutely at each z B
R
(z
0
);
for each r [0, R), the series
n=0
a
n
(zz
0
)
n
converges uniformly on B
r
[z
0
] :=
z C : [z z
0
[ r;
n=0
a
n
(z z
0
)
n
diverges for each z / B
R
[z
0
].
Moreover, R can be computed via the CauchyHadamard formula:
R =
1
limsup
n
n
_
[a
n
[
.
It is called the radius of convergence for
n=0
a
n
(z z
0
)
n
.
17
Figure 3.2: Surface plot of cos(z) in the complex plane, using an RGB color wheel to
represent the phase. Red indicates real positive values.
Proof. The uniqueness of R follows from the rst and the last property.
Let R [0, ] be dened by the CauchyHadamard formula (we set
1
0
= and
1
= 0).
Let r [0, R), and choose r
,
so that there exists n
0
N such that
n
_
[a
n
[ <
1
r
whenever n n
0
, i.e.
[a
n
[<
_
1
r
_
n
for all n n
0
. For n n
0
and z B
r
[z
0
], we then have
[a
n
(z z
0
)
n
[
_
r
r
_
n
.
18 CHAPTER 3. POWER SERIES
Figure 3.3: Surface plot of sin(z) in the complex plane, using an RGB color wheel to
represent the phase. Red indicates real positive values.
Since
r
r
< 1, we have
n=0
_
r
r
_
n
< . The Weierstra Mtest thus yields that
n=0
a
n
(z z
0
)
n
converges absolutely and uniformly on B
r
[z
0
].
Since every z B
R
(z
0
) is contained in B
r
[z
0
] for some r [0, R), it follows that
n=0
a
n
(z z
0
)
n
converges absolutely for each such z.
Let z / B
R
[z
0
], i.e. [z z
0
[> R, so that
1
[z z
0
[
<
1
R
= limsup
n
n
_
[a
n
[
and thus, for innitely many n N,
1
[z z
0
[
<
n
_
[a
n
[
or, equivalently,
1 < [a
n
(z z
0
)
n
[.
19
It follows that a
n
(zz
0
)
n
n=1
does not converge to zero. Consequently,
n=0
a
n
(z
z
0
)
n
diverges.
Examples.
1.
n=0
z
n
: R = 1.
2.
n=0
z
n
n!
: R = .
3.
n=0
n! z
n
: R = 0.
4.
n=0
(1)
n z
2n+1
(2n+1)!
and
n=0
(1)
n z
2n
(2n)!
: R = .
Theorem 3.2 (TermbyTerm Dierentiation). Let
n=0
a
n
(z z
0
)
n
be a complex
power series with radius of convergence R. Then
f : B
R
(z
0
) C, z
n=0
a
n
(z z
0
)
n
is complex dierentiable at each point z B
R
(z
0
) with
f
(z) =
n=1
na
n
(z z
0
)
n1
.
Proof. Without loss of generality, suppose that z
0
= 0.
We rst show that
n=1
na
n
z
n1
converges absolutely for each z B
R
(0).
Let z B
R
(0), and choose r such that [z[< r < R. Since
1
r
> limsup
n
n
_
[a
n
[,
there exists n
0
N such that [a
n
[<
_
1
r
_
n
for n n
0
and thus
[na
n
z
n1
[<
n
r
_
[z[
r
_
n1
for n n
0
. Since
z
r
< 1, we know from the Ratio Test that
n=1
n
r
_
z
r
_
n1
< ;
the Comparison Test then yields that
n=1
na
n
z
n1
converges absolutely.
In view of the foregoing, we may dene
g : B
R
(0) C, z
n=1
na
n
z
n1
.
We shall devote the rest of the proof to showing that f is complex dierentiable on
B
R
(0) with f
= g.
To this end, x > 0, and dene, for z B
R
(0) and n N,
S
n
(z) :=
n
k=0
a
k
z
k
and R
n
(z) :=
k=n+1
a
k
z
k
.
20 CHAPTER 3. POWER SERIES
Fix z B
R
(0) and let r (0, R) be such that z B
r
(0). Note that
f(w) f(z)
w z
g(z) =
_
S
n
(w) S
n
(z)
w z
S
n
(z)
_
+ (S
n
(z) g(z)) +
R
n
(w) R
n
(z)
w z
for all w B
R
(0) z. We shall see that each of the three summands on the right
hand side of this equation has modulus less than
3
, provided that n is suciently
large and w is suciently close to z.
We start with the last summand. First, note that
R
n
(w) R
n
(z)
w z
=
k=n+1
a
k
w
k
z
k
w z
for all w B
R
(0) z and also that
w
k
z
k
w z
j=1
w
kj
z
j1
j=1
[w[
kj
[z[
j1
kr
k1
for all w B
r
(0) z. Since r < R, we have
k=1
k[a
k
[r
k1
< . Consequently,
there exists n
1
N such that
k=n+1
k[a
k
[r
k1
<
3
for all n n
1
and therefore
R
n
(w) R
n
(z)
w z
k=n+1
a
k
w
k
z
k
w z
k=n+1
k[a
k
[r
k1
<
3
for all n n
1
and all w B
r
(0) z.
For the second summand, just note that lim
n
S
n
(z) = g(z); consequently, there
exists n
2
N such that [S
n
(z) g(z)[<
3
for all n n
2
.
For the rst summand, x n maxn
1
, n
2
. Since
lim
wz
S
n
(w) S
n
(z)
w z
= S
n
(z),
there exists (0, r) such that
S
n
(w) S
n
(z)
w z
S
n
(z)
<
3
for all w B
(z) B
r
(0) z. Consequently, we obtain for all w B
(z) z that
f(w) f(z)
w z
g(z)
S
n
(w) S
n
(z)
w z
S
n
(z)
+[S
n
(z) g(z)[+
R
n
(w) R
n
(z)
w z
<
3
+
3
+
3
= .
Since > 0 was arbitrary, we see that f
(z) = exp(z).
2. sin
(z) = cos(z).
3. cos
(z) = sin(z).
Corollary 3.2.1 (Higher Derivatives of Power Series). Let
n=0
a
n
(z z
0
)
n
be a
complex power series with radius of convergence R. Then
f : B
R
(z
0
) C, z
n=0
a
n
(z z
0
)
n
is innitely often complex dierentiable on B
R
(z
0
) with
f
(k)
(z) =
n=k
n(n 1) (n k + 1)a
n
(z z
0
)
nk
.
for z B
R
(z
0
) and k N. In particular, when z = z
0
we see that
a
n
=
1
n!
f
(n)
(z
0
)
holds for each n N
0
.
Corollary 3.2.2 (Integration of Power Series). Let
n=0
a
n
(z z
0
)
n
be a complex
power series with radius of convergence R. Then
F : B
R
(z
0
) C, z
n=0
a
n
n + 1
(z z
0
)
n+1
is complex dierentiable on B
R
(z
0
) with
F
(z) =
n=0
a
n
(z z
0
)
n
for z B
R
(z
0
).
Chapter 4
Complex Line Integrals
We call a function f : [a, b] C integrable if Re f, Imf : [a, b] R are integrable in
the sense of real variables. (The Riemann integral will do.) In this case, we dene
_
b
a
f(t) dt :=
_
b
a
Re f(t) dt + i
_
b
a
Imf(t) dt.
Denition. A curve (or path) in C is a continuous map : [a, b] C. We call
(a) the initial point of ,
(b) the endpoint (or terminal point) of , and
:= ([a, b]) the trajectory of .
Collectively, we call (a) and (b) the endpoints of .
Examples.
1. Let z, w C. Then
: [0, 1] C, t z
0
+ t(z z
0
)
has the initial point z
0
and the endpoint z, and is the line segment con
necting z
0
with z.
2. For k Z, let
k
: [0, 2] C, e
ik
.
Then
k
(0) = 1 =
k
(2) holds, and for k ,= 0, we have
k
= z C : [z[= 1.
Denition. A curve : [a, b] C is called piecewise smooth if there exists a partition
a = a
0
< a
1
< < a
n
= b such that [
[a
j1
,a
j
]
is continuously dierentiable for
j = 1, . . . , n.
22
23
Denition. The length of a piecewise smooth curve : [a, b] C is dened as
() :=
n
j=1
_
a
j
a
j1
[
(t)[ dt,
where a = a
0
< a
1
< < a
n
= b is a partition such that [
[a
j1
,a
j
]
is continuously
dierentiable for j = 1, . . . , n.
Denition. Let : [a, b] C be a piecewise smooth curve, let a = a
0
< a
1
< <
a
n
= b be a partition such that [
[a
j1
,a
j
]
is continuously dierentiable for j = 1, . . . , n,
and let f : C be continuous. Then the line integral (or contour integral) of f
along is dened as
_
f :=
_
f() d =
n
j=1
_
a
j
a
j1
f((t))
(t) dt.
Properties of the Line Integral. 1. Let be a piecewise smooth curve, let
, C, and let f, g : C be continuous. Then we have
_
(f + g) =
_
f +
_
g.
2. Let be a piecewise smooth curve, let f : C be continuous, and let
C 0 be such that [f()[ C for . Then
C ()
holds.
3. Let : [c, d] C be a piecewise smooth curve, let : [a, b] [c, d] be a
continuously dierentiable function with (a) = c and (b) = d, and let f :
C be continuous. Then we have
_
f =
_
f.
4. Let D C be open, and let f : D C be continuous with antiderivative
F : D C; i.e. F is complex dierentiable at each z D, with F
(z) = f(z).
Then
_
f = F((b)) F((a))
holds for every piecewise smooth curve : [a, b] D.
24 CHAPTER 4. COMPLEX LINE INTEGRALS
Denition. A curve : [a, b] C is called closed if (a) = (b).
Proposition 4.1. Let D C be open, and let f : D C be continuous with an
antiderivative. Then
_
( z
0
)
n
d.
For n ,= 1, let
F : C C, z
(z z
0
)
n+1
n + 1
,
so that F
(z) = (z z
0
)
n
for all z C. It follows that
_
( z
0
)
n
d = 0.
On the other hand, we have
_
( z
0
)
1
d =
_
2
0
rie
i
re
i
d =
_
2
0
i dt = 2i.
Consequently,
C z
0
C, z
1
z z
0
has no antiderivative.
Recall the following denition from multivariable calculus:
Denition. A subset D C is called connected if there are no open sets U, V C
with
U D ,= ,= V D;
U V D;
U V C D.
In other words, there are no open sets U and V , each containing points of D, such
that every point of D lies in exactly one of the sets U and V .
Denition. Let D C be open. A function f : D C is called locally constant if,
for each z
0
D, there exists > 0 such that B
(z
0
) D and f is constant on B
(z
0
).
The following curve constructions will be useful in understanding the relation
between locally constant functions and connectivity.
25
1. Given a < b < c and two curves
1
: [a, b] C and
2
: [b, c] C with
1
(b) =
2
(b), the concatenation of
1
and
2
is the curve
2
: [a, c] C, t
_
1
(t), t [a, b],
2
(t), t [b, c].
If
1
and
2
are piecewise smooth, then so is
1
2
, and we have
_
2
f =
_
1
f +
_
2
f
for each continuous f :
1
2
C.
2. For any curve : [a, b] C, the reversed curve is dened as
: [a, b] C, t (a + b t).
If is piecewise smooth, then so is
, and we have
_
f =
_
f
for each continuous f : C.
3. We denote the straight line segment z
0
+ t(z z
0
) : t [0, 1] by [z
0
, z].
Proposition 4.2 (Locally Constant vs. Connectivity). Let D C be open. Then
the following are equivalent:
(i) D is connected;
(ii) every locally constant function f : D C is constant;
(iii) for any z, w D, there exists a piecewise smooth curve : [a, b] D such that
(a) = z and (b) = w.
Proof. (iii) = (ii): Let f : D C be a locally constant function, and let z, w D.
Let : [a, b] D be a piecewise smooth curve with (a) = z and (b) = w. Since f
is locally constant, the function
[a, b] C, t f((t))
is dierentiable with zero derivative and therefore constant. It follows that f(z) =
f((a)) = f((b)) = f(w).
(ii) = (i): Suppose that D is not connected. Then there exist nonempty open
sets U, V C with U V = and U V = D. Dene
f : D C, z
_
0, z U,
1, z V.
26 CHAPTER 4. COMPLEX LINE INTEGRALS
Then f is locally constant, but not constant.
(i) = (iii): Let z D, and set
U := w D : a piecewise smooth curve : [a, b] D with (a) = z, (b) = w.
Obviously, U ,= (because z U).
We claim that U is open. To see this, let w
0
U, so that there exists a piecewise
smooth curve : [a, b] D with (a) = z and (b) = w
0
. Choose > 0 such that
B
(w
0
) D. For w B
(w
0
), we know that [w
0
, w] is a curve in B
(w
0
) D with
initial point w
0
and endpoint w. Consequently, [w
0
, w] is a piecewise smooth
curve in D with initial point z and endpoint w. It follows that w U, and since
w B
(w
0
) was arbitrary, we have B
(w
0
) U. This proves the openness of U.
Next, we claim that D U is also open. To see this, let w
0
D U, and let
> 0 be so small that B
(w
0
) D. Assume towards a contradiction that there exists
w B
(w
0
) U. Let : [a, b] D be a piecewise smooth curve with (a) = z and
(b) = w. Then [w, w
0
] is a piecewise smooth curve in D with initial point z and
endpoint w
0
, so that w
0
U. This contradicts the choice of w
0
D U. It follows
that B
(w
0
) U = , i.e. B
(w
0
) D U.
Since U and D U are both open with U (D U) = D and U (D U) = ,
the connectedness of D yields that D U = , i.e. D = U.
Lemma 4.1. Suppose D C is an open connected set (a region) and f : D C is
continuous. Let z
0
D. For each z D, let
z
: [a, b] D be a piecewise smooth
curve in D such that
z
(a) = z
0
and
z
(b) = z. Consider the function
F : D C, z
_
z
f() d.
For each z, let > 0 such that B
(z) D and
[ z[< [f() f(z)[< .
For all w B
(z), we nd
F(w) F(z)
w z
f(z)
=
1
[w z[
_
[z,w]
f
_
[z,w]
f(z)
=
1
[w z[
_
[z,w]
(f f(z))
[w z[
[w z[
sup[f() f(z)[: [z, w]
= sup[f() f(z)[: [z, w] < .
27
That is, F
(z) = f(z).
Theorem 4.1 (Antiderivative Theorem). Let D C be open and connected and let
f : D C be continuous. Then the following are equivalent:
(i) f has an antiderivative;
(ii)
_
f =
_
f +
_
f =
_
f
_
f.
(iii) = (i): Fix z
0
D. For each z D, choose a piecewise smooth curve
z
: [a, b] D with
z
(a) = z
0
and
z
(b) = z and let
F : D C, z
_
z
f() d.
For each z D, choose > 0 such that B
(z) that
F(w) =
_
w
f() d
=
_
z[z,w]
f() d by (iii)
=
_
z
f() d +
_
[z,w]
f() d
= F(z) +
_
[z,w]
f() d.
Lemma 4.1 then implies that F is an antiderivative for f.
From now on, we shall use the word curve as shorthand for piecewise smooth curve.
Chapter 5
Cauchys Integral Theorem and
Formula
Denition. Let D C be open. If f : D C is complex dierentiable at each
z D, then we call f holomorphic (or analytic) on D.
Let z
1
, z
2
, and z
3
be three dierent points in C. They span a triangle . Its
boundary can be parametrized as a curve with counterclockwise orientation.
z
1
z
2
z
3
f() d = 0.
Proof. First, we note that the result holds trivially whenever z
1
, z
2
, and z
3
are colin
ear. Otherwise, we can split at its medians into four subtriangles
(1)
,
(2)
,
(3)
,
and
(4)
as shown in the following gure:
28
29
(4)
(3)
(1)
(2)
As the line segments in the interior of also occur as their reversed paths, we
have
_
f =
4
j=1
_
(j)
f,
so that
j=1
(j)
f
.
Choose j 1, 2, 3, 4 such that
(j)
f
1
f
;
also, note that (
1
) =
1
2
().
Repeat this argument with
1
in place of , and obtain a triangle
2
1
with
(
2
) =
1
2
(
1
) =
1
4
()
and
1
f
2
f
,
so that
1
f
16
2
f
.
Inductively, we obtain triangles
1
2
30 CHAPTER 5. CAUCHYS INTEGRAL THEOREM AND FORMULA
with
(
n
) =
1
2
n
()
and
4
n
_
n
f
for n N.
Let z
0
n=1
n
, and dene
r : D C, z f(z) f(z
0
) f
(z
0
)(z z
0
),
so that lim
zz
0
[r(z)[
[z z
0
[
= 0 and
_
(r f) =
_
[f(z
0
) f
(z
0
)(z z
0
)] dz = 0 for each
closed curve in D (noting that the integrand has an antiderivative). Consequently,
4
n
_
n
r
r(z)
z z
0
[()]
2
for all z D with [z z
0
[< . Choose n N such that
n
B
(z
0
). For z
n
,
this means that
[z z
0
[ (
n
) =
1
2
n
().
We thus obtain:
4
n
_
n
r
4
n
(
n
) sup
n
[r()[= 2
n
() sup
n
[r()[
2
n
() sup
n
[()]
2
[ z
0
[
. .
1
2
n
()
.
As > 0 was arbitrary, this proves the claim.
Denition. A set D C is called star shaped if there exists z
0
D such that
[z
0
, z] D for each z D. The point z
0
is called a center for D.
Theorem 5.2. Let D C be open and star shaped with center z
0
, and let f : D C
be continuous such that _
f() d = 0
for each triangle D with z
0
as a vertex. Then f has an antiderivative.
31
Proof. Let z
0
D be a center for D. Dene
F : D C, z
_
[z
0
,z]
f.
Let z D, and let > 0 be such that B
(z) D. Let w B
(z). Since [z
0
, z]
[z, w] [w, z
0
] is the boundary of a triangle D,
_
[z
0
,z][z,w][w,z
0
]
f = 0,
so that
F(w) =
_
[z
0
,w]
f =
_
[w,z
0
]
f =
_
[z
0
,z]
f +
_
[z,w]
f = F(z) +
_
[z,w]
f.
Lemma 4.1 then implies that F is an antiderivative for f.
Corollary 5.2.1. Let D C be open and star shaped, and let f : D C be
holomorphic. Then f has an antiderivative.
Proof. Apply Goursats Lemma and Theorem 5.2.
Corollary 5.2.2. Let D C be open, and let f : D C be holomorphic. Then,
for each z
0
D, there exists a neighbourhood U D of z
0
such that f[
U
has an
antiderivative.
Corollary 5.2.3 (Cauchys Integral Theorem for StarShaped Domains). Let D C
be open and star shaped, and let f : D C be holomorphic. Then
_
f() d = 0 for
each closed curve in D.
Proof. This follows from Corollary 5.2.1 and Theorem 4.1.
Example. The sliced plane is dened as
C
:= z C : z / (, 0].
Then C
C, z
_
[1,z]
1
d
is an antiderivative of z
1
z
on C
, and let
z
be any curve in C
d = Log z.
For any z C
, the curve
: [0, ] C
,
t [z[e
it
.
32 CHAPTER 5. CAUCHYS INTEGRAL THEOREM AND FORMULA
has the initial point [z[ and the endpoint z. It follows that [1, [z[] is curve with
initial point 1 and endpoint z as shown:
y
x
z
[1, z]
1
z
[1, z]
It follows that
Log z :=
_
[1,z]
1
d +
_
d = log[z[+i
_
0
[z[e
it
[z[e
it
dt = log[z[+iArg z.
Lemma 5.1. Let D C be open and star shaped with center z
0
, and let f : D C
be continuous such that f[
D\{z
0
}
is holomorphic. Then f has an antiderivative on D.
Proof. Let be a triangle in D having a vertex z
0
:
z
0
w
z
z
1
z
3
z
2
Let z
1
be an interior point of [z
0
, w], let z
2
be an interior point of [z
0
, z], and let
z
3
be an interior point of [w, z]. As shown above, we use these points to split into
33
four subtriangles, denoted by (z
0
, z
1
, z
2
), (z
1
, z
3
, z
2
), (z
1
, w, z
3
), and (z
2
, z
3
, z).
As in the proof of Goursats Lemma, we have
_
f =
_
(z
0
,z
1
,z
2
)
f +
_
(z
1
,z
3
,z
2
)
f +
_
(z
1
,w,z
3
)
f +
_
(z
2
,z
3
,z)
f.
Since (z
1
, z
3
, z
2
), (z
1
, w, z
3
), (z
2
, z
3
, z) D z
0
, and since f is holomorphic on
D z
0
, Goursats Lemma yields
_
(z
1
,z
3
,z
2
)
f =
_
(z
1
,w,z
3
)
f =
_
(z
2
,z
3
,z)
f = 0,
so that _
f =
_
(z
0
,z
1
,z
2
)
f.
It follows that
_
(z
0
,z
1
,z
2
)
f
((z
0
, z
1
, z
2
)) sup
(z
0
,z
1
,z
2
)
[f()[.
Since [f[ is continuous on , it is bounded above by some M > 0. By placing z
1
and z
2
suciently close to z
0
, we see that ((z
0
, z
1
, z
2
)) can be made smaller than every
/M > 0. We deduce that
_
[z] B
r
(z
0
), so that
_
B(z)
1
z
d = 2i.
We need to show that
_
Br(z
0
)
1
z
d =
_
B(z)
1
z
d =
_
B(z)
1
z
d.
34 CHAPTER 5. CAUCHYS INTEGRAL THEOREM AND FORMULA
This is how the situation looks like:
z
0
B
z
We connect the boundaries of B
r
(z
0
) and B
2
z
0
z
Then it is clear that
_
1
1
z
d +
_
2
1
z
d =
_
Br(z
0
)
1
z
d +
_
B(z)
1
z
d.
The sketches also show that there exist starshaped open set D
j
C z with
j
D
j
for j = 1, 2. Cauchys integral theorem for starshaped domains thus
yields that
_
j
1
z
d = 0
for j = 1, 2, which proves the claim.
Theorem 5.3 (Cauchys Integral Formula for Circles). Let D C be open, let
f : D C be holomorphic, and let z
0
D and r > 0 be such that B
r
[z
0
] D. Then
we have
f(z) =
1
2i
_
Br(z
0
)
f()
z
d
for all z B
r
(z
0
).
Remark. A consequence of Theorem 5.3 is that the values of f on all of B
r
[z
0
] are
predetermined by those on B
r
(z
0
)!
Proof of Theorem 5.3. Since B
r
[z
0
] is compact, we may choose R > 0 be such that
B
r
[z
0
] B
R
(z
0
) D. Let z B
r
(z
0
), and note that z is a center for the starshaped
domain B
R
(z
0
).
Dene
g : D C,
_
f()f(z)
z
, ,= z,
f
(z), = z.
36 CHAPTER 5. CAUCHYS INTEGRAL THEOREM AND FORMULA
Then g is continuous on D and holomorphic on D z. By Lemma 5.1, g has an
antiderivative on B
R
(z
0
). Since B
r
(z
0
) is closed, this means that
0 =
_
Br(z
0
)
g() d =
_
Br(z
0
)
f()
z
d f(z)
_
Br(z
0
)
1
z
d
. .
=2i
,
using the identity
_
Br(z
0
)
1
z
d = 2i established by Lemma 5.2. Division by 2i
yields the claim.
Corollary 5.3.1 (Mean Value Equation). Let D C be open, let f : D C be
holomorphic, and let z
0
D and r > 0 be such that B
r
[z
0
] D. Then we have
f(z
0
) =
1
2
_
2
0
f(z
0
+ re
i
) d.
Proof. Parametrize B
r
(z
0
) as
: [0, 2] C, z
0
+ re
i
.
The Cauchy Integral Formula then yields
f(z
0
) =
1
2i
_
f()
z
0
d
=
1
2i
_
2
0
f(z
0
+ re
i
)
re
i
ire
i
d
=
1
2
_
2
0
f(z
0
+ re
i
) d.
Lemma 5.3. Let D R
N
be open, and let f : [a, b] D R be continuous such that
f
x
1
, . . . ,
f
x
N
: [a, b] D R all exist and are continuous. Dene
g : D R, x
_
b
a
f(t, x) dt.
Then g is continuously partially dierentiable with
g
x
j
(x) =
_
b
a
f
x
j
(t, x) dt
for x D and j = 1, . . . , N.
37
Proof. There is no loss of generality supposing that N = 1.
Let x
0
, x D be such that every number between them is also in D. By the
Mean Value Theorem of single variable calculus, there exists, for each t [a, b], a real
number
t
between x
0
and x such that
f(t, x) f(t, x
0
)
x x
0
=
f
x
(t,
t
).
Let > 0. By uniform continuity, there exists a > 0 such that
f
x
(t, x
1
)
f
x
(t, x
2
)
<
b a
for any t [a, b] and all x
1
, x
2
between x
0
and x with [x
1
x
2
[< .
Suppose that [x
0
x[< . Then we have:
g(x) g(x
0
)
x x
0
_
b
a
f
x
(t, x
0
) dt
_
b
a
_
f(t, x) f(t, x
0
)
x x
0
f
x
(t, x
0
)
_
dt
_
b
a
f(t, x) f(t, x
0
)
x x
0
f
x
(t, x
0
)
dt
=
_
b
a
f
x
(t,
t
)
f
x
(t, x
0
)
dt
_
b
a
b a
dt, because [
t
x
0
[< ,
= .
This proves that g is dierentiable at x
0
with g
(x
0
) =
_
b
a
f
x
(t, x
0
) dt.
A similar (but easier) argument shows that g
is continuous.
Lemma 5.4. Let D C be open, and let f : [a, b] D C be continuous such that
f
z
: [a, b] D C exists and is continuous. Dene
g : D R, z
_
b
a
f(t, z) dt.
Then g is holomorphic with
g
(z) =
_
b
a
f
z
(t, z) dt
for z D.
Proof. Apply Lemma 5.3 to Re f and Imf and note that the Cauchy Riemann dif
ferential equations are satised.
38 CHAPTER 5. CAUCHYS INTEGRAL THEOREM AND FORMULA
Theorem 5.4 (Higher Derivatives). Let D C be open, let z
0
D and r > 0 be
such that B
r
[z
0
] D, and let f : D C be continuous such that
f(z) =
1
2i
_
Br(z
0
)
f()
z
d
holds for all z B
r
(z
0
). Then f is innitely often complex dierentiable on B
r
(z
0
)
and satises
f
(n)
(z) =
n!
2i
_
Br(z
0
)
f()
( z)
n+1
d ()
holds for all z B
r
(z
0
) and n N
0
.
Proof. We prove by induction on n N
0
: f is ntimes complex dierentiable and ()
holds.
For n = 0, the claim is clear, so suppose that it is true for some n N
0
. Dene
F : [0, 2] B
r
(z
0
) C, (, z)
n!
2
f(z
0
+ re
i
)re
i
(z
0
+ re
i
z)
n+1
.
Then F is continuous and, by the induction hypothesis, satises
f
(n)
(z) =
n!
2i
_
Br(z
0
)
f()
( z)
n+1
d =
_
2
0
F(, z) d
for all z B
r
(z
0
). Furthermore,
F
z
(, z) =
(n + 1)!
2
f(z
0
+ re
i
)re
i
(z
0
+ re
i
z)
n+2
is continuous on [0, 2] B
r
(z
0
). From Lemma 5.4, we thus conclude that f
(n)
is
holomorphic on D with
f
(n+1)
(z) =
_
2
0
F
z
(, z) d =
(n + 1)!
2i
_
Br(z
0
)
f()
( z)
n+2
d.
Corollary 5.4.1 (Generalized Cauchy Integral Formula). Let D C be open, and
let f : D C be holomorphic. Then f is innitely often complex dierentiable on D.
Moreover, for any z
0
D and r > 0 such that B
r
[z
0
] D, the generalized Cauchy
integral formula
f
(n)
(z) =
n!
2i
_
Br(z
0
)
f()
( z)
n+1
d
holds for all z B
r
(z
0
) and n N
0
.
39
Proof. Apply Theorem 5.3 and 5.4.
Example. We shall use Cauchys integral theorem and formula to evaluate the line
integral
_
( 1)
d
for various curves :
(a) = B
( 1)
d =
_
B(3)
f()
d = 2i f(0) = 2i.
(b) = B1
2
(i): As the integrand is holomorphic in the starshaped domain B3
4
(i),
Cauchys integral theorem yields that
_
B
1
2
(i)
e
( 1)
d = 0.
(c) = B
2
(0): the method of partial fractions yields
1
z(z 1)
=
1
z 1
1
z
.
Since 0, 1 B
2
(0), we obtain with the help of Cauchys integral formula:
_
B
2
(0)
e
( 1)
d =
_
B
2
(0)
e
1
d
_
B
2
(0)
e
d = 2i (e 1).
Theorem 5.5 (Characterizations of Holomorphic Functions). Let D C be open,
and let f : D C be continuous. Then the following are equivalent:
(i) f is holomorphic;
(ii) the Morera condition holds, i.e.
_
0.
Let C 0 be such that [f(z)[ C for all z C. Let z C be arbitrary, and let
r > 0. By the generalized Cauchy integral formula, we have
[f
(z)[=
1
2
_
Br(z)
f()
( z)
2
d
1
2
(B
r
(z)) sup
Br(z)
[f()[
[ z[
2
1
2
2r
C
r
2
=
C
r
.
Letting r , we obtain f
(z)
f(z)
.
(b) If D is star shaped, and f : D C is holomorphic such that 0 / f(D), show that
there exists a holomorphic function g : D C such that f = exp g.
Problem 5.2. Let z
0
C, let r > 0, and let f : B
r
[z
0
] C be continuous such that
f[
Br(z
0
)
is holomorphic. Show that
f(z) =
1
2i
_
Br(z
0
)
f()
z
d
for all z B
r
(z
0
). Hint: For (0, 1), apply the Cauchy integral formula to
z f((z z
0
) + z
0
) on B
r
(z
0
); then let 1
.
Chapter 6
Convergence of Holomorphic
Functions
Recall the following denition:
Denition. Let D C be open. A sequence (f
n
)
n=1
of Cvalued functions on D is
said to converge uniformly on D to f : D C if, for each > 0, there exists N N
such that [f
n
(z) f(z)[< for all n N and all z D.
We recall the following theorem from analysis:
Theorem 6.1 (Uniform Convergence Preserves Continuity). Let D C be open, and
let (f
n
)
n=1
be a sequence of continuous, Cvalued functions on D converging uniformly
on D to f : D C. Then f is continuous.
We now localize the notion of uniform convergence:
Denition. Let D C be open. Then a sequence (f
n
)
n=1
of Cvalued functions on
D is said to converge locally uniformly on D to f : D C if, for each z
0
D, there
exists a neighbourhood U D of z
0
such that (f
n
[
U
)
n=1
converges to f[
U
uniformly
on U.
Proposition 6.1 (Local Uniform Convergence). Let D C be open, and let (f
n
)
n=1
be a sequence of continuous, Cvalued functions on D converging locally uniformly on
D to f : D C. Then f is continuous.
Proof. Let z
0
D, and let U D be a neighbourhood of z
0
such that f
n
[
U
f[
U
uniformly on U By Theorem 6.1, f[
U
is continuous. Hence, f is continuous at z
0
.
Proposition 6.2 (Compact Convergence). Let D C be open, and let f, f
1
, f
2
, . . . :
D C be functions. Then the following are equivalent:
(i) (f
n
)
n=1
converges to f locally uniformly on D;
42
43
(ii) for each compact K D, the sequence (f
n
[
K
)
n=1
converges to f[
K
uniformly on
K.
Proof. (i) = (ii): Let K D be compact. For each z K, there exists a neigh
bourhood U
z
D of z such that f
n
[
Uz
f[
Uz
uniformly on U
z
. Since K is compact,
there exist z
1
, . . . , z
m
K such that
K U
z
1
U
zm
.
Let > 0. For each j = 1, . . . , m, there exists n
j
N such that [f
n
(z) f(z)[< for
all n n
j
and all z U
z
j
. Set N := maxn
1
, . . . , n
m
. Then [f
n
(z) f(z)[< holds
for all n N and z K.
(ii) =(i): Let z
0
D, and let r > 0 be such that B
r
[z
0
] D. Since B
r
[z
0
] is com
pact, (f
n
[
Br[z
0
]
)
n=1
converges uniformly on B
r
[z
0
] to f[
Br[z
0
]
. Trivially, (f
n
[
Br(z
0
)
)
n=1
thus converges uniformly on B
r
(z
0
) to f[
Br(z
0
)
.
Instead of locally uniform convergence, we therefore often speak of compact con
vergence.
Lemma 6.1. Let D C be open, let be a curve in D, and let f, f
1
, f
2
, . . . : D C
be continuous functions such that (f
n
[
{}
)
n=1
converges to f[
{}
uniformly on .
Then we have _
f() d = lim
n
_
f
n
() d.
Proof. Let > 0, and choose N N such that
[f
n
() f()[<
() + 1
for all n N and z . For n N, we thus obtain:
f
n
(f
n
f)
() sup[f
n
() f()[:
()
() + 1
< .
Lemma 6.2. Let D C be open, let be a curve in D, and let f
1
, f
2
, . . . : D C be
continuous functions converging compactly to f : D C. Then f is continuous, and
we have _
f() d = lim
n
_
f
n
() d.
44 CHAPTER 6. CONVERGENCE OF HOLOMORPHIC FUNCTIONS
Proof. If : [a, b] C is a curve (and thus continuous) then = ([a, b]) is
compact. Hence, Lemma 6.1 applies.
Theorem 6.2 (Weierstra Theorem). Let D C be open, let f
1
, f
2
, . . . : D C
be holomorphic such that (f
n
)
n=1
converges to f : D C compactly. Then f is
holomorphic, and (f
(k)
n
)
n=1
converges compactly to f
(k)
for each k N.
Proof. By Theorem 6.1, f is continuous.
To see that f is holomorphic, let D be a triangle. By Goursats Lemma,
_
f
n
() d = 0 holds for all n N. From Lemma 6.2, we conclude that
_
f() d = lim
n
_
f
n
() d = 0,
i.e. f satises the Morera condition and thus is holomorphic.
Let z
0
D, and let 0 < r < R be such that B
r
[z
0
] B
R
(z
0
) B
R
[z
0
] D. For
any z B
r
(z
0
), we have
[f
n
(z) f
(z)[ =
1
2
_
B
R
(z
0
)
f
n
() f()
( z)
2
d
1
2
(B
R
(z
0
)) sup
B
R
(z
0
)
f
n
() f()
( z)
2
R
(R r)
2
sup
B
R
(z
0
)
[f
n
() f()[.
Let > 0, and choose N N such that
[f
n
() f()[<
(R r)
2
R
for all n N and B
R
(z
0
). Then it follows from the above estimates that
[f
n
(z)f
n
[
Br(z
0
)
)
n=1
converges
to f
[
Br(z
0
)
uniformly on B
r
(z
0
). As z
0
D is arbitrary, this means that (f
n
)
n=1
converges to f locally uniformly, i.e. compactly, on D.
For higher derivatives, the claim now follows by induction.
Lemma 6.3. Let z
0
C, let r > 0, and let z B
r
(z
0
). Then
1
z
=
n=0
1
( z
0
)
n+1
(z z
0
)
n
converges absolutely and uniformly on B
r
(z
0
).
45
Proof. Let B
r
(z
0
), and note that
1
z
=
1
( z
0
) (z z
0
)
=
1
z
0
1
1
zz
0
z
0
.
Since [z z
0
[< r and [ z
0
[= r, we have
zz
0
z
0
< 1, so that
1
z
=
1
( z
0
)
n=0
_
z z
0
z
0
_
n
=
n=0
1
( z
0
)
n+1
(z z
0
)
n
.
Since
(zz
0
)
n
(z
0
)
n+1
=
zz
0

n
r
n+1
and
n=0
zz
0

n
r
n+1
< , the Weierstra Mtest yields abso
lute and uniform convergence on B
r
(z
0
).
Theorem 6.3 (Power Series for Holomorphic Functions). Let D C be open. Then
the following are equivalent for f : D C:
(i) f is holomorphic;
(ii) for each z
0
D, there exists r > 0 with B
r
(z
0
) D and a
0
, a
1
, a
2
, . . . C such
that f(z) =
n=0
a
n
(z z
0
)
n
for all z B
r
(z
0
);
(iii) for each z
0
D and r > 0 with B
r
(z
0
) D, we have
f(z) =
n=0
f
(n)
(z
0
)
n!
(z z
0
)
n
for all z B
r
(z
0
).
Proof. (iii) = (ii) is trivial; (ii) = (i) follows from Theorem 3.2.
(i) = (iii): Let z
0
D, and let r > 0 be such that B
r
(z
0
) D. Let z B
r
(z
0
)
and choose (0, r) such that z B
(z
0
). Then we have:
f(z) =
1
2i
_
B(z
0
)
f()
z
d
=
1
2i
_
B(z
0
)
n=0
f()
( z
0
)
n+1
(z z
0
)
n
d, by Lemma 6.3,
=
n=0
_
1
2i
_
B(z
0
)
f()
( z
0
)
n+1
d
_
(z z
0
)
n
, by Lemma 6.1,
=
n=0
f
(n)
(z
0
)
n!
(z z
0
)
n
.
Chapter 7
Elementary Properties of
Holomorphic Functions
Theorem 7.1 (Identity Theorem). Let D C be open and connected, and let f, g :
D C be holomorphic. Then the following are equivalent:
(i) f = g;
(ii) the set z D : f(z) = g(z) has a cluster point in D;
(iii) there exists z
0
D such that f
(n)
(z
0
) = g
(n)
(z
0
) for all n N
0
.
Proof. Without loss of generality, it suces to prove the case where g = 0.
(i) = (iii) is trivial.
(iii) = (ii): Let z
0
D be as in (iii), and let r > 0 be such that B
r
(z
0
) D.
Then we have by Theorem 6.3 that
f(z) =
n=0
f
(n)
(z
0
)
n!
(z z
0
)
n
= 0
for all z B
r
(z
0
), so that
B
r
(z
0
) Z(f) := z D : f(z) = 0.
Every point in B
r
(z
0
) is therefore a cluster point of Z(f).
(ii) = (i): Let
V := z D : z is a cluster point of Z(f),
so that V ,= by (ii).
We claim that V is open. Let z
0
V , and let r > 0 be such that B
r
(z
0
) D.
Then
f(z) =
n=0
a
n
(z z
0
)
n
46
47
holds for some a
0
, a
1
, a
2
, . . . C and all z B
r
(z
0
). As z
0
is a cluster point of Z(f),
there exists a sequence (z
k
)
k=1
in Z(f) z
0
such that z
0
= lim
k
z
k
. Inductively, we
nd that a
n
= 0 for all n N
0
: given that a
n
= 0 for n = 0, 1, 2, m1 we see that
0 = lim
k
f(z
k
)
(z
k
z
0
)
m
= lim
k
n=m
a
n
(z
k
z
0
)
nm
= a
m
,
on noting, in view of Theorems 3.1 and 6.3, that the uniform convergence of the
power series on, say, B
r/2
[z
0
] justies the interchange of limits. Hence f(z) = 0 for
all z B
r
(z
0
). That is, B
r
(z
0
) Z(f) so that B
r
(z
0
) V .
We now claim that D V is also open. Let z
0
D V , and > 0 be such that
B
(z
0
) D and (B
(z
0
) z
0
) Z(f) = . For any z B
(z
0
) z
0
and > 0 such
that B
(z) B
(z
0
) z
0
, we thus have (B
(z
0
) D V .
In summary, V and D V are both open and clearly satisfy D = V (D V ) and
V (D V ) = . The connectedness of D yields D = V and thus Z(f) = D.
Examples.
1. There is no nonzero entire function f : C C such that f
_
1
n
_
= 0 for n N.
For any entire function f with this property, 0 is a cluster point of Z(f). By
the Identity Theorem, this means f 0.
2. Since R has cluster points in C, the holomorphic extensions of exp, cos, and sin
from R to C are unique. For analogous reasons, Log is the only holomorphic
extension of log to C
.
3. There is no entire function f such that
f
_
1
n
_
=
1
n
and f
_
1
n
_
=
1
n
2
for n N. In the view of the identity theorem, the rst condition necessitates
that f(z) = z whereas the second one implies that f(z) = (z)
2
for all z C.
Lemma 7.1. Let D C be open, let f : D C be holomorphic, and let z
0
D and
r > 0 be such that B
r
[z
0
] D. Suppose that
[f(z
0
)[< inf
zBr(z
0
)
[f(z)[.
Then f has a zero in B
r
(z
0
).
48CHAPTER 7. ELEMENTARYPROPERTIES OF HOLOMORPHIC FUNCTIONS
Proof. Assume otherwise, i.e. f has no zero in B
r
(z
0
). The hypothesis implies that f
has no zero on B
r
(z
0
), so that f has no zero in B
r
[z
0
]. Assume however, that for each
R > 0 such that B
r
[z
0
] B
R
(z
0
) D, there is a zero of f in B
R
(z
0
). Then we have
a sequence (R
n
)
n=1
in (r, ) with r = lim
n
R
n
such that B
r
[z
0
] B
Rn
(z
0
) D and
Z(f) B
Rn
(z
0
) ,= for each n N. For each n N, pick z
n
Z(f) B
Rn
(z
0
). Then
(z
n
)
n=1
is bounded, and thus has a convergent subsequence (z
n
k
)
k=1
with limit z
.
Clearly, z
B
r
[z
0
], which is impossible.
Consequently, f has no zero on some B
R
(z
0
) with B
r
[z
0
] B
R
(z
0
) D.
From the Cauchy Integral Formula, we obtain
1
[f(z
0
)[
=
1
2i
_
Br(z
0
)
1
f()
1
z
0
d
1
2
2r sup
Br(z
0
)
1
[f()[r
=
1
inf
Br(z
0
)
[f()[
and thus
[f(z
0
)[ inf
Br(z
0
)
[f()[,
which is a contradiction.
Theorem 7.2 (Open Mapping Theorem). Let D C be open and connected, and let
f : D C be holomorphic and not constant. Then f(D) C is open and connected.
Proof. By the continuity of f, it is clear that f(D) is connected.
Let w
0
f(D), and let z
0
D be such that w
0
= f(z
0
). Choose r > 0 such
that B
r
[z
0
] D and such that z B
r
[z
0
] : f(z) = w
0
= z
0
. (This can be
accomplished with the help of the Identity Theorem.) Let =
1
2
inf
Br(z
0
)
[f(z) w
0
[> 0.
We claim that B
(w
0
) f(D). Let w B
(w
0
). For z B
r
(z
0
), we have
[f(z) w[ [f(z) w
0
[[w w
0
[> 2 = .
It follows that
[f(z
0
) w[= [w w
0
[< inf
zBr(z
0
)
[f(z) w[.
By Lemma 7.1, this means that
D C, z f(z) w
has a zero in B
r
(z
0
). It follows that w f(D).
Theorem 7.3 (Maximum Modulus Principle). Let D C be open and connected,
and let f : D C be holomorphic such that the function
[f[: D C, z [f(z)[
attains a local maximum on D. Then f is constant.
49
Proof. Let z
0
D be such that [f[ attains a local maximum at z
0
, i.e. there exists
> 0 such that B
(z
0
) D and [f(z
0
)[ [f(z)[ for all z B
(z
0
). Then f(z
0
) is
not an interior point of f(B
(z
0
)), so that f[
B(z
0
)
is constant by the Open Mapping
Theorem. The Identity Theorem then yields that f is constant.
Corollary 7.3.1. Let D C be open and connected, and let f : D C be holo
morphic such that [f[ attains a local minimum on D. Then f is constant or f has a
zero.
Proof. Suppose that f has no zero. Applying the Maximum Modulus Principle to 1/f
yields that f is constant.
Corollary 7.3.2 (Maximum Modulus Principle for Bounded Domains). Let D C
be open, connected, and bounded, and let f : D C be continuous such that f[
D
is
holomorphic. Then [f[ attains its maximum over D on D.
Proof. The claim is trivial if f is constant, so suppose that f is not constant.
Since f is continuous and D is compact, there exists a point z
0
D with [f(z
0
)[=
max
_
[f(z)[: z D
_
. If z
0
D, then [f[ would attain a local maximum at z
0
, which
is impossible by the Maximum Modulus Principle. Therefore z
0
D must hold.
From now on, we shall use D to denote the open unit disc B
1
(0).
Theorem 7.4 (Schwarzs Lemma). Let f : D D be holomorphic such that f(0) = 0.
Then one has
[f(z)[ [z[ for z D and [f
(0)[ 1.
Moreover, if there exists z
0
D 0 such that [f(z
0
)[= [z
0
[ or if [f
(0)[= 1, then
there exists c C with [c[= 1 such that f(z) = cz for z D.
Proof. Let f(z) =
n=0
a
n
z
n
be the power series expansion of f. Since f(0) = 0, we
have a
0
= 0. Dene
g : D C, z
n=1
a
n
z
n1
.
Then g is holomorphic with g(0) = a
1
= f
(0)[= [g(0)[ 1.
Suppose that there exists z
0
D0 such that [f(z
0
)[= [z
0
[ or [f
(0)[= [g(0)[= 1.
Then [g[ has a maximum at z
0
or 0, respectively, so that g is constant. Hence, there
exists c C with [c[= 1 such that f(z) = zg(z) = cz for z D.
50CHAPTER 7. ELEMENTARYPROPERTIES OF HOLOMORPHIC FUNCTIONS
Denition. Let D
1
, D
2
C be open. Then f : D
1
D
2
is called biholomorphic (or
conformal) if
(a) f is bijective and
(b) both f and f
1
are holomorphic.
Corollary 7.4.1. Let f : D D be biholomorphic such that f(0) = 0. Then there
exists c C with [c[= 1 such that f(z) = cz for z D.
Proof. Let z D. Then [f(z)[ [z[ holds by Schwarzs Lemma, as does
[z[= [f
1
(f(z))[ [f(z)[.
The result then follows from Schwarzs Lemma.
Lemma 7.2. Let w D, and dene
w
: D C, z
w z
1 wz
.
Then:
(i)
w
maps D bijectively onto D;
(ii)
w
(w) = 0;
(iii)
w
(0) = w;
(iv)
1
w
=
w
.
Proof. Obviously,
w
is holomorphic and extends continuously to D.
Then for [z[= 1 we may express
[
w
(z)[=
w z
1 wz
1
[ z[
=
w z
z w
= 1.
By the Maximum Modulus Principle,
w
(D) D holds. Since
w
is not constant,
w
(D) is open and thus contained in the interior of D, i.e. in D.
It is obvious that
w
(w) = 0 and
w
(0) = w.
Moreover, we have for z D:
(
w
w
)(z) =
w
wz
1 wz
1 w
wz
1 wz
=
w(1 wz) (w z)
(1 wz) w(w z)
=
[w[
2
z + z
1 [w[
2
= z.
Hence,
w
is bijective with
1
w
=
w
.
51
Theorem 7.5 (Biholomorphisms of D). Let f : D D be biholomorphic. Then there
exist w D and c D with f(z) = c
w
(z) for z D.
Proof. Set w := f
1
(0). Then f
w
: D D is biholomorphic with (f
w
)(0) = 0.
By Corollary 7.4.1, there exists c C with [c[= 1 such that f(
w
(z)) = cz for z D,
so that
f(z) = f(
w
(
w
(z))) = c
w
(z)
for z D.
Chapter 8
The Singularities of a Holomorphic
Function
Denition. Let D C be open, and let f : D C be holomorphic. We call
z
0
CD an isolated singularity of f if there exists > 0 such that B
(z
0
)z
0
D.
We say that the isolated singularity z
0
is removable if there exists a holomorphic
function g : D z
0
C such that g[
D
= f.
Theorem 8.1 (Riemanns Removability Condition). Let D C be open, let f :
D C be holomorphic, and let z
0
C D be an isolated singularity of f. Then the
following are equivalent:
(i) z
0
is removable;
(ii) there is a continuous function g : D z
0
C such that g[
D
= f;
(iii) there exists > 0 with B
(z
0
)z
0
D such that f is bounded on B
(z
0
)z
0
.
Proof. (i) = (ii) follows from the continuity of a dierentiable function.
(ii) = (iii) follows from the boundedness of g on a compact set B
[z
0
] D.
(iii) = (i): Let C 0 be such that [f(z)[ C for z B
(z
0
) z
0
. Dene
h: D z
0
C, z
_
(z z
0
)
2
f(z), z ,= z
0
,
0, z = z
0
.
Then we have for z B
(z
0
) z
0
that
h(z) h(z
0
)
z z
0
= [(z z
0
)f(z)[ C[z z
0
[.
Hence, h is holomorphic with h
(z
0
) = h(z
0
) = 0. Let h(z) =
n=0
a
n
(z z
0
)
n
be the power series representation of h on B
(z
0
). Then h
(z
0
) = h(z
0
) = 0 means
that a
0
= a
1
= 0, so that h(z) =
n=2
a
n
(z z
0
)
n
for z B
(z
0
) and thus f(z) =
n=0
a
n+2
(z z
0
)
n
for z B
(z
0
) z
0
.
52
53
Dene
g : D z
0
C, z
_
n=0
a
n+2
(z z
0
)
n
, z B
(z
0
),
f(z), z D B
(z
0
).
Then g is a holomorphic function extending f.
Denition. Let D C be open, let f : D C be holomorphic, and let z
0
C D
be an isolated singularity of f. Then z
0
is called a pole of f if lim
zz
0
[f(z)[= .
Example. For n N, the function
C 0 C, z
1
z
n
has a pole at 0.
Theorem 8.2 (Poles). Let D C be open, let f : D C be holomorphic, and let
z
0
C D be an isolated singularity of f. Then z
0
is a pole of f there exist a
unique k N and a holomorphic function g : D z
0
C such that g(z
0
) ,= 0 and
f(z) =
g(z)
(z z
0
)
k
for z D.
Proof.
This follows directly from the denition of a pole.
Let us prove the uniqueness rst. Suppose that there exist natural numbers
k
1
k
2
and holomorphic functions g
1
, g
2
: D z
0
C such that g
j
(z
0
) ,= 0 and
f(z) =
g
j
(z)
(z z
0
)
k
j
for z D and j = 1, 2. If k
2
> k
1
, we then nd for all z D that
g
2
(z
0
) = lim
zz
0
g
2
(z) = lim
zz
0
(z z
0
)
k
2
k
1
g
1
(z) = 0 g
1
(z
0
) = 0,
which is a contradiction. Hence k
1
= k
2
and thus g
1
= g
2
on D and, by continuity,
on D z
0
.
To establish the existence of k and g, choose r > 0 such that B
r
(z
0
) z
0
D
and [f(z)[ 1 for all z B
r
(z
0
) z
0
. Then
B
r
(z
0
) z
0
C, z
1
f(z)
54 CHAPTER 8. THE SINGULARITIES OF A HOLOMORPHIC FUNCTION
is holomorphic and bounded and thus, by Riemanns Removability Criterion, has a
holomorphic extension h: B
r
(z
0
) C with h(z
0
) = lim
zz
0
1/f(z) = 0. Note that z
0
is
the only zero of h. Let
h(z) =
n=0
a
n
(z z
0
)
n
for z B
r
(z
0
) be the power series representation of h. Set k := minn N
0
: a
n
,= 0.
Since a
0
= h(z
0
) = 0, we have k 1. Dene
h: B
r
(z
0
) C, z
n=k
a
n
(z z
0
)
nk
.
Then
h is holomorphic, has no zeros, and satises h(z) = (z z
0
)
k
h(z) for z B
r
(z
0
).
For z B
r
(z
0
) z
0
, we thus have
f(z) =
1
(z z
0
)
k
h(z)
,
so that we can construct the holomorphic function
g : D z
0
C, z
_
(z z
0
)
k
f(z), z ,= z
0
,
1
h(z
0
)
, z = z
0
.
Denition. Let D C be open, let f : D C be holomorphic, and let z
0
C D
be a pole of f. Then the positive integer k in Theorem 8.2 is called the order of z
0
and denoted by ord(f, z
0
). If ord(f, z
0
) = 1, we call z
0
a simple pole of f.
Example. For m N, consider
f
m
: C 0 C, z
sin z
z
m
.
We claim that f
1
has a removable singularity at 0 whereas f
m
has a pole of order
m1 at 0 for m 2.
Recall that
sin z =
n=0
(1)
n
z
2n+1
(2n + 1)!
for z C. For z ,= 0, we thus have
f
1
(z) =
sin z
z
=
n=0
(1)
n
z
2n
(2n + 1)!
.
55
Dene
g : C C, z
n=0
(1)
n
z
2n
(2n + 1)!
.
Then g is holomorphic and extends f
1
. Hence, f
1
has a removable singularity at 0.
For m 2 and z ,= 0, note that f
m
(z) =
g(z)
z
m1
. Since g(0) ,= 0, we see that f
m
has
a pole of order m1 at 0.
Example. Consider
f : C 0 C, z e
1
z
.
Then 0 is not removable because lim
n
f
_
1
n
_
= lim
n
e
n
= . But 0 is not a pole
for f either: for n N, we have
f
_
i
n
_
= [e
in
[= 1.
Denition. Let D C be open, let f : D C be holomorphic, and let z
0
C D
be an isolated singularity of f. Then z
0
is called essential if it is neither removable
nor a pole.
Theorem 8.3 (CasoratiWeierstra Theorem). Let D C be open, let f : D C be
holomorphic, and let z
0
C D be an isolated singularity of f. Then z
0
is essential
f(B
(z
0
) D) = C for each > 0.
Proof. For each n N choose z
n
B1
n
(z
0
) D such that [f(z
n
) n[<
1
n
. It
follows that lim
n
[f(z
n
)[= . Hence, z
0
cannot be removable.
For each n N, choose z
n
B1
n
(z
0
) D such that [f(z
n
)[<
1
n
. This means that
lim
n
f(z
n
) = 0, so that z
0
is not a pole either.
Assume for some
0
> 0 that f(B
0
(z
0
) D) ,= C. Without loss of generality,
suppose that B
0
(z
0
) z
0
D. Let w
0
C and > 0 be such that B
(w
0
)
C f(B
0
(z
0
) z
0
). Consider
g : B
0
(z
0
) z
0
C, z
1
f(z) w
0
.
Then g is holomorphic with
[g(z)[=
1
[f(z) w
0
[
1
for z B
0
(z
0
) z
0
. Hence, z
0
is a removable singularity of g. Let g : B
0
(z
0
) C
be a holomorphic extension of g.
Case 1: g(z
0
) ,= 0. Since f(z) =
1
g(z)
+ w
0
for z B
0
(z
0
) z
0
, this means
that z
0
is a removable singularity of f, contradicting the fact that z
0
is an essential
singularity.
56 CHAPTER 8. THE SINGULARITIES OF A HOLOMORPHIC FUNCTION
Case 2: g(z
0
) = 0. For z B
0
(z
0
), we have
[f(z)[=
1
g(z)
+ w
0
1
[ g(z)[
[w
0
[
zz
0
.
Hence, z
0
is a pole of f, again contradicting the fact that z
0
is an essential singularity.
Problem 8.1.
Let D C be open, let f : D C be holomorphic, and let z
0
D. Show that the
following are equivalent for n N:
(i) f
(k)
(z
0
) = 0 for k = 0, . . . , n 1 and f
(n)
(z
0
) ,= 0;
(ii) there exists a holomorphic function g : D C with g(z
0
) ,= 0 such that f(z) =
(z z
0
)
n
g(z) for z D.
If either condition holds, we say that z
0
is a zero of f of order n.
Problem 8.2.
Let D C be open, let f, g : D C be holomorphic, and let z
0
D be a zero of
order n for f and of order m 1 for g. Show the singularity z
0
of
f
g
is
(i) removable if m n and
(ii) a pole of order mn otherwise.
Problem 8.3.
Let D C be open, let f : D C be holomorphic, and let z
0
C D be an isolated
singularity of f.
(a) Show that, if z
0
is a pole of order k of f, then it is a pole of order k + 1 of f
.
(b) Show that exp f has either a removable or an essential singularity at z
0
.
Chapter 9
Holomorphic Functions on Annuli
Denition. Let z
0
C, and let r, R [0, ] be such that r < R. Then the annulus
centered at z
0
with inner radius r and outer radius R is dened as
A
r,R
(z
0
) := z C : r < [z z
0
[< R.
Theorem 9.1 (Cauchys Integral Theorem for Annuli). Let z
0
C, let r, , P, R
[0, ] be such that r < < P < R, and let f : A
r,R
(z
0
) C be holomorphic. Then
we have _
B
P
(z
0
)
f() d =
_
B(z
0
)
f() d.
Proof. The claim is equivalent to
_
B
P
(z
0
)
f() d +
_
B(z
0
)
f() d = 0.
Consider
z
0
r
P
R
57
58 CHAPTER 9. HOLOMORPHIC FUNCTIONS ON ANNULI
Split B
P
(z
0
) and B
(z
0
)
3
We thus obtain
_
B
P
(z
0
)
f() d +
_
B(z
0
)
f() d =
n
j=1
_
j
f() d +
n
j=1
_
j
f() d
=
n
j=1
_
j
j
(j+1)mod n
f() d
By making the arc segments
1
, . . . ,
n
and
1
, . . . ,
n
suciently small, we can ensure
that each of the closed curves
1
1
1
2
, . . . ,
n1
n1
n1
n
,
n
n
n
1
lies inside a starshaped open subset of A
r,R
(z
0
):
59
It follows that
_
B
P
(z
0
)
f() d +
_
B(z
0
)
f() d
=
n
j=1
_
j
j
(j+1)mod n
f() d = 0
as claimed.
Theorem 9.2 (Laurent Decomposition). Let z
0
C, let r, R [0, ] be such that
r < R, and let f : A
r,R
(z
0
) C be holomorphic. Then there exists a holomorphic
function
g : B
R
(z
0
) C and h: C B
r
[z
0
] C
with f = g + h on A
r,R
(z
0
). Moreover, h can be chosen such that lim
z
h(z) = 0, in
which case g and h are uniquely determined.
Proof. We prove the uniqueness assertion rst.
Let g
1
, g
2
: B
R
(z
0
) C and h
1
, h
2
: C B
r
[z
0
] C be holomorphic such that
lim
z
h
j
(z) = 0 for j = 1, 2 and
f = g
1
+ h
1
= g
2
+ h
2
.
It follows that g
1
g
2
= h
2
h
1
on A
r,R
(z
0
). Dene
F : C C, z
_
g
1
(z) g
2
(z), z B
R
(z
0
),
h
2
(z) h
1
(z), z C B
r
[z
0
].
60 CHAPTER 9. HOLOMORPHIC FUNCTIONS ON ANNULI
Then F is entire with lim
z
[F(z)[= lim
z
[h
2
(z) h
1
(z)[= 0. Hence, F is bounded
and entire and thus constant by Liouvilles theorem. Since lim
z
[F(z)[= 0, this means
that F 0, so that g
1
= g
2
and h
1
= h
2
.
To show that g and h exists, for z A
r,R
(z
0
) choose and P such that
r < < [z z
0
[< P < R.
Dene
G: A
r,R
(z
0
) C,
_
f()f(z)
z
, ,= z,
f
(z), = z.
Then G is certainly holomorphic on A
r,R
(z
0
)z. Because it is continuous on A
r,R
(z
0
),
Riemanns Removability Criterion implies that G is in fact holomorphic on all of A
r,R
(z
0
).
It follows from Cauchys Integral Theorem for Annuli that
_
B(z
0
)
G() d =
_
B
P
(z
0
)
G() d,
i.e.
_
B(z
0
)
f()
z
d f(z)
_
B(z
0
)
1
z
d
. .
=0
=
_
B
P
(z
0
)
f()
z
d f(z)
_
B
P
(z
0
)
1
z
d
. .
=2i
,
Let us dene the holomorphic functions (cf. Lemma 5.4)
h(z) :=
1
2i
_
B(z
0
)
f()
z
d
on C B
[z
0
] and
g(z) :=
1
2i
_
B
P
(z
0
)
f()
z
d
on B
P
(z
0
), noting from Cauchys Integral Theorem for Annuli that these denitions
are independent of the precise choices of (r, [z z
0
[) and P ([z z
0
[, R). Then
the above result may be expressed as
2i h(z) = 2i g(z) 2i f(z)
and thus
f(z) = g(z) + h(z).
Finally, we note that h satises
[h(z)[ sup
B(z
0
)
f()
z
sup
B(z
0
)
[f()[
dist(z, B
(z
0
))
z
0.
61
Denition. The function h in Theorem 9.2 is called the principal part and g is called
the secondary part of the Laurent decomposition f = g + h.
Theorem 9.3 (Laurent Coecients). Let z
0
C, let r, R [0, ] be such that
r < R, and let f : A
r,R
(z
0
) C be holomorphic. Then f has a representation
f(z) =
n=
a
n
(z z
0
)
n
for z A
r,R
(z
0
) as a Laurent series, which converges uniformly and absolutely on
compact subsets of A
r,R
(z
0
). Moreover, for every n Z and (r, R), the coe
cients a
n
are uniquely determined as
a
n
=
1
2i
_
B(z
0
)
f()
( z
0
)
n+1
d.
Proof. Let g and h be as in Theorem 9.2 (in particular, with lim
z
h(z) = 0).
For z B
R
(z
0
), we have the Taylor series
g(z) =
n=0
a
n
(z z
0
)
n
,
which converges uniformly and absolutely on compact subsets of B
R
(z
0
).
Dene
h: A
0,
1
r
(0) C, z h
_
z
0
+
1
z
_
,
so that
h is holomorphic with lim
z0
h(z) = 0. Hence,
h has a removable singularity at 0
and thus extends to B1
r
(0) as a holomorphic function. This holomorphic function,
which we also denote by
h, can then be expanded in a Taylor series for z B1
r
(0):
h(z) =
n=0
b
n
z
n
=
n=1
b
n
z
n
,
so that
h(z) =
n=1
b
n
(z z
0
)
n
converges uniformly and absolutely on compact subsets of C B
r
[z
0
].
Set a
n
:= b
n
for n < 0. For z A
r,R
(z
0
), we obtain
f(z) = g(z) + h(z) =
n=0
a
n
(z z
0
)
n
+
n=1
a
n
(z z
0
)
n
=
n=
a
n
(z z
0
)
n
.
62 CHAPTER 9. HOLOMORPHIC FUNCTIONS ON ANNULI
Finally, pick m Z and (r, R). Note that
f(z)
(z z
0
)
m+1
=
n=
a
n
(z z
0
)
nm1
=
n=
a
n+m+1
(z z
0
)
n
converges uniformly on B
(z
0
). Hence, we nd
_
B(z
0
)
f()
( z
0
)
m+1
d =
n=
a
n+m+1
_
B(z
0
)
(z
0
)
n
d = a
m
_
B(z
0
)
1
z
0
d = 2i a
m
,
noting that ( z
0
)
n
has an antiderivative for all n ,= 1. Thus
a
m
=
1
2i
_
B(z
0
)
f()
( z
0
)
m+1
d
Corollary 9.3.1. Let z
0
C, let r > 0, and let f : B
r
(z
0
) z
0
C be holomorphic
with Laurent representation f(z) =
n=
a
n
(z z
0
)
n
. Then the singularity z
0
of f
is
(i) removable if and only if a
n
= 0 for n < 0;
(ii) a pole of order k N if and only if a
k
,= 0 and a
n
= 0 for all n < k;
(iii) essential if and only if a
n
,= 0 for innitely many n < 0.
Proof.
(i) The if part follows from Theorem 6.3.
Conversely, suppose that z
0
is a removable singularity, and let
f : B
r
(z
0
) C
be a holomorphic extension of f with Taylor expansion
f(z) =
n=0
b
n
(z z
0
)
n
for z B
r
(z
0
). The uniqueness of the Laurent representation yields a
n
= b
n
for
n N
0
and a
n
= 0 for n < 0.
(ii) For the if part, set
g(z) := (z z
0
)
k
f(z) =
n=k
a
n
(z z
0
)
n+k
for z B
r
(z
0
) z
0
. Then g extends holomorphically to B
r
(z
0
) with g(z
0
) =
a
k
,= 0. By denition, we have f(z) =
g(z)
(zz
0
)
k
for z B
r
(z
0
) z
0
. Hence, f
has a pole of order k at z
0
.
63
For the converse, let g : B
r
(z
0
) C be holomorphic such that g(z
0
) ,= 0 and
f(z) =
g(z)
(zz
0
)
k
for z B
r
(z
0
) z
0
. Let g(z) =
n=0
b
n
(z z
0
)
n
for z B
r
(z
0
)
be the Taylor series of g, so that
f(z) =
n=0
b
n
(z z
0
)
nk
for z B
r
(z
0
) z
0
. The uniqueness of the Laurent representation yields that
a
n
= b
n+k
for n k and a
n
= 0 for n < k.
(iii) This follows from (i) and (ii) by elimination.
Examples.
1. Let
f : C 0 C, z e
1
z
2
.
Then f has the Laurent representation
f(z) =
n=0
(1)
n
n!
1
z
2n
for z C 0 and thus has an essential singularity at 0.
2. Let
f : C 0 C, z
e
z
1
z
3
,
so that
f(z) =
1
z
3
n=1
z
n
n!
=
n=1
z
n3
n!
for z C 0. Hence, f has a pole of order two at 0.
Remark. The Laurent representation of a holomorphic function on an annulus A
r,R
(z
0
)
depends not only on z
0
, but also on r and R.
Example. Consider the function
f : C 1, 3 C, z
2
z
2
4z + 3
,
and note that
f(z) =
1
1 z
1
3 z
.
Then f has the following Laurent representations:
64 CHAPTER 9. HOLOMORPHIC FUNCTIONS ON ANNULI
(a) On A
0,1
(0): For [z[< 1, we have
1
1 z
=
n
n=0
z
n
and, for [z[< 3,
1
3 z
=
1
3
_
1
z
3
_ =
1
3
n=0
_
z
3
_
n
.
We thus have for z A
0,1
(0) that
f(z) =
n=0
_
1
1
3
n+1
_
z
n
.
(b) On A
1,3
(0): For [z[> 1, we have
1
1 z
=
1
z 1
=
1
z
_
1
1
z
_ =
n=0
1
z
n+1
,
so that, for z A
1,3
(0):
f(z) =
_
n=1
1
z
n
+
n=0
z
n
3
n+1
_
.
(c) On A
3,
(0): For [z[> 3, we have
1
3 z
=
1
z 3
=
1
z
_
1
3
z
_ =
n=0
3
n
z
n+1
and thus, for z A
3,
(0):
f(z) =
n=1
(3
n1
1)
1
z
n
.
Chapter 10
The Winding Number of a Curve
Denition. Let be a closed curve in C, and let z C . Then the winding
number of with respect to z is dened as
(, z) :=
1
2i
_
1
z
d.
Remark. Geometrically, (, z) is the number of times winds around z in the coun
terclockwise direction.
Lemma 10.1. Let : [0, 1] C be a curve, and let z C . Then there exist
open discs D
1
, . . . , D
n
C z and a partition 0 = t
0
< t
1
< < t
n
= 1 such that
([t
j1
, t
j
]) D
j
for j = 1, . . . , n.
Proof. Let := dist(z, ) > 0. Since is uniformly continuous, there exists > 0
such that [(t) (t
[< . Choose 0 = t
0
<
t
1
< < t
n
= 1 such that [t
j1
t
j
[< for j = 1, . . . , n, and set D
j
:= B
((t
j
)) for
j = 1, . . . , n. By the choice of , it is clear that D
1
, . . . , D
n
Cz. For j = 1, . . . , n,
let t [t
j1
, t
j
], and note that [t t
j
[ [t
j1
t
j
[< , so that [(t) (t
j
)[< , i.e.
(t) D
j
; consequently, ([t
j1
, t
j
]) D
j
holds.
Proposition 10.1. Let be a closed curve in C, and let z C . Then
(, z) Z.
Proof. Choose a partition 0 = t
0
< t
1
< < t
n
= 1 and open discs D
1
, . . . , D
n
such
that ([t
j1
, t
j
]) D
j
for j = 1, . . . , n.
Let j 1, . . . , n. Since z / D
j
, there exists a holomorphic function L
j
: D
j
C
such that
e
L
j
(w)
= w z for w D
j
.
On noting that (t
n
) = (t
0
), it is convenient to denote D
n+1
:= D
1
and L
n+1
:= L
1
.
For j = 1, . . . , n we then see that (t
j
) D
j
D
j+1
and hence
exp(L
j
((t
j
))) = (t
j
) z = exp(L
j+1
((t
j
))),
65
66 CHAPTER 10. THE WINDING NUMBER OF A CURVE
so that
exp(L
j
((t
j
)) L
j+1
((t
j
))) = 1
and thus
L
j
((t
j
)) L
j+1
((t
j
)) 2i Z.
On dierentiating e
L
j
(w)
= w z, we nd that e
L
j
(w)
L
j
(w) = 1. Thus
L
j
(w) =
1
w z
for w D
j
;
this allows us to express
_
1
z
d =
n
j=1
_

[t
j1
,t
j
]
1
z
d
=
n
j=1
[L
j
((t
j
)) L
j
((t
j1
))]
=
n
j=1
L
j
((t
j
))
n1
j=0
L
j+1
((t
j
))
= L
n
((t
n
)) L
1
((t
0
)) +
n1
j=1
[L
j
((t
j
)) L
j+1
((t
j
))].
= L
n
((t
n
)) L
n+1
((t
n
)) +
n1
j=1
[L
j
((t
j
)) L
j+1
((t
j
))].
=
n
j=1
[L
j
((t
j
)) L
j+1
((t
j
))].
We thus see that
_
1
z
d 2i Z.
Denition. Let be a closed curve in C. We dene the interior and exterior of
to be
int := z C : (, z) ,= 0
and
ext := z C : (, z) = 0.
67
Proposition 10.2 (Winding Numbers Are Locally Constant). Let be a closed curve
in C. Then:
(i) the map
C C, z (, z)
is locally constant;
(ii) there exists R > 0 such that C B
R
[0] ext .
Proof. (i): Let z
0
C and choose R > r > 0 such that B
R
(z
0
) C .
Consider the function
F : B
r
[z
0
] C, (, z)
1
z
.
Then F is continuous and thus uniformly continuous. Choose (0, r) such that
z B
(z
0
), [F(, z) F(, z
0
)[<
() + 1
.
Then
[(, z) (, z
0
)[ =
1
2i
_
_
1
z
1
z
0
_
d
()
2
sup
{}
[F(, z) F(, z
0
)[
()
2
() + 1
<
1
2
.
Since (, z) (, z
0
) Z, this means that (, z) = (, z
0
).
(ii): For any z C , we have
[(, z)[=
1
2i
_
1
z
d
()
2
1
dist(z, )
.
Since lim
z
dist(z, ) = , there exists R > 0 such that [(, z)[
()
2
1
dist(z,{})
< 1
for all z C such that [z[> R. Since (, z) Z for all z C , this implies that
(, z) = 0 for all z C with [z[> R.
Chapter 11
The General Cauchy Integral
Theorem
Denition. Let D C be open. We call a closed curve in D homologous to zero
if (, z) = 0 for each z C D. That is, the interior of is a subset of D.
Denition. An open connected subset D of C is simply connected if every closed
curve in D is homologous to zero. Equivalently, the interior of every closed curve in
D is a subset of D.
Theorem 11.1 (Cauchys Integral Formula). Let D C be open, let f : D C be
holomorphic, and let be a closed curve in D that is homologous to zero. Then, for
n N
0
and z D , we have
(, z)f
(n)
(z) =
n!
2i
_
f()
( z)
n+1
d.
Proof. It is enough to prove the claim for n = 0: for n 1, dierentiate the integral
with respect to z and use induction.
Dene
g : D D C, (w, z)
_
f(w)f(z)
wz
, w ,= z,
f
(z), w = z.
We claim that g is continuous. To see this, let (w
0
, z
0
) D D. As g is clearly
continuous at (w
0
, z
0
) if w
0
,= z
0
, we need only show that g is continuous at (z
0
, z
0
).
Given > 0, choose > 0 small enough that B
[z
0
] D and [f
(z) f
(z
0
)[< for
all z B
[z
0
]. For (w, z) B
(z
0
) B
(z
0
) we nd:
if w = z:
[g(w, z) g(z
0
, z
0
)[= [f
(z) f
(z
0
)[< ;
68
69
if w ,= z:
[g(w, z) g(z
0
, z
0
)[ =
f(w) f(z)
w z
f
(z
0
)
1
w z
_
[z,w]
[f
() f
(z
0
)] d
sup
{[z,w]}
[f
() f
(z
0
)[ .
Thus, g is continuous.
Next, dene
h
0
: D C, z
_
g(, z) d.
We claim that h
0
is holomorphic. It is easy to see that h
0
is continuous. To see that
it is indeed holomorphic, we shall show that it satises the Morera condition. Let
D be a triangle. For xed , the function
D C, z g(, z)
is holomorphic as a consequence of Riemanns Removability Condition. Goursats Lemma
thus yields
_
g(, z) dz = 0
for each . As a consequence, we nd
0 =
_
__
g(, z) dz
_
d
=
_
__
g(, z) d
_
dz
=
_
h
0
(z) dz,
so that h
0
is holomorphic as claimed.
Dene
h
1
: ext C, z
_
f()
z
d.
70 CHAPTER 11. THE GENERAL CAUCHY INTEGRAL THEOREM
Then h
1
is holomorphic. For z D ext , we note that
h
0
(z) =
_
g(, z) d
=
_
f() f(z)
z
d
=
_
f()
z
d f(z)
_
1
z
d
. .
=0
=
_
f()
z
d
= h
1
(z).
Dene
h: D ext , z
_
h
0
(z), z D,
h
1
(z), z ext .
Then h is holomorphic. Since is homologous to zero, we have C D ext .
Hence, h is entire.
For any z ext , we have the estimate
[h(z)[= [h
1
(z)[
()
dist(z, )
sup
{}
[f()[. ()
Let R > 0 be such that C B
R
(0) ext . Since () implies that h is bounded
on C B
R
(0) and h is trivially bounded by continuity on B
R
[0], we see that h is
bounded on C and hence constant by Liouvilles Theorem. From () again, we see
that lim
z
[h(z)[= 0. Hence, h 0.
In summary, we have for z D that
0 = h(z) = h
0
(z) =
_
f() f(z)
z
d =
_
f()
z
d 2i(, z)f(z).
Theorem 11.2 (Cauchys Integral Theorem). Let D C be open, let f : D C
be holomorphic, and let be a closed curve in D that is homologous to zero. Then
_
f() d = 0.
Proof. Let z
0
D be arbitrary, and dene
g : D C, z (z z
0
)f(z),
so that
0 = 2i(, z
0
)g(z
0
) =
_
g()
z
0
d =
_
f() d.
71
Corollary 11.2.1. Let D be an open, connected subset of C. Then D is simply
connected every holomorphic function on D has an antiderivative.
Problem 11.1. Let D C be open and connected such that, for each holomorphic
f : D C, there is a sequence (p
n
)
n=1
of polynomials converging to f compactly
on D. Show that D is simply connected.
Denition. Let D C be open and connected and n N. We say that D admits
(a) holomorphic logarithms if, for every holomorphic function f : D C with
Z(f) = , there exists a holomorphic function g : D C with f = exp g;
(b) holomorphic nth roots if for every holomorphic function f : D C with Z(f) = ,
there exists a holomorphic function h
n
: D C with f(z) = [h
n
(z)]
n
for z D;
(c) holomorphic roots if D admits holomorphic nth roots for each n N.
Corollary 11.2.2 (Holomorphic Logarithms). A simply connected domain admits
holomorphic logarithms.
Proof. This follows from Corollary 11.2.1 and Problem 5.1(a).
Corollary 11.2.3 (Holomorphic Roots). A simply connected domain admits holo
morphic roots.
Proof. Let g be such that f = exp g, and set h
n
:= exp
_
g
n
_
for n N.
Chapter 12
The Residue Theorem and
Applications
Denition. Let z
0
C, let r > 0, and let f : B
r
(z
0
) z
0
C be holomorphic with
Laurent series representation
f(z) =
n=
a
n
(z z
0
)
n
for z B
r
(z
0
) z
0
. Then a
1
is called the residue of f at z
0
and denoted by
res(f, z
0
).
Remarks. 1. By Theorem 9.3, we have
res(f, z
0
) =
1
2i
_
B(z
0
)
f() d
for any (0, r).
2. If f has a removable singularity at z
0
, then res(f, z
0
) = 0.
3. Suppose that f has a simple pole at z
0
, i.e.
f(z) =
n=1
a
n
(z z
0
)
n
with a
1
,= 0, then
res(f, z
0
) = lim
zz
0
(z z
0
)f(z).
4. Suppose that f(z) =
n=k
a
n
(z z
0
)
n
has a pole of order k at z
0
. On letting
g(z) = (z z
0
)
k
f(z), we see that res(f, z
0
) is the coecient in the Taylor series
of g(z) =
n=k
a
n
(zz
0
)
n+k
=
n=0
a
nk
(zz
0
)
n
corresponding to n = k1:
res(f, z
0
) =
g
(k1)
(z
0
)
(k 1)!
=
1
(k 1)!
d
k1
dz
k1
_
(z z
0
)
k
f(z)
z=z
0
.
72
73
Examples.
1. Let
f(z) =
e
iz
z
2
+ 1
,
so that f has a simple pole at z
0
= i. It follows that
res(f, i) = lim
zi
(z i)f(z) = lim
zi
e
iz
z + i
=
i
2e
.
2. Let
f(z) =
cos(z)
sin(z)
,
so that f has a simple pole at each n Z. For n Z, we thus have:
res(f, n) = lim
zn
(z n)
cos(z)
sin(z)
= lim
zn
(z n)
cos(z)
sin(z) sin(n)
=
1
lim
zn
z n
sin(z) sin(n)
cos(z)
=
1
.
3. Let
f(z) =
1
(z
2
+ 1)
3
;
then f has a pole of order 3 at z
0
= i. With
g(z) = (z i)
3
f(z) =
1
(z + i)
3
,
we have
g
(z) =
3
(z + i)
4
and g
(z) =
12
(z + i)
5
,
so that
res(f, i) =
1
2
12
(2i)
5
=
3i
16
.
Theorem 12.1 (Residue Theorem). Let D C be open and simply connected,
z
1
, . . . , z
n
D be such that z
j
,= z
k
for j ,= k, f : D z
1
, . . . , z
n
C be holo
morphic, and be a closed curve in D z
1
, . . . , z
n
. Then we have
_
f() d = 2i
n
j=1
(, z
j
) res(f, z
j
).
74 CHAPTER 12. THE RESIDUE THEOREM AND APPLICATIONS
Proof. Let > 0 be such that B
(z
j
) D for j = 1, . . . , n, with z
k
/ B
(z
j
) for k ,= j.
For j = 1, . . . , n, we have Laurent representations
f(z) =
k=
a
(j)
k
(z z
j
)
k
for z B
(z
j
) z
j
, so that res(f, z
j
) = a
(j)
1
. For j = 1, . . . , n, dene
h
j
: C z
j
C, z
1
k=
a
(j)
k
(z z
j
)
k
,
so that h
j
is holomorphic on C z
j
. Dene
g : D z
1
, . . . , z
n
C, z f(z)
n
j=1
h
j
(z),
and note that z
1
, . . . , z
n
are removable singularities for g.
Since D is simply connected, Cauchys Integral Theorem yields:
0 =
_
g() d
=
_
f() d
n
j=1
_
h
j
() d
=
_
f() d
n
j=1
_
_
1
k=
a
(j)
k
( z
j
)
k
_
d
=
_
f() d
n
j=1
1
k=
a
(j)
k
_
( z
j
)
k
d
=
_
f() d
n
j=1
a
(j)
1
_
1
z
j
d
=
_
f() d
n
j=1
res(f, z
j
) 2i (, z
j
).
Corollary 12.1.1. Let D C be open and simply connected, f : D C be
holomorphic, and be a closed curve in D. Then we have
(, z) f(z) =
1
2i
_
f()
z
d
for z D .
12.1. APPLICATIONS OF THE RESIDUE THEOREMTO REAL INTEGRALS75
Proof. Fix z D , and dene
g : D z C, w
f(w)
w z
.
Then g is holomorphic with an isolated singularity at z. Let
f(w) =
n=0
a
n
(w z)
n
be the Taylor series expansion of f near z, so that
g(w) =
n=1
a
n+1
(w z)
n
,
and thus res(g, z) = a
0
= f(z). The Residue Theorem then yields:
2i (, z) f(z) = 2i (, z) res(g, z) =
_
g() d =
_
f()
z
d.
12.1 Applications of the Residue Theorem to Real
Integrals
Proposition 12.1 (Rational Trigonometric Polynomials). Let p and q be polynomials
of two real variables such that q(x, y) ,= 0 for all (x, y) R
2
with x
2
+ y
2
= 1. Then
we have
_
2
0
p(cos t, sin t)
q(cos t, sin t)
dt = 2i
zD
res(f, z),
where
f(z) =
1
iz
p
_
1
2
_
z +
1
z
_
,
1
2i
_
z
1
z
__
q
_
1
2
_
z +
1
z
_
,
1
2i
_
z
1
z
__.
Proof. Just note that, by the Residue Theorem,
2i
zD
res(f, z) =
_
D
f() d
=
_
2
0
f(e
i
)i e
i
d
=
_
2
0
p(cos , sin )
q(cos , sin )
d.
76 CHAPTER 12. THE RESIDUE THEOREM AND APPLICATIONS
Examples.
1. Let a > 1. What is
_
0
dt
a + cos t
?
First, note that
_
0
dt
a + cos t
=
1
2
_
dt
a + cos t
=
1
2
_
2
0
dt
a + cos t
.
Let
p(x, y) = 1 and q(x, y) = a + x,
so that
f(z) =
1
iz
1
a +
1
2
_
z +
1
z
_
=
i
az +
z
2
2
+
1
2
=
2i
z
2
+ 2az + 1
=
2i
(z z
1
)(z z
2
)
,
where
z
1
= a +
a
2
1 D and z
2
= a
a
2
1 / D, (12.1)
on noting that 1 +
a
2
1 > a implies that z
1
> 1.
By Proposition 12.1, we thus obtain
_
0
dt
a + cos t
=
1
2
_
2
0
dt
a + cos t
= i res(f, z
1
)
= i
2i
z
1
z
2
=
2
2
a
2
1
=
a
2
1
.
2. Let a > 0. What is
_
2
0
dt
(a + cos t)
2
?
Let
p(x, y) = 1 and q(x, y) = (a + x)
2
,
12.1. APPLICATIONS OF THE RESIDUE THEOREMTO REAL INTEGRALS77
so that
f(z) =
1
iz
1
_
a +
1
2
_
z +
1
z
__
2
=
4iz
(z
2
+ 2az + 1)
2
=
4iz
(z z
1
)
2
(z z
2
)
2
,
where z
1
and z
2
are again given by Eq. 12.1.
At z
1
, the function f has a pole of order two. In order to calculate res(f, z
1
),
set
g(z) := (z z
1
)
2
f(z) =
4iz
(z z
2
)
2
,
so that
g
(z) = 4i
_
1
(z z
2
)
2
2z
(z z
2
)
3
_
=
4i
(z z
2
)
3
[(z z
2
) 2z]
=
4i(z + z
2
)
(z z
2
)
3
;
it follows that
res(f, z
1
) = g
(z
1
)
=
4i(2a)
8
_
a
2
1
_
3
=
ai
_
a
2
1
_
3
.
From Proposition 12.1, we conclude that
_
2
0
dt
(a + cos t)
2
= 2i res(f, z
1
) =
2a
_
a
2
1
_
3
.
Problem 12.1.
Let D C be open, let f : D C be holomorphic, and let z
0
D be a zero of order
one of f. Show that
res
_
1
f
; z
0
_
=
1
f
(z
0
)
.
78 CHAPTER 12. THE RESIDUE THEOREM AND APPLICATIONS
Problem 12.2.
Let D C be open, let f : D C be holomorphic, and let z
0
C D be a simple
pole of f. Show that
res(gf; z
0
) = g(z
0
) res(f; z
0
)
for every holomorphic function g : D z
0
C.
Proposition 12.2 (Rational Functions). Let p and q be polynomials of one real
variable with deg q deg p + 2 and such that q(x) ,= 0 for x R. Then we have
_
p(x)
q(x)
dx = 2i
zH
res
_
p
q
, z
_
,
where
H := z C : Imz > 0.
Proof. Since deg q deg p+2, the Comparison Test yields that the indenite integral
exists.
For r > 0 consider the semicircle
r
: [0, ] C, re
i
.
Let > 0 be such that, for D := z C : Imz > , we have
z H : q(z) = 0 = z D : q(z) = 0.
Then D is simply connected and
p
q
is holomorphic on D except at the zeros of q in
H. For r large enough so that all zeros of q in D lie in the interior of [r, r]
r
, we
see by the Residue Theorem that
_
[r,r]r
p()
q()
d = 2i
zD
res
_
p
q
, z
_
= 2i
zH
res
_
p
q
, z
_
.
Since deg q deg p + 2, there exist numbers R > 0 and C 0 such that
p(z)
q(z)
C
[z[
2
for all z C with [z[ R. It follows that
_
r
p()
q()
d
r sup
{r}
C
[[
2
C
r
for r R and thus
lim
r
_
r
p()
q()
d = 0.
12.1. APPLICATIONS OF THE RESIDUE THEOREMTO REAL INTEGRALS79
We then nd that
_
p(x)
q(x)
dx = lim
r
_
[r,r]
p()
q()
d
= lim
r
_
[r,r]
p()
q()
d + lim
r
_
r
p()
q()
d
= lim
r
_
[r,r]r
p()
q()
d
= 2i
zH
res
_
p
q
, z
_
.
Examples.
1. What is
_
0
1
1 + x
6
dx?
The zeros of q(z) = 1 + z
6
are of the form e
i
where [0, 2) is such that
e
i6
= 1 = e
i
, i.e. 6 2 Z, so that =
6
,
2
,
5
6
,
7
6
,
3
2
,
11
6
. For
k = 1, . . . , 6, let
z
k
= e
i(2k1)
6
.
Then
1
q
has a simple pole at z
k
for k = 1, . . . , 6.
By Problem 12.1, we have
res
_
1
q
, z
k
_
=
1
q
(z
k
)
=
1
6z
5
k
=
z
k
6
,
so that by, Proposition 12.2,
_
0
1
1 + x
6
dx =
1
2
_
1
1 + x
6
dx
= i
3
k=1
res
_
1
q
, z
k
_
=
i
6
_
e
i
6
+ e
i
2
+ e
i
5
6
_
=
i
6
_
cos
6
+ i sin
6
+ i + cos
5
6
+ i sin
5
6
_
=
6
_
2 sin
6
+ 1
_
=
3
.
80 CHAPTER 12. THE RESIDUE THEOREM AND APPLICATIONS
2. What is
_
1
(x
2
+ 1)
n
dx, where n N?
The polynomial q(z) := (z
2
+ 1)
n
has zeros of order n at i. Dene
g(z) = (z i)
n
1
q(z)
= (z + i)
n
,
so that
g
(n1)
(z) = (n) (2n + 2)(z + i)
2n+1
and thus
res
_
1
q
, i
_
=
g
(n1)
(i)
(n 1)!
=
1
(n 1)! 2
2n1
i
n (2n 2)
=
(2n 2)!
i2
2n1
(n 1)!
2
.
It follows that
_
1
(x
2
+ 1)
n
dx = 2i res
_
1
q
, i
_
=
2
2n2
(2n 2)!
(n 1)!
2
;
in particular, we have
_
1
x
2
+ 1
dx = ,
_
1
(x
2
+ 1)
2
dx =
2
, and
_
1
(x
2
+ 1)
3
dx =
3
8
.
Problem 12.3.
(a) Prove that sin
2
for 0
2
.
(b) Use part (a) to show that for R > 0 that
_
0
e
Rsin
d <
R
.
(c) Let C
R
be the semicircular contour Re
i
: 0 , with R > 0. Use part (b)
to establish Jordans Lemma:
_
C
R
e
iz
dz
< .
Problem 12.4.
Let D be an open set. If f : D z
0
C is holomorphic, where f has a simple pole
at z
0
, and C
r
= z
0
+ re
i
: , prove the Fractional Residue Theorem:
lim
r0
_
Cr
f(z) dz = ( )i res(f, z
0
).
12.2. THE GAMMA FUNCTION 81
12.2 The Gamma Function
For Re(z) > 0, dene
+
(z) :=
_
0
+
e
t
t
z1
dt,
where the integration is performed along the positive real axis. Then
+
is holomor
phic in the right half plane z C : Re(z) > 0. A single integration by parts yields
the following recurrence relation
(12.2)
+
(z + 1) =
_
0
+
e
t
t
z
dt = e
t
t
z
0
+ z
_
0
+
e
t
t
z1
dt
= z
+
(z).
Since
+
(1) =
_
0
e
t
dt = 1 = 0!, we see that
+
(n+1) = n! for n N
0
. Continuing
in this manner we nd that
+
(z +n) = (z +n1) . . . (z +1)z
+
(z). On rearranging
this formula,
+
(z) =
+
(z + n)
z(z + 1) . . . (z + n 1)
,
it is possible to analytically continue the function to the lefthalf plane:
(z) :=
_
_
_
+
(z) Re(z) > 0,
+
(z + n)
z(z + 1) . . . (z + n 1)
n < Re(z) n + 1, z ,= n + 1, n = 1, 2, 3, . . .
The resulting function (z) is holomorphic in the complex plane except at z =
0, 1, 2, . . . , where it has simple poles. The graph of (x) for x R is shown
in Figure 12.1 and an interactive threedimensional plot of the surface (z) for z C
is shown in Figure 12.2.
We proceed to derive a few useful relationships involving the function.
For (0, 1) we have
() =
_
0
+
e
t
t
1
dt = 2
_
0
+
e
y
2
y
21
dy (letting t = y
2
),
which leads to
()(1 ) =
_
2
_
0
+
e
y
2
y
21
dy
__
2
_
0
+
e
x
2
x
12
dx
_
= 4
_
0
+
_
0
+
e
(x
2
+y
2
)
_
y
x
_
21
dx dy
= 4
_
/2
0
tan
21
_
0
e
r
2
r dr d
= 2
_
/2
0
tan
21
d.
82 CHAPTER 12. THE RESIDUE THEOREM AND APPLICATIONS
In particular, we see for = 1/2 that
2
(1/2) = 2
_
/2
0
d = and
_
e
x
2
dx = 2
_
0
e
x
2
dx =
_
1
2
_
=
.
A substitution then leads to the important result
_
e
ax
2
dx =
_
/a for
a > 0.
For arbitrary (0, 1), we nd, on substituting z = tan
2
,
I() := ()(1 ) = 2
_
/2
0
+
tan
21
d =
_
0
+
z
1
1 + z
dz.
The integral here can be evaluated by a contour integration in the complex
plane, noting that the function z
1
= e
(1) log z
is holomorphic on the star
shaped domain obtained by slicing the complex plane along the positive real
axis. This branch cut is shown in red in the following gure. In other words we
choose the antiderivative log z = log [z[ +i arg z of the function z 1/z, where
arg z [0, 2).
Imz
Re z
ir
1
iR
C
R
C
r
Here the large circular contour C
R
is chosen to have radius R 2, so that
[1 + z[ R/2 on C
R
, and the small semicircular contour C
r
is chosen to have
radius r 1/2, so that [1 + z[ 1/2 on C
r
. On denoting
f(z) :=
z
1
1 + z
=
e
(1) log z
1 + z
,
we then see, accounting for the residue from the pole of f at z = 1, that
2ie
(1)i
=
_
R+ir
ir
f +
_
C
R
f +
_
ir
Rir
f +
_
Cr
f.
12.2. THE GAMMA FUNCTION 83
Since < 1, we see that the contribution from the circular arc C
R
is
_
C
R
f
R
1
R
2
2R = 4R
1
R
0.
Likewise, since > 0, the contribution from the semicircular contour C
r
is
_
Cr
f
r
1
1
2
r = 2r
r0
0.
We thus deduce that
2ie
(1)i
= lim
r0
R
__
R+ir
ir
f
_
Rir
ir
f
_
=
_
0
+
e
(1) logz
1 + z
dz
_
0
+
e
(1)(logz+i2)
1 + z
dz
= I()
_
1 e
(1)2i
_
.
Thus
= I()
e
(1)i
e
(1)i
2i
= I()
e
i
+ e
i
2i
,
from which we see that
I() = ()(1 ) =
sin
,
On extending this result by analytic continuation, one nds for all z C Z
that
(z)(1 z) =
sin z
.
For 1 and positive x and , another frequently encountered integral can be
expressed in terms of using the substitution u = xt
:
_
0
e
xt
t
1
dt =
1
x
_
0
e
u
u
1
du =
_
x
. (12.3)
For the special case = x = 1, this result simplies to
_
0
e
t
dt =
1
_
1
_
=
_
1 +
1
_
.
For 0 < < 1 and x > 0, a related integral is
(12.4)
_
0
e
ixt
t
1
dt =
i
()
x
.
To establish this result, it is convenient to introduce a branch cut, shown in
red, along the negative real axis:
84 CHAPTER 12. THE RESIDUE THEOREM AND APPLICATIONS
Imz
Re z
r
R
ir
iR
C
R
C
r
We note that f(z) = e
ixz
z
1
is holomorphic inside the blue contour. Cauchys
Integral Theorem thus implies that
0 =
_
R
r
f(t) dt +
_
C
R
f + i
_
r
R
f(it) dt +
_
Cr
f.
Since < 1, we see on using Problem 12.3 (a) that
_
C
R
f
_
/2
0
e
xRsin
R
1
R d
R
1
_
/2
0
e
2xR/
R d = R
1
2x
_
1 e
xR
_
R
0.
Likewise, since > 0, we see that
_
Cr
f
2x
_
1 e
xr
r
_
r0
0.
Hence
_
0
f(t) dt =
_
0
f(it) idt = i
_
0
e
xt
t
1
dt =
i
()
x
,
as claimed.
12.2. THE GAMMA FUNCTION 85
6
5
4
3
2
1
1
2
3
4
5
6
4 3 2 1 1 2 3 4
x
(x)
Figure 12.1: Graph of (x) on the real line.
86 CHAPTER 12. THE RESIDUE THEOREM AND APPLICATIONS
Figure 12.2: Surface plot of (z) in the complex plane, using an RGB color wheel
to represent the phase. Red indicates real positive values. The poles at the negative
integers and 0 are evident.
Chapter 13
Function Theoretic Consequences
of the Residue Theorem
Denition. Let D C be open. We call S D discrete in D if it has no cluster
points in D.
Example. If D is open and connected, and f : D C is holomorphic and not identi
cally zero, then Z(f) is discrete.
Remark. Let S D be discrete, and let K D be compact. If K S were innite,
then K S would have cluster points, which would lie in K D. Thus, K S must
be nite.
Remark. Let S be a discrete subset of an open set D. For suitably small radii r(z) > 0,
we note that D can be expressed as a countable union of compact sets:
D =
_
zDQ
2
B
r(z)
[z].
On denoting these compact sets as K
n
n=1
, we see that each K
n
S is nite. Thus
S =
_
n=1
(K
n
S).
is either nite or countably innite.
Example. If D is open and connected, and f : D C is holomorphic and not identi
cally zero, then Z(f) is at most countably innite.
Proposition 13.1. Let D C be open, let be a closed curve in D, and let S D
be discrete. Then S int is nite.
Proof. By Proposition 10.2(ii), there exists R > 0 such that int B
R
[0].
Denition. Denote the set of poles of a holomorphic function f by P(f).
87
88CHAPTER 13. FUNCTIONTHEORETIC CONSEQUENCES OF THE RESIDUE THEOREM
Denition. Let D C be open. A meromorphic function on D is a holomorphic
function f : D P(f) C such that P(f) D is discrete in D.
Remark. If D is open and connected, and f, g : D C are holomorphic, then
f
g
is
meromorphic if g is not identically zero.
Proposition 13.2. Let D C be open, and let f be meromorphic on D. Then,
for each z
0
D, there exist > 0 with B
(z
0
) D and holomorphic functions
g, h: B
(z
0
) C such that f(z) =
g(z)
h(z)
for z B
(z
0
) z
0
.
Proof. If z
0
is not a pole of f, the claim is clear.
Otherwise, choose > 0 so small that B
(z
0
) D and P(f) B
(z
0
) = z
0
. We
can then nd a holomorphic function g : B
(z
0
) C with g(z
0
) ,= 0 and k N such
that
f(z) =
g(z)
(z z
0
)
k
for z B
(z
0
) z
0
. Setting h(z) := (z z
0
)
k
yields the claim.
Denition. It is convenient to dene the set of extended complex numbers C
=
C and extend the domain of meromorphic functions to include their poles,
assigning the function value at the poles. Here is identied with for all
D. Furthermore, we dene 1/0 = and 1/ = 0.
Denition. We can reinterpret C
S
2
. Under this mapping, 0 maps to the south pole, maps to the
north pole, and the unit circle D maps to the equator. One can readily show that P
is continuous, and that the inverse P
1
: S
2
C
1
(z) S = . Also, since D is open, there exists
2
> 0 with B
2
(z) D. Setting
:= min
1
,
2
, we get B
(s) D
and B
(s) S = s. Set W := B
(s) s U
and
S
V
:= s S : there exists > 0 such that B
(s) s V .
By the foregoing, we have S = S
U
S
V
, and trivially, S
U
S
V
= holds. Set
U := U S
U
and
V := V S
V
.
Then
U ,= ,=
V are easily seen to be open and clearly satisfy
U
V = and
U
V = D, which contradicts the connectedness of D.
Theorem 13.1 (Meromorphic Functions Form a Field). Let D C be open and
connected. Then the meromorphic functions on D, where we dene (f + g)(z) =
lim
wz
[f(w) + g(w)] and (fg)(z) = lim
wz
[f(w)g(w)], form a eld.
90CHAPTER 13. FUNCTIONTHEORETIC CONSEQUENCES OF THE RESIDUE THEOREM
Proof. It is easily checked that the meromorphic functions do indeed form a commu
tative ring. For each meromorphic function f , 0 on D dene
f : D C, z
1
f(z)
.
As P(f) is discrete, DP(f) is connected by Lemma 13.1. From the Identity Theorem,
we then conclude that Z(f) is discrete, too. Thus
f is meromorphic and (f
f)(z) = 1
(the multiplicative identity) for z D.
Denition. Let z
0
Z(f). If f(z) = (zz
0
)
k
g(z), where g is a holomorphic function
with g(z
0
) ,= 0, we say that k := ord(f, z
0
).
Theorem 13.2 (Argument Principle). Let D C be open and simply connected, let
f be meromorphic on D, and let be a closed curve in D (P(f) Z(f)). Then we
have
1
2i
_
()
f()
d =
zZ(f)
(, z) ord(f, z)
zP(f)
(, z) ord(f, z).
Proof. By the Residue Theorem, we have
1
2i
_
()
f()
d =
zZ(f)
(, z) res
_
f
f
, z
_
+
zP(f)
(, z) res
_
f
f
, z
_
.
Let z
0
Z(f), and let k := ord(f, z). Then there is a holomorphic function g
with g(z
0
) ,= 0 such that f(z) = (z z
0
)
k
g(z) and thus
f
(z) = k (z z
0
)
k1
g(z) + (z z
0
)
k
g
(z).
It follows that
f
(z)
f(z)
=
k
z z
0
+
g
(z)
g(z)
for z near z
0
, so that
res
_
f
f
, z
_
= k.
Let z
0
P(f), and let k := ord(f, z
0
). Then f(z) =
g(z)
(z z
0
)
k
holds with g
holomorphic such that g(z
0
) ,= 0 and, consequently,
f
(z) = k(z z
0
)
(k+1)
g(z) + (z z
0
)
k
g
(z).
It follows that
f
(z)
f(z)
=
k
z z
0
+
g
(z)
g(z)
for z ,= z
0
near z
0
, so that
res
_
f
f
, z
_
= k.
91
Denition. Let D C be open, and let f : D C be holomorphic. We say that f
attains w
0
C with multiplicity k N at z
0
D if the function
D C, z f(z) w
0
has a zero of order k at z
0
.
Theorem 13.3 (Bifurcation Theorem). Let D C be open, let f : D C be
holomorphic, and suppose that, at z
0
D, the function f attains w
0
with multiplicity
k N. Then there exist neighbourhoods V D of z
0
and W f(V ) of w
0
such that,
for each w W w
0
, there exist distinct z
1
, . . . , z
k
V with f(z
1
) = = f(z
k
) =
w, where f attains w at each z
j
with multiplicity one.
Proof. In view of the Identity Theorem, we may choose > 0 with B
[z
0
] D such
that f(z) ,= w
0
and f
(z
0
) z
0
.
Set V := B
(z
0
) and := B
(z
0
). Choose > 0 such that B
(w
0
) C f ,
and set W := B
(w
0
). Let w W. By the Argument Principle, the number of
times w is attained in V (counting multiplicity) is
1
2i
_
()
f() w
d =
1
2i
_
f
d
w
= (f , w).
As (f , ) is constant on W, the number of times w is attained in V is the same as
the number of times w
0
is attained in V , i.e. k. Since f
n=1
converges to f compactly on D, and suppose
that Z(f
n
) = for n N. Then f 0 or Z(f) = .
Proof. In view of Theorem 6.2, we note that f itself is holomorphic. Suppose that
f , 0, but that there exists z
0
Z(f). Choose > 0 such that B
[z
0
] D and
f(z) ,= 0 for all z B
[z
0
] z
0
, and note that
0 = lim
n
1
2i
_
B(z
0
)
f
n
()
f
n
()
d =
1
2i
_
B(z
0
)
f
()
f()
d = ord(f, z
0
),
which is a contradiction.
Corollary 13.4.1. Let D C be open and connected, let f
1
, f
2
, . . . : D C be
holomorphic such that (f
n
)
n=1
converges to f compactly on D, and suppose that f
n
is injective for n N. Then f is constant or injective.
92CHAPTER 13. FUNCTIONTHEORETIC CONSEQUENCES OF THE RESIDUE THEOREM
Proof. Suppose that f is not constant. Let z
0
D be arbitrary, and dene
g
n
: D z
0
C, z f
n
(z) f
n
(z
0
)
for n N. Then g
1
, g
2
, . . . have no zeros. Since f is not constant, the function
D z
0
C, z f(z) f(z
0
)
is not zero, so that it has no zeros by Hurwitzs theorem, i.e. f(z) ,= f(z
0
) for all
z D, z ,= z
0
.
Theorem 13.5 (Rouches Theorem). Let D C be open and simply connected, and
let f, g : D C be holomorphic. Suppose that is a closed curve in D such that
int = z D : (, z) = 1 and that
[f() g()[< [f()[
for . Then f and g have the same number of zeros in int (counting multi
plicity).
Proof. For t [0, 1], dene h
t
:= f +t(g f), so that h
0
= f and h
1
= g. Also, since
[t(g f)[ [g f[< [f[
for any t [0, 1] on , the functions h
t
have no zeros on . For t [0, 1], let
n(t) N
0
denote the number of zeros of h
t
in int . Since the functions h
t
have no
poles in D, we know from the Argument Principle that
n(t) =
1
2i
_
t
()
h
t
()
d =
1
2i
_
() + t(g
() f
())
f() + t(g() f())
d.
We thus see that n(t) is a continuous function of t. But since n(t) can only taken on
integer values, it must be constant on [0, 1]; in particular, n(0) = n(1).
Example. How many zeros does z
4
4z + 2 have in D?
Set
g(z) := z
4
4z + 2 and f(z) = 4z + 2.
For D, we have [f()[ [4z[ 2 = 4 2 = 2, so that
[f() g()[= [
4
[= 1 < 2 [f()[.
Since f has precisely one zero in D, so does g.
Corollary 13.5.1 (Fundamental Theorem of Algebra). Let p be a polynomial with
n := deg p 1. Then p has n zeros (counting multiplicity).
93
Proof. Let
p(z) = a
n
z
n
+ + a
1
z + a
0
with a
n
,= 0, and let g(z) := a
n
z
n
, so that lim
z
p(z) g(z)
g(z)
= 0. Choose R > 0
such that
p(z) g(z)
g(z)
< 1
for z C with [z[ R. Consequently, if z B
R
(0), we have [p(z) g(z)[< [g(z)[.
By Rouches Theorem, p thus has as many zeros in B
R
(0) as g, namely n. Since p
has at most n zeros, these are all of the zeros of p.
Chapter 14
Harmonic Functions
Denition. Let D R
N
be open, and let u: D R be twice continuously partially
dierentiable. Then u is called harmonic if
u =
2
u
x
2
1
+ +
2
u
x
2
N
0.
We will only be concerned with harmonic functions on R
2
, i.e. on C.
Proposition 14.1 (Harmonic Components). Let D C be open, and let f : D C
be holomorphic. Then Re f and Imf are harmonic.
Proof. Clearly, Re f and Imf are twice continuously dierentiable.
We have
2
(Re f)
x
2
=
x
x
Re f
=
x
y
Imf, by CauchyRiemann,
=
y
x
Imf
=
2
(Re f)
y
2
, by CauchyRiemann again,
so that Re f 0, i.e. Re f is harmonic. Similarly, one sees that Imf is harmonic.
Remark. The converse of Proposition 14.1 is not true: a harmonic function need not
be the real part of some holomorphic function. Consider
u: C 0 R, z log[z[,
so that
u(x, y) = log
_
x
2
+ y
2
=
1
2
log(x
2
+ y
2
)
94
95
for (x, y) R
2
(0, 0). The partial derivatives of u with respect to x are
u
x
=
x
x
2
+ y
2
.
and
2
u
x
2
=
(x
2
+ y
2
) 2x
2
(x
2
+ y
2
)
2
=
x
2
+ y
2
(x
2
+ y
2
)
2
Moreover, since u is symmetric in x and y, we nd
2
u
y
2
=
y
2
+ x
2
(x
2
+ y
2
)
2
.
Consequently, u is harmonic.
Now suppose that there is a holomorphic function f : C 0 C such that
Re f = u. On C
(z) =
Re f
x
(z) + i
(Imf)
x
(z) =
Re Log
x
(z) + i
(ImLog)
x
(z) = Log
z =
1
z
for z C
(z) =
1
z
for all z C 0, so that f is
an antiderivative of z
1
z
on C 0. This is impossible (cf. page 24).
Denition. Let D C be open, and let u: D R be harmonic. We call a harmonic
function v : D R a harmonic conjugate of u if u + i v is holomorphic.
Theorem 14.1 (Harmonic Conjugates). Let D C be open and suppose that there
exists (x
0
, y
0
) D with the following property: for each (x, y) D, we have
(x, t) D for each t between y and y
0
and
(s, y
0
) D for each s between x and x
0
.
Then every harmonic function on D has a harmonic conjugate.
Proof. Let u: D R be harmonic. We will nd a harmonic v : D R such that
u
x
=
v
y
and
v
x
=
u
y
. ()
For (x, y) D, dene
v(x, y) =
_
y
y
0
u
x
(x, t) dt + (x),
96 CHAPTER 14. HARMONIC FUNCTIONS
where will be specied later. First, note that
v
x
(x, y) =
_
y
y
0
2
u
x
2
(x, t) dt +
2
u
y
2
(x, t) dt +
(x)
=
u
y
(x, y) +
u
y
(x, y
0
) +
(x).
Hence, if we want the CauchyRiemann dierential equations to hold for u + iv, we
require that
(x) =
u
y
(x, y
0
). We thus set
v(x, y) =
_
y
y
0
u
x
(x, t) dt
_
x
x
0
u
y
(s, y
0
) ds.
Then () holds, so that
2
v
x
2
=
x
v
x
=
x
u
y
=
y
u
x
=
2
v
y
2
,
i.e. v is harmonic.
Example. Let
u: R
2
R, (x, y) xy.
Then u is harmonic and
v(x, y) =
_
y
0
t dt
_
x
0
s ds =
y
2
2
x
2
2
is a harmonic conjugate for u.
Corollary 14.1.1. Let D C be open, and let u: D R be harmonic. Then, for
each z
0
D, there is a neighbourhood U D of z
0
such that u[
U
has a harmonic
conjugate.
Corollary 14.1.2. Let D C be open, and let u: D R be harmonic. Then u is
innitely often partially dierentiable.
Corollary 14.1.3. Let D C be open and connected, and let u : D R be
harmonic. Then the following are equivalent:
(i) u 0;
(ii) there exists a nonempty open set U D with u[
U
0.
97
Proof. Of course, only (ii) = (i) needs proof.
Given a nonempty open set U D with u[
U
0, let z
0
U. Corollary 14.1.1
implies that there exists a holomorphic function f = u+iv on an open ball B
(z
0
) U
of z
0
. But then
f
=
u
x
i
u
y
0
on B
(z
0
). Consequently, f = u+iv is constant on B
(z
0
) D. The Identity Theorem
then implies that f is constant throughout the open connected set D.
Corollary 14.1.4. Let D C be open, let u: D R be harmonic, and let z
0
D
and r > 0 be such that B
r
[z
0
] D. Then we have
u(z
0
) =
1
2
_
2
0
u(z
0
+ re
i
) d.
Corollary 14.1.5. Let D C be open and connected, and let u : D R be
harmonic with a local maximum or minimum on D. Then u is constant.
Proof. It is enough to consider the case of a local maximum: otherwise, replace u
by u.
Let z
0
D be a point where u attains a local maximum. Let > 0 be such that
B
(z
0
) D and u(z) u(z
0
) for all z B
(z
0
). Let v be a harmonic conjugate of
u on B
(z
0
). Hence, f := u + i v : B
(z
0
) C is holomorphic such that Re f has
a local maximum at z
0
. On considering the holomorphic function e
f
, with modulus
e
Re f
, we then see that e
Re f
has a local maximum at z
0
. The Maximum Modulus
Principle then implies that e
f
, and hence its modulus e
Re f
, must be constant on
B
(z
0
). On taking the real logarithm, we see that u = Re f also is constant on B
(z
0
).
On applying Corollary 14.1.3 to u minus this constant value, we then see that u is
constant throughout D.
Corollary 14.1.6. Let D C be open, connected, and bounded, and let u: D R
be continuous such that u[
D
is harmonic. Then u attains its maximum and minimum
over D on D.
The Dirichlet Problem. Let D C be open, connected, and bounded, and let
f : D R be continuous. Is there a continuous g : D R such that g[
D
= f and
g[
D
is harmonic?
Remark. If the Dirichlet problem has a solution, it must be unique. To see this, let
g
1
, g
2
: D R be such that g
j
[
D
= f and g
j
[
D
is harmonic for j = 1, 2. Then g
1
g
2
vanishes on D. Since g
1
g
2
attains both its maximum and minimum on D, it
follows that g
1
g
2
0 on D.
98 CHAPTER 14. HARMONIC FUNCTIONS
Denition. Let r > 0. The Poisson kernel for B
r
(0) is dened as
P
r
(, z) :=
r
2
[z[
2
2[ z[
2
for z B
r
(0) and B
r
(0).
Lemma 14.1. Let D C be open, let r > 0 be such that B
r
[0] D, and let
f : D C be holomorphic. Then we have
f(z) =
1
2
_
2
0
f(re
i
)
r
2
[z[
2
[re
i
z[
2
d =
_
2
0
f(re
i
)P
r
(re
i
, z) d.
for z B
r
(0).
Proof. Fix z B
r
(0), and dene g(w) :=
f(w)
r
2
wz
, which is holomorphic for w in
B
r+
(0) for some > 0. The Cauchy Integral Formula then yields
f(z)
r
2
[z[
2
= g(z) =
1
2i
_
Br(0)
g()
z
d =
1
2i
_
2
0
g(re
i
)ire
i
re
i
z
d
=
1
2
_
2
0
f(re
i
)re
i
(r
2
re
i
z)(re
i
z)
d =
1
2
_
2
0
f(re
i
)
(re
i
z)(re
i
z)
d =
1
2
_
2
0
f(re
i
)
[re
i
z[
2
d,
so that
f(z) =
1
2
_
2
0
f(re
i
)
r
2
[z[
2
[re
i
z[
2
d.
Remark. If we apply Lemma 14.1 to the function f = 1 we see for all z B
r
(0) that
_
2
0
P
r
(re
i
, z) d = 1.
Theorem 14.2 (Poissons Integral Formula). Let r > 0, and let u : B
r
[0] R be
continuous such that u[
Br(0)
is harmonic. Then
u(z) =
_
2
0
u(re
i
)P
r
(re
i
, z) d
holds for all z B
r
(0).
Proof. Suppose rst that u extends to B
R
(0) for some R > r as a harmonic function.
Then u has a harmonic conjugate v on B
R
(0), so that f := u + i v is holomorphic.
By Lemma 14.1, we have, for z B
r
(0), that
u(z) + i v(z) = f(z)
=
_
2
0
f(re
i
)P
r
(re
i
, z) d =
_
2
0
u(re
i
)P
r
(re
i
, z) d +i
_
2
0
v(re
i
)P
r
(re
i
, z) d,
99
so that
u(z) =
_
2
0
u(re
i
)P
r
(re
i
, z) d.
Suppose now that u is arbitrary. For t (0, 1), dene
u
t
: B
r
t
(0) R, z u(tz).
Then u
t
is harmonic, and by the foregoing we have
u
t
(z) =
_
2
0
u
t
(re
i
)P
r
(re
i
, z) d
for z B
r
(0). Letting t 1
z)
[ z[
2
=
1
[ z[
2
Re([[
2
[z[
2
+z
z) =
1 [z[
2
[ z[
2
= 2P
1
(, z).
As the real part of a holomorphic function,
D R, z P
1
(, z)
is therefore harmonic for each D. We thus obtain for z = x + iy D:
(g)(z) =
2
g
x
2
(z) +
2
g
y
2
(z) =
_
2
0
f(e
i
)
_
2
x
2
P
1
(e
i
, z) +
2
y
2
P
1
(e
i
, z)
_
d = 0.
Consequently, g is harmonic on B
r
(0).
What remains to be shown is that g is continuous at any point z
0
D.
Let z
0
= e
i
0
, and suppose without loss of generality (if necessary considering
instead g(z)) that
0
(0, 2). Let > 0. We need to nd > 0 such that
[g(z
0
) g(z)[< for all z D with [z
0
z[< .
For
0
> 0, let J := [
0
2
0
,
0
+ 2
0
]. By making
0
> 0 suciently small, we
can ensure that J [0, 2] and [f(e
i
) f(z
0
)[<
2
for J. Set
S := se
i
: s [0, 1), [
0
0
,
0
+
0
],
and note that C := inf[e
i
z[: [0, 2] J, z S > 0.
100 CHAPTER 14. HARMONIC FUNCTIONS
z
0
S
0
+ 2
0
0
0
2
0
Since
_
2
0
P
1
(e
i
, z) d = 1 for all z D, we have
g(z) g(z
0
) =
_
2
0
(f(e
i
) f(z
0
))P
1
(e
i
, z) d
=
_
J
(f(e
i
) f(z
0
))P
1
(e
i
, z) d
. .
I
1
+
_
[0,2]\J
(f(e
i
) f(z
0
))P
1
(e
i
, z) d
. .
I
2
.
Note that
[I
1
[
_
J
[f(e
i
) f(z
0
)[
. .
<
2
P
1
(e
i
, z) d
2
_
2
0
P
1
(e
i
, z) d =
2
.
101
Set K := sup
D
[f()[. For z S, we then have
[I
2
[
_
[0,2]\J
([f(e
i
)[+[f(z
0
)[)P
1
(e
i
, z) d
=
1
2
_
[0,2]\J
([f(e
i
)[+[f(z
0
)[)
1 [z[
2
[e
i
z[
2
d
_
[0,2]\J
1 [z[
2
[e
i
z[
2
d
K
C
2
_
[0,2]\J
(1 [z[
2
) d, because z S,
2K
C
2
(1 [z[
2
)
Choose (0,
0
) so small that [z
0
z[< for z D implies
1 [z[
2
<
C
2
2K
2
.
For z D with [z
0
z[< , we then have z S and hence [I
2
[<
2
. On combining
these results, we see that [g(z
0
) g(z)[< .
Denition. Let D C be open, and let f : D C be continuous. We say that f
has the mean value property if, for every z
0
D, there exists R > 0 with B
R
[z
0
] D
such that
f(z
0
) =
1
2
_
2
0
f(z
0
+ re
i
) d
for all r [0, R].
Theorem 14.4. Let D C be open, and let f : D C have the mean value
property such that [f[ attains a local maximum at z
0
D. Then f is constant on a
neighbourhood of z
0
.
Proof. Choose R > 0 with B
R
[z
0
] D such that [f(z
0
)[ [f(z)[ for all z B
R
[z
0
]
and f(z
0
) =
1
2
_
2
0
f(z
0
+ re
i
) d for all r [0, R]. If f(z
0
) = 0 then the result is
trivial. Otherwise, let
h(z) =
[f(z
0
)[
f(z
0
)
f(z)
and set g := Re h [h(z
0
)[. Then g has the mean value property and satises
g(z) [h(z)[[h(z
0
)[ 0
for z B
R
[z
0
]. It follows that
0 = g(z
0
) =
_
2
0
g(z
0
+ re
i
)
. .
0
d
102 CHAPTER 14. HARMONIC FUNCTIONS
for all r [0, R]. As the integrand is continuous, we conclude that g(z
0
+ re
i
) = 0
for all r [0, R] and [0, 2], i.e. g 0 on B
R
[z
0
]. This means that, for z B
R
[z
0
],
we have
[h(z)[ [h(z
0
)[= Re h(z) [h(z)[,
so that Re h(z) = [h(z)[= [h(z
0
)[ for z B
R
[z
0
]. That is, h(z) = [h(z)[= [h(z
0
)[=
[f(z
0
)[, so that f(z) = f(z
0
) for z B
R
[z
0
].
Corollary 14.4.1. Let D C be open, let f : D R be continuous and have
the mean value property, and suppose that f has a local maximum or minimum at
z
0
D. Then f is constant on a neighbourhood of z
0
.
Proof. We only consider the case of a local maximum (for a local minimum, replace
f by f).
Let R > 0 be such that B
R
[z
0
] D and f(z) f(z
0
) for all z B
R
[z
0
]. Choose C
such that f(z)+C 0 for all z B
R
[z
0
]. It follows that [f +C[ has a local maximum
at z
0
. Hence, f + C is constant on a neighbourhood of z
0
, as is f.
Corollary 14.4.2. Let D C be open, connected, and bounded, and let f : D
R be continuous such that f[
D
has the mean value property. Then f attains its
maximum and minimum on D.
Proof. Without loss of generality, suppose that f is not constant. Let z
0
D be such
that f(z
0
) is maximal. Set
V := z D : f(z) < f(z
0
).
Then V is open and not empty. Let z D V , i.e. f(z) = f(z
0
). Then f has a
local maximum at z, so that, by Corollary 14.4.1, f(w) = f(z) = f(z
0
) for w in a
neighbourhood, say W D, of z. Consequently, W D V holds, so that z is an
interior point of D V . Since z D V is arbitrary, this shows that D V is open.
Since D is connected, and V ,= , we must have D V = , i.e. V = D.
The case of a minimum is treated analogously.
Corollary 14.4.3 (Equivalence of Harmonic and MeanValue Properties). Let D C
be open, and let f : D R be continuous. Then the following are equivalent:
(i) f is harmonic;
(ii) f has the mean value property.
Proof. Only (ii) = (i) needs proof (cf. Corollary 14.1.4).
Let z
0
D, and let R > 0 be such that B
R
[z
0
] D. By Theorem 14.3, there is a
continuous function g : B
R
[z
0
] R such that g[
B
R
[z
0
]
= f[
B
R
[z
0
]
and g[
Br(z
0
)
is har
monic. Consequently, (gf)[
B
R
(z
0
)
has the mean value property. By Corollary 14.4.2,
this means that g f attains its maximum and minimum over B
R
[z
0
] on B
R
[z
0
],
so that g = f on B
R
[z
0
]. Hence, f[
B
R
(z
0
)
is harmonic, i.e. f 0 on B
R
[z
0
]. Since
z
0
D was arbitrary, this means that f 0.
Chapter 15
Analytic Continuation along a
Curve
Example. Let
D
1
:= z C : Re z > 0,
D
2
:= z C : Imz > Re z,
and
D
3
:= z C : Imz < Re z,
so that
D
1
D
2
D
3
= C 0.
Let
g : C 0 C, z
1
z
,
and let f
1
= Log[
D
1
, so that f
1
is an antiderivative of g on D
1
. Since D
2
is simply
connected, g also has an antiderivative on D
2
; since (f
1
f
2
)[
D
1
D
2
= (gg)[
D
1
D
2
0,
it follows that (f
1
f
2
)[
D
1
D
2
is constant, and by altering f
2
by an additive constant,
we can achieve that f
1
[
D
1
D
2
= f
2
[
D
1
D
2
. In the same fashion, we can nd an an
tiderivative f
3
of g on D
3
such that f
2
[
D
2
D
3
= f
3
[
D
2
D
3
. However, f
1
[
D
1
D
3
,= f
3
[
D
1
D
3
because otherwise, we would have an antiderivative of g on all of C 0, which we
know to be impossible.
Since f
1
f
3
[
D
1
D
3
= g g
D
1
D
3
0, however, there exists c C such that
f
3
(z) = f
1
(z)+c for z D
1
D
2
. We claim that c = 2i. To see this, let z
1
, z
2
, z
3
D
be such that z
1
D
1
D
3
, z
2
D
2
D
1
, and z
3
D
3
D
2
. Let
z
1
,z
2
,
z
2
,z
3
, and
z
3
,z
1
be the arc segments of D from z
1
to z
2
, from z
2
to z
3
, and from z
3
to z
1
, respectively.
103
104 CHAPTER 15. ANALYTIC CONTINUATION ALONG A CURVE
Since f
j
is an antiderivative of g on D
j
for j = 1, 2, 3, we obtain
_
z
1
,z
2
g = f
1
(z
2
) f
1
(z
1
),
_
z
2
,z
3
g = f
2
(z
3
) f
2
(z
2
),
and
_
z
3
,z
1
g = f
3
(z
1
) f
3
(z
3
).
It follows that
c = f
3
(z
1
) f
1
(z
1
)
= f
3
(z
1
) f
3
(z
3
) + f
2
(z
3
) f
2
(z
2
) + f
1
(z
2
) f
1
(z
1
)
=
_
z
3
,z
1
g +
_
z
2
,z
3
g +
_
z
1
,z
2
g
=
_
z
1
,z
2
z
2
,z
3
z
3
,z
1
g
=
_
D
1
d
= 2i.
y
x
z
1
z
2
z
3
D
1
D
2
D
3
Denition. A function element is a pair (D, f), where D C is open and connected,
and f : D C is a holomorphic function. For a given function element (D, f) and
105
z
0
D, the germ of f at z
0
denoted by f
z
0
is the collection of all function
elements (E, g) such that z
0
E and there is a neighbourhood U DE of z
0
such
that f(z) = g(z) for all z U.
Denition. Let : [0, 1] C be a path, and suppose that, for each t [0, 1], there
is a function element (D
t
, f
t
) such that:
(a) (t) D
t
;
(b) there exists > 0 such that, whenever s [0, 1] is such that [s t[< , then
(s) D
t
and f
s
(s)
= f
t
(s)
.
Then we call (D
t
, f
t
) : t [0, 1] an analytic continuation along and say that
(D
1
, f
1
) is obtained by analytic continuation of (D
0
, f
0
) along .
Remark. Since is continuous and D
t
is open for each t D
t
, it is clear that there
exists > 0 such that (s) D
t
for all s [0, 1] such that [s t[< . What is
important about part (b) of the denition is that f
s
(s)
= f
t
(s)
, i.e. there is a
neighbourhood U
s
D
s
D
t
of (s) such that f
s
(z) = f
t
(z) for z U
s
.
(0)
(1)
D
0
D
1
Theorem 15.1 (Monodromy Theorem). Let : [0, 1] C be a path, and let
(D
t
, f
t
) : t [0, 1] and (E
t
, g
t
) : t [0, 1] be analytic continuations along
such that f
0
(0)
= g
0
(0)
. Then we have f
1
(1)
= g
1
(1)
.
106 CHAPTER 15. ANALYTIC CONTINUATION ALONG A CURVE
Proof. Let
I = t [0, 1] : f
t
(t)
= g
t
(t)
,
so that 0 I.
We rst claim that I is closed. Let t I, and let > 0 be such that (s) D
t
E
t
and
f
s
(s)
= f
t
(s)
and g
s
(s)
= g
t
(s)
for all s [0, 1] with [s t[< . Since t I, there exists s I with [s t[< . There
is thus a neighbourhood U D
t
D
s
E
t
E
s
of (s) such that f
s
(z) = g
s
(z) for
all z U by the denition of I. From the choice of , we also haveafter possibly
making U smallerthat f
s
(z) = f
t
(z) and g
s
(z) = g
t
(z) for z U. It follows that
f
t
(z) = g
t
(z) for z U, so that t I.
Let t
0
:= sup I. Let > 0 be such that (s) D
t
0
E
t
0
and
f
s
(s)
= f
t
0
(s)
and g
s
(s)
= g
t
0
(s)
for all s [0, 1] with [s t[< . Since I is closed, we have t
0
I and thus f
t
0
(z) =
g
t
0
(z) for all z in some neighbourhood V of (t
0
) contained in D
t
0
E
t
0
. It follows
that f
t
0
(s)
= g
t
0
(s)
for all s [0, 1] such that (s) V . For > 0 suciently
small, we thus have f
s
(s)
= g
s
(s)
for any s [0, 1] with [s t[< . It follows
that [0, 1] (t
0
, t
0
+ ) I. Since t
0
= sup I, this means that t
0
= 1, so that
I = [0, 1].
Chapter 16
Montels Theorem
Denition. Let S R
N
. A family T of functions on S into R
M
is called equicontinuous
if, for each > 0, there exists > 0 such that [f(x) f(y)[< for all f T and for
all x, y S such that [x y[< .
Lemma 16.1. Let S R
N
. Then S contains a countable dense subset.
Proof. Let x
1
, x
2
, x
3
, . . . be a dense, countable subset of R
N
, e.g., Q
N
. For n, m N
with S B 1
m
(x
n
) ,= , choose y
n,m
S B 1
m
(x
n
). Then
_
y
n,m
: n, m N, S B 1
m
(x
n
) ,=
_
S
is countable.
Let > 0 and x S. Choose m N so large that
1
m
<
2
. Since x
1
, x
2
, x
3
, . . . is
dense in R
N
, there exists n N such that [x
n
x[<
1
m
and thus x S B 1
m
(x
n
) ,= .
It follows that
[y
n,m
x[ [y
n,m
x
n
[+[x
n
x[<
2
m
<
Theorem 16.1 (Arzel` aAscoli Theorem). Let K R
N
be compact, and let T be
an equicontinuous and uniformly bounded family of functions from K to R
M
. Then
every sequence in T has a subsequence that converges uniformly on K.
Proof. Let (f
n
)
n=1
be a sequence in T, and let x
1
, x
2
, x
3
, . . . be a countable dense
subset of K.
Since (f
n
(x
1
))
n=1
is a bounded sequence in R
M
, there exists a subsequence (f
n,1
)
n=1
of (f
n
)
n=1
such that (f
n,1
(x
1
))
n=1
converges.
Since (f
n,1
(x
2
))
n=1
is a bounded sequence in R
M
, there exists a subsequence
(f
n,2
)
n=1
of (f
n,1
)
n=1
such that (f
n,2
(x
2
))
n=1
converges.
Continuing inductively in this fashion, we obtain, for each k N, a subsequence
(f
n,k
)
n=1
of (f
n
)
n=1
such that, for each k N,
107
108 CHAPTER 16. MONTELS THEOREM
(f
n,k+1
)
n=1
is a subsequence of (f
n,k
)
n=1
, and
(f
n,k
(x
k
))
n=1
converges.
For n N, set g
n
:= f
n,n
. Then (g
n
)
n=1
is a subsequence of (f
n
)
n=1
, and
(g
n
(x
k
))
n=1
converges for each k N.
We show that (g
n
)
n=1
is a uniform Cauchy sequence on K (and thus convergent).
Let > 0. Choose > 0 such that [f(x) f(y)[<
3
for all f T and for all
x, y K with [x y[< . Since K is compact, there exist y
1
, . . . , y
K such that
K
j=1
B
2
(y
j
). Since x
1
, x
2
, x
3
, . . . is dense in K, there exist k
1
, . . . , k
N such
that x
k
j
B
2
(y
j
). It follows that K
j=1
B
(x
k
j
).
By construction, (g
n
(x
k
))
n=1
is a Cauchy sequence for each k N. Choose N N
such that
[g
n
(x
k
j
) g
m
(x
k
j
)[<
3
for n, m N and j = 1, . . . , . Let x K be arbitrary, and let n, m N. Choose
j 1, . . . , such that x B
(x
k
j
), and note that
[g
n
(x) g
m
(x)[ [g
n
(x) g
n
(x
k
j
)[
. .
<
3
+[g
n
(x
k
j
) g
m
(x
k
j
)[
. .
<
3
+[g
m
(x
k
j
) g
m
(x)[
. .
<
3
< .
Hence, (g
n
)
n=1
is a uniform Cauchy sequence on K.
Proposition 16.1. Let D R
N
be open, and let T be a family of functions from D
to R
M
that is equicontinuous and uniformly bounded on compact subsets of D. Then
every sequence in T has a compactly convergent subsequence.
Proof. For each k N, dene K
k
:= B
k
[0] if D = R
N
and K
k
:= B
k
[0] x D :
dist(x, D)
1
k
if D ,= R
N
. Notice that
k=1
K
k
= D and
K
k
int K
k+1
for n N.
Let (f
n
)
n=1
be a sequence in T. By the Arzel` aAscoli Theorem, there exists a
subsequence (f
n,1
)
n=1
of (f
n
)
n=1
and a function g
1
: K
1
R
M
such that f
n,1
[
K
1
g
1
uniformly on K
1
. Invoking the Arzel` aAscoli Theorem again, we obtain a subsequence
(f
n,2
)
n=1
of (f
n,1
)
n=1
and a function g
2
: K
2
R
M
such that f
n,2
[
K
2
g
2
uniformly
on K
2
. Inductively, we thus obtain, for each k N, a subsequence (f
n,k
)
n=1
of (f
n
)
n=1
and a function g
k
: K
k
R
M
such that, for each k N,
(f
n,k+1
)
n=1
is a subsequence of (f
n,k
)
n=1
, and
f
n,k
[
K
k
g
k
uniformly on K
k
.
109
Dene g : D R
M
as follows: for x D, let k be the smallest natural number
such that x K
k
, set g(x) := g
k
(x). Then f
n,n
[
K
k
g[
K
k
uniformly on K
k
.
Let K D be compact. By the choices of K
1
, K
2
, . . ., we have K D =
k=1
int K
k
, so that int K
k
: k N is an open cover for K. Since K is compact,
and since int K
k
int K
k+1
for k N, there exists k
0
N such that K int K
k
0
K
k
0
. Since f
n,n
[
K
k
0
g[
K
k
0
uniformly on K
k
0
, it follows that f
n,n
[
K
g[
K
uniformly
on K.
Theorem 16.2 (Montels Theorem). Let D C be open, and let T be a family of
holomorphic functions on D that is uniformly bounded on compact subsets of D. Then
every sequence in T has a subsequence that converges compactly to a holomorphic
function on D.
Proof. In view of Proposition 16.1, we only need to show that T is equicontinuous
on compact subsets of D.
Let z
0
D, and let r > 0 be such that B
2r
[z
0
] D. There exists C > 0 such that
[f()[ C for all f T and all B
2r
(z
0
).
Let f T, and let z, w B
r
(z
0
). Then we have:
[f(z) f(w)[ =
1
2
_
B
2r
(z
0
)
_
f()
z
f()
w
_
d
=
1
2
_
B
2r
(z
0
)
f()( w + z )
( z)( w)
d
=
[z w[
2
_
B
2r
(z
0
)
f()
( z)( w)
d
[z w[
2
4r
C
r
2
=
2C
r
[z w[.
For > 0, choose :=
r
2C
, so that [f(z) f(w)[< for all z, w B
r
(z
0
) with
[z w[< .
Chapter 17
The Riemann Mapping Theorem
Denition. Let D
1
, D
2
C be open and connected. We say that D
1
and D
2
are
biholomorphically equivalent if there is a biholomorphic map from D
1
onto D
2
.
Examples.
1. Let z
1
, z
2
C, and let r
1
, r
2
> 0. Then B
r
1
(z
1
) and B
r
2
(z
2
) are biholomorphi
cally equivalent because
B
r
1
(z
1
) B
r
2
(z
2
), z
r
2
r
1
(z z
1
) + z
2
is biholomorphic.
2. Consider the Cayley transform
f : H C, z
z i
z + i
.
Let x, y R with y > 0, and let z = x + iy. Then
[z i[
2
= [x + i(y 1)[
2
= x
2
+ y
2
2y + 1
< x
2
+ y
2
+ 2y + 1
= [x + i(y + 1)[
2
= [z + i[
2
holds, so that [f(z)[< 1. Consequently, we have f(H) D. Consider
g : D C, z i
1 + z
1 z
,
110
111
and note that
g(f(z)) = i
1 +
zi
z+i
1
zi
z+i
= i
z + i + z i
z + i z + i
= i
2z
2i
= z
for z H. Hence, f is injective. Let x
2
+ y
2
< 1, and note that
g(x + iy) = i
(1 + x) + iy
(1 x) iy
= i
[(1 + x) + iy][(1 x) + iy]
(1 x)
2
+ y
2
=
2y
(1 x)
2
+ y
2
+ i
1 (x
2
+ y
2
)
(1 x)
2
+ y
2
. .
>0
H.
For z D, we can thus evaluate
f(g(z)) =
i
1+z
1z
i
i
1+z
1z
+ i
=
1 + z 1 + z
1 + z + 1 z
=
2z
2
= z.
Hence, f is also surjective and thus bijective with inverse g. Since f and g
are obviously holomorphic, this means that D and H are biholomorphically
equivalent.
3. There is no biholomorphic map f : C D because any holomorphic map fromC
to D is bounded and thus constant by Liouvilles theorem. Hence, C and D are
not biholomorphically equivalent.
112 CHAPTER 17. THE RIEMANN MAPPING THEOREM
Proposition 17.1. Let D
1
, D
2
C be open and connected such that D
1
is simply
connected, and suppose that D
1
and D
2
are biholomorphically equivalent. Then D
2
is
simply connected.
Proof. Let f : D
1
D
2
be biholomorphic. Let g : D
2
C be holomorphic, and let
be a closed curve in D
2
. Since f
1
is a closed curve in the simply connected set
D
1
, we see that
_
g() d =
_
ff
1
g() d =
_
f
1
g(f())f
() d = 0.
Example. Let r, R [0, ] be such that r < R. Then A
r,R
(0) is not biholomorphically
equivalent to D or C.
Biholomorphic maps have a very interesting geometric property.
Given and open set D R
N
and curves
1
,
2
: [0, 1] D, suppose there exist
t
1
, t
2
(0, 1) such that
1
(t
1
) =
2
(t
2
) = x
0
. In order to dene the angle between
1
and
2
at x
0
, we further suppose that there exists > 0 such that
j
is dierentiable
on (t
j
, t
j
+ ) for j = 1, 2 with
j
(t
j
) ,= 0. The angle is then dened to be the
unique [0, ] such that
cos =
1
(t
1
)
2
(t
2
)
[
1
(t
1
)[[
2
(t
2
)[
.
Given two open sets D
1
, D
2
R
N
, a dierentiable map f : D
1
D
2
is called angle
preserving at x
0
D
1
if, for any two curves
1
and
2
in D
1
, the angle between f
1
and f
2
at f(x
0
) is the same between
1
and
2
at x
0
.
Recall that a real N N matrix A is called orthogonal if it is invertible with
A
1
= A
t
.
Lemma 17.1. Let D
1
, D
2
R
N
be open, let x
0
D
1
, and let f : D
1
D
2
be
dierentiable such that J
f
(x
0
) is orthogonal. Then f is angle preserving at x
0
.
Proof. Let
1
and
2
be two curves in D
1
satisfying the necessary requirements, and
note that
cosine of the angle between f
1
and f
2
at f(x
0
)
=
(f
1
)
(t
1
) (f
2
)
(t
2
)
[(f
1
)
(t
1
)[[(f
2
)
(t
2
)[
=
J
f
(x
0
)
1
(t
1
) J
f
(x
0
)
2
(t
2
)
[J
f
(x
0
)
1
(t
1
)[[J
f
(x
0
)
2
(t
2
)[
, by the chain rule,
=
J
f
(x
0
)
t
J
f
(x
0
)
1
(t
1
)
2
(t
2
)
[J
f
(x
0
)
1
(t
1
)[[J
f
(x
0
)
2
(t
2
)[
=
1
(t
1
)
2
(t
2
)
[
1
(t
1
)[[
2
(t
2
)[
= cosine of the angle between
1
and
2
at x
0
.
113
Example. Let z be a complex number. Then multiplication by z is a Rlinear map
from C = R
2
into itself and thus uniquely represented by a real 2 2 matrix A of the
form
_
a b
b a
_
,
where a = Re z and b = Imz. It follows that A
t
is the matrix representing z. Hence,
A is orthogonal if and only if [z[= 1.
Theorem 17.1 (Conformality at Nondegenerate Points). Let D
1
, D
2
C be open,
and let f : D
1
D
2
be holomorphic. Then f is angle preserving at z
0
D
1
whenever
f
(z
0
) ,= 0.
Proof. Let z
0
D
1
be such that f
(z
0
) ,= 0. In view of Lemma 17.1 and the example
following it, the claim is clear if [f
(z
0
)[= 1.
For the general case, let
1
[f
(z
0
)[
D
2
:=
_
z
[f
(z
0
)[
: z D
2
_
,
and dene
g : D
1
1
[f
(z
0
)[
D
2
, z
f(z)
[f
(z
0
)[
and
h:
1
[f
(z
0
)[
D
2
D
2
, z [f
(z
0
)[z.
Then g is angle preserving at z
0
because [g
(z
0
)[= 1, and it is easily seen that h is
angle preserving at g(z
0
). Consequently, f = h g is angle preserving at z
0
.
Corollary 17.1.1 (Conformality of Biholomorphic Maps). Let D
1
, D
2
C be open
and connected, and let f : D
1
D
2
be biholomorphic. Then f is angle preserving at
every point of D
1
.
Theorem 17.2 (Holomorphic Inverses). Let D
1
, D
2
C be open and connected,
and let f : D
1
D
2
be holomorphic and bijective. Then f is biholomorphic and
Z(f
) = .
Proof. We rst show that f
1
is continuous.
Let w
0
D
2
, and let > 0 be such that B
(f
1
(w
0
)) D
1
. By the Open
Mapping Theorem, f(B
(f
1
(w
0
))) is open. Hence, there exists > 0 such that
B
(w
0
) f(B
(f
1
(w
0
))). Hence, if w B
(w
0
), then f
1
(w) B
(f
1
(w
0
)), That
is, f
1
is continuous at w
0
.
114 CHAPTER 17. THE RIEMANN MAPPING THEOREM
Let z
0
D
1
and let w
0
:= f(z
0
). We see that f
is dierentiable at z
0
only if
f
(z
0
) ,= 0: for w D
2
w
0
,
f
1
(w) f
1
(w
0
)
w w
0
=
f
1
(w) f
1
(w
0
)
f(f
1
(w)) f(f
1
(w
0
))
ww
0
1
f
(z
0
)
.
Since f is not constant, f
) is discrete. We
claim that f(Z(f
n=1
in Z(f
) such that w
0
,= f(z
n
) for n N, but
w
0
= lim
n
f(z
n
). By the bijectivity and continuity of f
1
, we have f
1
(w
0
) ,= z
n
for
n N and f
1
(w
0
) = lim
n
z
n
. Hence, f
1
(w
0
) is a cluster point of Z(f
), which is
impossible.
Hence, f
1
is holomorphic on D
2
f(Z(f
)). Since f
1
is continuous and f(Z(f
))
is discrete, Riemanns Removability Criterion then yields the holomorphy of f
1
on
all of D
2
. Thus Z(f
) = .
Corollary 17.2.1. Let D C be open and connected, and let f : D C be
holomorphic and injective. Then Z(f
) = .
Proof. If f is injective, it is not constant. By the Open Mapping Theorem, f(D) is
therefore open and connected. Apply Theorem 17.2 with D
1
= D and D
2
= f(D).
Theorem 17.3 (Riemann Mapping Theorem). Let D C be open and connected and
admit holomorphic square roots, and let z
0
D. Then there is a unique biholomorphic
function f : D D with f(z
0
) = 0 and f
(z
0
) > 0.
Proof. Uniqueness: Let g : D D be another such function. Then f g
1
: D D is
biholomorphic with (f g
1
)(0) = f(z
0
) = 0. By Corollary 7.4.1, there exists c C
with [c[= 1 such that
f(g
1
(z)) = cz
for z D and thus
f(z) = f(g
1
(g(z))) = c g(z)
for z D. Dierentiation yields f
(z) = c g
(z
0
) and g
(z
0
) are real and positive, we conclude that c = 1.
Existence: Let
T := f : D D : f is injective and holomorphic with f(z
0
) = 0 and f
(z
0
) > 0.
Claim 1. T ,= .
Since D ,= C, there exists w C D. Since D admits holomorphic square roots,
there is a holomorphic function g : D C such that [g(z)]
2
= z w for z D. Note
for z
1
, z
2
D that g(z
1
) = g(z
2
) z
1
= z
2
. In particular, this means that g is
injective and thus not constant.
115
By the Open Mapping Theorem, there exists r > 0 with B
r
(g(z
0
)) g(D).
Assume that there exists a point z D with g(z) B
r
(g(z
0
)):
r > [g(z) + g(z
0
)[= [g(z) g(z
0
)[;
this means that g(z) B
r
(g(z
0
)) g(D). Hence, there exists z D with g( z) =
g(z) and thus z = z, which in turn yields that 0 = [g(z)]
2
= z w. This contradicts
w / D. Hence, g(D) B
r
(g(z
0
)) = must hold.
Dene
g : D D, z
r
g(z) + g(z
0
)
,
Then g : D D is holomorphic and injective. Let a := g(z
0
). Then
a
g :
D D is holomorphic and injective with (
a
g)(z
0
) = 0 and (
a
g)
(z
0
) ,= 0
by Corollary 17.2.1. Let c C with [c[= 1 be such that c (
a
g)
(z
0
) > 0. Then
c (
a
g) T, so that indeed T ,= .
Claim 2. Let (f
n
)
n=1
be a sequence in T converging compactly to f : D C. Then
either f 0 or f T.
It is straightforward that f(z
0
) = 0, f
(z
0
) 0, and f(D) D. By Corol
lary 13.4.1, f 0 or f is injective. If f is injective, then f
(z
0
) ,= 0 must hold by
Corollary 17.2.1, so that f
(z
0
) > 0. Also, since f(D) D is open, we have f(D) D,
and hence f T.
Claim 3. There exists f T such that f(D) = D.
Choose a sequence (f
n
)
n=1
in T such that
lim
n
f
n
(z
0
) = sup
(z
0
) :
f T (0, ].
Notice that [f
n
(z)[ 1 for all z D. By Montels Theorem, there exists a subsequence
(f
n
k
)
k=1
that converges compactly to some f : D D. In particular,
f
(z
0
) = sup
(z
0
) :
f T > 0 ()
holds, so that f T by Claim 2.
Assume that there exists w Df(D). Since D admits holomorphic square roots,
there is a holomorphic function h: D C such that
[h(z)]
2
= (
w
f)(z) =
f(z) w
1 wf(z)
.
for z D. In particular, h(D) D, h is injective and hence h
(z
0
) ,= 0. We then
evaluate the derivative of each side of the above equation at z = z
0
, noting that
f(z
0
) = 0:
2h(z
0
)h
(z
0
) = f
(z
0
) + wf
(z
0
)(w) = (1 [w[
2
)f
(z
0
).
116 CHAPTER 17. THE RIEMANN MAPPING THEOREM
We also note that [h(z
0
)[
2
= [w[ and dene
g : D C, z
[h
(z
0
)[
h
(z
0
)
(
h(z
0
)
h)(z).
Then g is injective with g(D) D and g(z
0
) = 0. Since
a
(a) = 1/(1 [a[
2
),
g
(z
0
) =
[h
(z
0
)[
h
(z
0
)
h(z
0
)
(h(z
0
))h
(z
0
)
=
[h
(z
0
)[
1 [h(z
0
)[
2
=
(1 [w[
2
)f
(z
0
)
2
_
[w[(1 [w[)
=
1 +[w[
2
_
[w[
f
(z
0
) > f
(z
0
) > 0,
so that g T and g
(z
0
) > f
(z
0
). This contradicts ().
Theorem 17.4 (Simply Connected Domains). The following are equivalent for an
open and connected set D C:
(i) D is simply connected;
(ii) D admits holomorphic logarithms;
(iii) D admits holomorphic roots;
(iv) D admits holomorphic square roots;
(v) D is all of C or biholomorphically equivalent to D;
(vi) every holomorphic function f : D C has an antiderivative;
(vii)
_
f()
z
d
for each closed curve in D and all z D ;
(ix) every harmonic function u: D R has a harmonic conjugate.
117
Proof. (i) = (ii) is Corollary 11.2.2, (ii) = (iii) is shown in the proof of Corol
lary 11.2.3, (iii) = (iv) is trivial, (iv) = (v) follows from Theorem 17.3, and (v)
= (i) is implied by Proposition 17.1.
(i) (vi) is Corollary 11.2.1 and (vi) (vii) follows from Theorem 4.1.
(i) = (viii) follows from Theorem 11.1, and (viii) = (vii) is established in the
proof of Theorem 11.2.
(v) = (ix): Let u: D R be harmonic. If D = C, the existence of a harmonic
conjugate is immediate by Theorem 14.1. So suppose that D ,= C. Hence, there is a
biholomorphic map f : D D. It is easily seen that u := uf
1
: D R is harmonic
and by Theorem 14.1 has a harmonic conjugate v : D R. Then v := v f : D R
is a harmonic conjugate of u.
(ix) =(ii): Let f : D C be holomorphic such that Z(f) = . Then u := log[f[
is harmonic and thus has a harmonic conjugate v : D R so that g := u + i v is
holomorphic. On D we have
[exp g[= [exp(u + i v)[= exp u = [f[,
so that
D C, z
f(z)
exp(g(z))
is a holomorphic function whose range lies on D and therefore isnt open. By the
Open Mapping Theorem, this means that there exists c D such that f(z) =
c exp(g(z)) for z D. Choose R with exp(i) = c, and note that f(z) =
exp(g(z) + i) for z D.
Denition. Two (not necessarily piecewise smooth) closed curves
1
,
2
: [0, 1] D
with
1
(0) =
2
(0) and
1
(1) =
2
(1) are called path homotopic if there is a continuous
function : [0, 1] [0, 1] D such that,
(0, t) =
1
(t) and (1, t) =
2
(t)
for t [0, 1] and
(s, 0) =
1
(0) and (s, 1) =
1
(1)
for all s [0, 1].
Denition. A closed curve is called homotopic to zero if and the constant curve
(0) are path homotopic.
Further Characterizations of Simply Connected Domains. There are further
conditions that characterize simply connected domains. We will only state them,
without giving proofs. Simple connectedness is also equivalent to:
(x) every (not necessarily smooth) curve in D is homotopic to zero;
(xi) D is homeomorphic to D.
118 CHAPTER 17. THE RIEMANN MAPPING THEOREM
Condition (xi) makes no reference to holomorphic functions and is entirely topo
logical in nature. It means that there is a bijective, continuous map f : D D with
a continuous inverse. Since (x) is preserved under homeomorphisms, we see that (xi)
implies (x). For the converse, it is sucient to show that C is homeomorphic to D
(for D ,= C, this is clear by Theorem 17.3). Since
C D, z
z
1 +[z[
and
D C, z
z
1 [z[
are continuous and inverse to each other, this is indeed the case.
The converse to Problem 11.1 states that the property
(xii) for every holomorphic function f : D C, there exists a sequence of polynomials
converging to f compactly on D
always holds for a simply connected domain. The proof relies on Runges Approxi
mation Theorem.
There is also an equivalent condition for simple connectedness involving the ex
tended complex plane C
:
(xiii) C
D is connected.
Index
C is a Field, 5
Biholomorphisms of D, 51
analytic, 28
analytic continuation, 105
angle preserving, 112
annulus, 57
Antiderivative Theorem, 27
Argument Principle, 90
Arzel` aAscoli Theorem, 107
Bifurcation Theorem, 91
biholomorphic, 50
biholomorphically equivalent, 110
branch cut, 82
CasoratiWeierstra Theorem, 55
Cauchys Integral Formula, 68
Cauchys Integral Formula for Circles, 35
Cauchys Integral Theorem, 70
Cauchys Integral Theorem for Annuli, 57
Cauchys Integral Theorem for StarShaped
Domains, 31
CauchyHadamard formula, 16
CauchyRiemann Equations, 11
Cayley transform, 110
center, 30
closed, 24
Compact Convergence, 42
compact convergence, 43
complex conjugate, 6
complex dierentiable, 9
complex numbers, 5
concatenation, 25
conformal, 50
Conformality at Nondegenerate Points, 113
Conformality of Biholomorphic Maps, 113
connected, 24
contour integral, 23
cosine, 15
curve, 22, 27
deMoivres Theorem, 8
discrete, 87
endpoint, 22
endpoints, 22
entire, 40
equicontinuous, 107
Equivalence of Harmonic and MeanValue
Properties, 102
essential, 55
exponential function, 15
extended complex numbers, 88
exterior, 66
Fractional Residue Theorem, 80
function element, 104
Fundamental Theorem of Algebra, 40, 92
Gamma Function, 81
Generalized Cauchy Integral Formula, 38
generalized Cauchy integral formula, 38
germ, 105
Goursats Lemma, 28
harmonic, 94
Harmonic Components, 94
harmonic conjugate, 95
Harmonic Conjugates, 95
Higher Derivatives, 38
Higher Derivatives of Power Series, 21
holomorphic, 28
119
120 INDEX
holomorphic nth roots, 71
Holomorphic Inverses, 113
Holomorphic Logarithms, 71
holomorphic logarithms, 71
Holomorphic Roots, 71
holomorphic roots, 71
homologous to zero, 68
homotopic to zero, 117
Hurwitzs Theorem, 91
Identity Theorem, 46
imaginary part, 6
imaginary unit, 6
initial point, 22
integrable, 22
Integration of Power Series, 21
interior, 66
isolated singularity, 52
Jordans Lemma:, 80
Laurent Coecients, 61
Laurent Decomposition, 59
Laurent decomposition, 61
Laurent series, 61
length, 23
line integral, 23
Liouvilles Theorem, 40
Local Uniform Convergence, 42
locally constant, 24
Locally Constant vs. Connectivity, 25
locally uniformly, 42
Maximum Modulus Principle, 48
Maximum Modulus Principle for Bounded
Domains, 49
Mean Value Equation, 36
mean value property, 101
meromorphic function, 88
Meromorphic Functions Form a Field, 89
Monodromy Theorem, 105
Montels Theorem, 109
Morera condition, 39
Open Mapping Theorem, 48
order, 54
ordered, 6
orthogonal, 112
path, 22
path homotopic, 117
piecewise smooth, 22
point of expansion, 14
Poisson kernel, 98
Poissons Integral Formula, 98
pole, 53
power series, 14
Power Series for Holomorphic Functions,
45
principal argument, 31
principal branch of the logarithm, 31
principal part, 61
Radius of Convergence, 16
radius of convergence, 16
Rational Functions, 78
Rational Trigonometric Polynomials, 75
real part, 6
region, 26
removable, 52
residue, 72
Residue Theorem, 73
reversed curve, 25
Riemann Mapping Theorem, 114
Riemann sphere, 88
Riemanns Removability Condition, 52
Rouches Theorem, 92
Runges Approximation Theorem, 118
Schwarzs Lemma, 49
secondary part, 61
simple pole, 54
simply connected, 68
Simply Connected Domains, 116
sine, 15
sliced plane, 31
star shaped, 30
INDEX 121
straight line segment, 25
TermbyTerm Dierentiation, 19
terminal point, 22
trajectory, 22
Uniform Convergence Preserves Continu
ity, 42
uniformly, 42
Weierstra Theorem, 44
winding number, 65
Winding Numbers Are Locally Constant,
67
zero, 56