Contents

1 The Field of Complex Numbers 3

2 Complex Functions 5

3 Power Series and Some Elementary Functions 11

4 Riemann-Stieltjes integrals 19

5 Line Integrals 22

6 Cauchy’s Theorem in a Ball 28

7 Cauchy’s Theorem, General Case 40

8 Harmonic Functions 46

9 Zeros and Isolated Singularities 49

10 The Maximum Modulus Principle 61

11 Essential Singularities 63

12 Sequences of Holomorphic Functions 70

13 Picard’s Big Theorem 74

14 Entire Functions 75
14.1 In…nite Products . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
14.2 Entire Functions of Finite Order . . . . . . . . . . . . . . . . . . 79
14.3 Weierstrass Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 84

15 Prime Number Theorem 87

16 Conformal Mappings 98

17 Runge’s Theorem 104

17.1 Mittag-Le- er Theorem . . . . . . . . . . . . . . . . . . . . . . . 111

18 Simply Connected Domains 113

19 Proof of Caratheodory’s Theorem 117

20 Elliptic Functions 124

21 Proof of Hadamard Theorem 135

22 Runge 137

1
23 Picard 137

24 Caratheodory 137

25 Old stu¤ 137

26 Prime 138

2
Monday, January 13, 2020

1 The Field of Complex Numbers

We de…ne C, the complex numbers, to be the set of all ordered pairs z = (x; y)
of real numbers x; y with operations of addition and multiplication de…ned by

(x1 ; y1 ) + (x2 ; y2 ) := (x1 + x2 ; y1 + y2 ); (1)

(x1 ; y1 )(x2 ; y2 ) := (x1 x2 y1 y2 ; y1 x2 + x1 y2 ); (2)

for all z1 = (x1 ; y1 ), z2 = (x2 ; y2 ). It can be checked that with these two oper-
ations C is a …eld. This means that addition and multiplication are associative
and commutative, (0; 0) and (1; 0) are the identities for addition and multiplica-
tion, respectively, every complex number has an additive inverse, every complex
number di¤erent from zero has a multiplicative inverse, and distributivity of
multiplication over addition holds. The set of complex numbers of the form
(x; 0), x 2 R is a sub…eld of C, and it is the isomorphic image of R through the
mapping
x 7! (x; 0):
Hence, from now on we will consider R as a subset of C by identifying the pair
(x; 0) with the real number x. Using this identi…cation, if we de…ne i := (0; 1)
then x + iy = (x; y). From now on we will use notation. The real numbers x
and y are called the real and imaginary parts of z, and we write

Re z = x Im z = y:

Complex numbers of the form yi are called purely imaginary numbers.

Observe that using (2) we have that i2 = 1 and so the equation z 2 + 1 = 0
has a root in C. Indeed, z 2 + 1 = (z + i)(z i). More generally, if z; w 2 C we
have that
z 2 + w2 = (z + iw)(z iw):
Using the previous formula, given z = x + iy 6= 0 we have
1 1 1 x iy x iy
= = = 2 ;
z x + iy x + iy x iy x + y2
which is the formula for the multiplicative inverse, or the opposite, of z.
Given a complex number z = x + iy, x; y 2 R, we de…ne the absolute value
of or modulus of z as p
jzj = x2 + y 2 :
Note that this is the norm of the vector (x; y) 2 R2 . Hence, we have

jzj = 0 if and only if z = 0;

jz + wj jzj + jwj for all z; w 2 C;
jtzj = jtjjzj for all z 2 C and t 2 R:

3
We leave as an exercise to show that

jzwj = jzjjwj for all z; w 2 C;

z jzj
= for all z; w 2 C; with w 6= 0:
w jwj

Since the absolute value of z = x + iy is the norm in R2 of (x; y), if we de…ne

the open ball centered at z0 = x0 + iy0 2 C and radius r > 0 as

B(z0 ; r) := fz 2 C : jz z0 j < rg;

this is nothing else than the ball B((x0 ; y0 ); r) R2 . Hence, the topology in
2
C coincides with the topology in R . So we will have the same open sets, the
same closed sets, the same compact sets, the same connected sets, and so on.
Given a complex number z = x + iy 2 C, the complex conjugate of z is
de…ned as the complex number

z := x = iy:

The following properties are left as an exercise:

z+z z z
jzj2 = zz; Re z = ; Im z = for all z 2 C; (3)
2 2i
z + w = z + w; zw = zw for all z; w 2 C; (4)
z z
= for all z; w 2 C; with w 6= 0: (5)
w w
A complex number z = x + iy 2 C n f0g can be written in polar form as

z = rei ;

where r = jzj and (we will justify this later)

ei = cos + i sin ; (6)

where is the angle between the positive real axis and the half-line starting at
the origin and passing through z. The number is called the argument of z and
is denoted arg z.
The following properties are left as an exercise:

if z = rei and w = sei' , then zw = rsei( +')

;
i n n in
if z = re and n 2 N, then z = r e :

Exercise 1 Given n 2 N, solve the equation z n = 1.

4
2 Complex Functions
De…nition 2 Let E C, let z0 2 C be an accumulation point of E and let
f : E ! C. We say that ` 2 C is the limit of f as z approaches z0 and we write

lim f (z) = `
z!z0

if for every " > 0 there exists = (z0 ; ") > 0 such that

jf (z) `j < "

for all z 2 E with 0 < jz z0 j < .

Given E C and a function f : E ! C, since the absolute value in C is the

norm in R2 , the the basic properties of limits (sum, composition, multiplication
by a scalar) will not change. The only additional property is the product of
limits.

Exercise 3 Let E C, let z0 2 C be an accumulation point of E and let

f : E ! C and g : E ! C. Assume that there exist

lim f (z) = ` 2 C; lim g(z) = L 2 C:

z!z0 z!z0

Prove that

(i) there exist

lim (f + g)(z) = ` + L;
z!z0

(ii) there exist

lim (f g)(z) = `L;
z!z0

(iii) if L 6= 0, then z0 is an accumulation point for E0 := fz 2 E : g(z) 6= 0g,

and if we restrict f =g to E0 , then there exists

f `
lim (z) = :
z!z0 g L

Exercise 4 State and prove a similar result for the limit of compositions.

Next we discuss di¤erentiation.

De…nition 5 Let E C, let z0 2 E be an accumulation point of E and let

f : E ! C. We say that f is di¤erentiable at z0 if there exists the limit
f (z) f (z0 )
lim = ` 2 C:
z!z0 z z0
df
We call the limit ` the derivative of f at z0 and we denote it by f 0 (z0 ) or dz (z0 ).

5
De…nition 6 Let U C be an open set and let f : U ! C. We say that f is
holomorphic in U , if f is di¤ erentiable in U .

The following properties are left as an exercise;

Exercise 7 Let E C, let z0 2 E be an accumulation point of E and let

f : E ! C and g : E ! C be di¤ erentiable at z0 . Prove that

(i) f + g is di¤ erentiable at z0 and (f + g)0 (z0 ) = f 0 (z0 ) + g 0 (z0 ),

(ii) f g is di¤ erentiable at z0 and (f g)0 (z0 ) = g(z0 )f 0 (z0 ) + f (z0 )g 0 (z0 ),
f
(iii) if g(z0 ) 6= 0 then g : E0 ! C is di¤ erentiable at z0 and
0
f g(z0 )f 0 (z0 ) f (z0 )g 0 (z0 )
(z0 ) = ;
g (g(z0 ))2

where E0 := fz 2 E : g(z) 6= 0g.

Exercise 8 Let E; F C, let z0 2 E be an accumulation point of E , let

f : E ! F be di¤ erentiable at z0 , let f (z0 ) be an accumulation point of f (E)
and let g : F ! C be di¤ erentiable at f (z0 ). Prove that g f is di¤ erentiable at
z0 and
(g f )0 (z0 ) = g 0 (f (z0 ))f 0 (z0 ):

Exercise 9 Let U; V C be open sets, let f : U ! V be continuous and let

g : V ! C be di¤ erentiable and such that

g(f (z)) = z for all z 2 U:

Let z0 2 U be such that g 0 (f (z0 )) 6= 0. Prove that f is di¤ erentiable at z0 and

1
f 0 (z0 ) = :
g 0 (f (z0 ))

Let’s discuss the relation between complex and real di¤erentiation. Given
E C and f : E ! C, let

F := f(x; y) 2 R2 : x + iy 2 Eg

and de…ne u : F ! R and v : F ! R by

u(x; y) := Re f (x + iy); v(x; y) := Im f (x + iy): (7)

The following example shows that di¤erentiability of u and v does not imply
di¤erentiability of f .

6
Example 10 Consider the function f (z) = z. Then u(x; y) = x and v(x; y) =
y, which are C 1 and even analytic functions. However, f is not di¤ erentiable
at 0, since
f (z) f (0) z
lim = lim
z!0 z 0 z!0 z

and this limit does not exist, since taking z = x + i0 gives

z x
lim = lim = 1;
z!0 z x!0 x

while taking z = 0 + iy gives

z y
lim = lim = 1:
z!0 z y!0 y
Wednesday, January 15, 2020
We recall that given a set F RN , a point x0 2 F \ acc F , and a real-valued
function u : F ! R, we say that u is di¤erentiable at x0 if there exists a linear
function L : RN ! R such that
u(x) u(x0 ) L(x x0 )
lim = 0:
x!x0 kx x0 k

The linear function L is called the di¤ erential of f at x0 and is denoted df (x0 ).

Exercise 11 Let F RN , let x0 2 F , and let u : F ! R be di¤ erentiable at

x0 .

(i) Prove that u is continuous at x0 .

@u
(ii) Prove that there exist all partial derivatives @xi (x0 ), all directional deriv-
atives @@u (x0 ) and that

@u
ru(x0 ) = (x0 ) (8)
@
for all 2 RN n f0g.

Exercise 12 Let u : R2 ! R be de…ned by

x if y = x2 ; x 6= 0;
u(x; y) :=
0 otherwise.

Prove that u is continuous at (0; 0), all partial and directional derivatives exist
at (0; 0) and that (8) holds but that u is not di¤ erentiable at (0; 0).

Next we show that di¤erentiability of f implies the di¤erentiability of u and

v. In what follows, given a set E C, we denote by E the set of interior points
of E.

7
Theorem 13 (Cauchy–Riemann Equations) Let E C, let z0 = x0 +
iy0 2 E be an accumulation point of E and let f : E ! C be di¤ erentiable
at z0 . Then the functions u and v de…ned in (7) are di¤ erentiable at (x0 ; y0 ).
Moreover if z0 is an interior point of E, then
@u @v
(x0 ; y0 ) = (x0 ; y0 ) = Re f 0 (x0 + iy0 ); (9)
@x @y
@u @v
(x0 ; y0 ) = (x0 ; y0 ) = Im f 0 (x0 + iy0 ):
@y @x
In particular,
!
@u @u
@x (x0 ; y0 ) @y (x0 ; y0 )
det @v @v = jf 0 (z0 )j2 (10)
@x (x0 ; y0 ) @y (x0 ; y0 )

The relations
@u @v @u @v
(x0 ; y0 ) = (x0 ; y0 ); (x0 ; y0 ) = (x0 ; y0 ) (11)
@x @y @y @x
are known as the Cauchy–Riemann equations.
Proof. We have
f (z) f (z0 ) f 0 (z0 )(z z0 )
0 = lim
z!z0 z z0
and so (since the product of a bounded function and a function going to zero
goes to zero)
f (z) f (z0 ) f 0 (z0 )(z z0 )
0 = lim :
z!z0 jz z0 j
In turn,
Re(f (z) f (z0 ) f 0 (z0 )(z z0 ))
lim = 0; (12)
z!z0 jz z0 j
Im(f (z) f (z0 ) f 0 (z0 )(z z0 ))
lim = 0: (13)
z!z0 jz z0 j

Now by (2), writing z = x + iy and z0 = x0 + iy0 ,

f 0 (z0 )(z z0 ) = Re f 0 (z0 )(x x0 ) Im f 0 (z0 )(y y0 )

0
+ i(Im f (z0 )(x x0 ) + Re f 0 (z0 )(y y0 ));

and so (12) and (13) become

Re f (x + iy) Re f (x0 + iy0 ) Re f 0 (x0 + iy0 )(x x0 ) + Im f 0 (x0 + iy0 )(y y0 )

lim p = 0;
(x;y)!(x0 ;y0 ) (x x0 )2 + (y y0 )2
Im f (x + iy) Im f (x0 + iy0 ) Im f 0 (x0 + iy0 )(x x0 ) Re f 0 (x0 + iy0 )(y y0 )
lim p = 0:
(x;y)!(x0 ;y0 ) (x x0 )2 + (y y0 )2

8
These can be written as
u(x; y) u(x0 ; y0 ) (Re f 0 (x0 + iy0 ); Im f 0 (x0 + iy0 )) ((x x0 ); (y y0 ))
lim p = 0;
(x;y)!(x0 ;y0 ) (x x0 )2 + (y y0 )2
v(x; y) v(x0 ; y0 ) (Im f (x0 + iy0 ); Re f 0 (x0 + iy0 )) ((x
0
x0 ); (y y0 ))
lim p = 0:
(x;y)!(x0 ;y0 ) (x x0 )2 + (y y0 )2

These implies that u and v are di¤erentiable at (x0 ; y0 ) with

du(x0 ; y0 )(s; t) = Re f 0 (x0 + iy0 )s Im f 0 (x0 + iy0 )t; (s; t) 2 R2; ;

dv(x0 ; y0 )(s; t) = Im f 0 (x0 + iy0 )s + Re f 0 (x0 + iy0 )t; (s; t) 2 R2 :

In particular, if z0 belongs to the interior of E then ru(x0 ; y0 ) and rv(x0 ; y0 )

exist with
@u @u
(x0 ; y0 ) = Re f 0 (x0 + iy0 ); (x0 ; y0 ) = Im f 0 (x0 + iy0 ) (14)
@x @y
@v @v
(x0 ; y0 ) = Im f 0 (x0 + iy0 ); (x0 ; y0 ) = Re f 0 (x0 + iy0 ): (15)
@x @y

Comparing (14) and (15) gives the Cauchy–Riemann equations. In turn, (10)
follows by direct computation.

Corollary 14 Let U C be an open and connected set and let f : U ! C be a

di¤ erentiable function with f 0 = 0 in U . Then f is constant.

Proof. By the previous theorem the functions u and v de…ned in (7) are
di¤erentiable in V = f(x; y) 2 R2 : x + iy 2 U g, with ru = rv (0; 0) in
V . Thus, by a result in Analysis, u and v are constant in V . Again by (7), it
follows that f is constant.

Theorem 15 Let F R2 , let (x0 ; y0 ) 2 F be an interior point of F and let

u; v : E ! R be di¤ erentiable at (x0 ; y0 ). Assume that the Cauchy–Riemann
equations (11) hold at (x0 ; y0 ). Let E := fz = x + iy 2 C : (x; y) 2 F g and let
f : E ! C be de…ned by

f (z) = u(x; y) + iv(x; y); z = x + iy 2 E: (16)

Then f is di¤ erentiable at z0 .

Proof. Set
@u @v
f 0 (z0 ) := (x0 ; y0 ) + i (x0 ; y0 ):
@x @x

9
Now by (2), writing z = x + iy and z0 = x0 + iy0 ,

f 0 (z0 )(z z0 ) = Re f 0 (z0 )(x x0 ) Im f 0 (z0 )(y y0 )

+ i(Im f (z0 )(x x0 ) + Re f 0 (z0 )(y y0 ))
0

@u @v
= (x0 ; y0 )(x x0 ) (x0 ; y0 )(y y0 )
@x @x
@v @u
+ i (x0 ; y0 )(x x0 ) + i (x0 ; y0 )(y y0 )
@x @x
@u @u
= (x0 ; y0 )(x x0 ) + (x0 ; y0 )(y y0 )
@x @y
@v @v
+ i (x0 ; y0 )(x x0 ) + i (x0 ; y0 )(y y0 );
@x @y

where in the last equality we used (11) ( @u

@x (x0 ; y0 ) =
@v
@y (x0 ; y0 ) and @u
@y (x0 ; y0 ) =
@v
@x (x0 ; y0 )). Hence, also by (16),

f (z) f (z0 ) f 0 (z0 )(z z0 ) = u(x; y) u(x0 ; y0 ) ru(x0 ; y0 ) (x x0 ; y y0 )

+ i(v(x; y) v(x0 ; y0 ) rv(x0 ; y0 ) (x x0 ; y y0 )):
p
Dividing by jz z0 j = (x x0 )2 + (y y0 )2 gives

f (z) f (z0 ) f 0 (z0 )(z z0 ) u(x; y) u(x0 ; y0 ) ru(x0 ; y0 ) (x x0 ; y y0 )

= p
jz z0 j (x x0 )2 + (y y0 )2
v(x; y) v(x0 ; y0 ) rv(x0 ; y0 ) (x x0 ; y y0 )
+i p :
(x x0 )2 + (y y0 )2

Since u and v are di¤erentiable at (x0 ; y0 ), it follows that

f (z) f (z0 ) f 0 (z0 )(z z0 )

0 = lim ;
z!z0 jz z0 j
which implies that

f (z) f (z0 ) f 0 (z0 )(z z0 )

0 = lim ;
z!z0 z z0
and the proof is complete.
The following example shows that the previous theorem fails without assum-
ing that u and v are di¤erentiable. We refer to Section 3 for the de…nition of
ez .

Example 16 Let
4
exp( z ) if z 6= 0;
f (z) =
0 if z = 0:
Prove that the Cauchy–Riemann equations are satis…ed but that f is not di¤ er-
entiable at the origin.

10
Note that the previous function is not continuous at z = 0. There is a
beautiful theorem, due to Looman and Mencho¤, which we will not prove, which
says the following.
Theorem 17 (Looman–Mencho¤) Let V R2 be an open set, let u; v :
V ! R be continuous functions in V . Assume that @u @u @v @v
@x , @y , @x , and @y exist in
V and satisfy the Cauchy–Riemann equations (11) in V . Let U := fz = x+iy 2
C : (x; y) 2 V g and let f : U ! C be de…ned by
f (z) = u(x; y) + iv(x; y); z = x + iy 2 U:
Then f is di¤ erentiable in U .

3 Power Series and Some Elementary Functions

De…nition 18 Given a sequence fzn gn of complex numbers, we call the n-th
partial sum the number
sn = z1 + + zn :
The sequence fsn gn of partial sums is called in…nite series or series and is
denoted
X1
zn :
n=1
P1
If there exists limn!1 sn = S 2 C, we say that the series n=1 zn is conver-
gent. The number S is called sum P1of the series. If the limit limn!1 sn does
not exist, we say that the series
P1 n=1 zn oscillates. P1
We say that the series n=1 zn converges absolutely if the series n=1 jzn j
converges.
RemarkP 19 There isPnothing special about 1, we will also consider series of
1 1
the type n=0 zn or n=n0 zn , where n0 2 N. The only change is that in the
partial sums, one should consider sn = z0 + + zn and sn = zn0 + + zn ,
respectively.
P1
Theorem 20 If the series n=1 zn converges, then there exists
lim zn = 0:
n!1
P1
Proof. Since the series n=1 zn converges, there exists limn!1 sn = S 2 C.
Hence,
zn = sn+1 sn ! S S = 0
as n ! 1. Note that here it is important that S 2 C.
De…nition 21 A power series is a series of the form
1
X
an z n ; z 2 C;
n=0

where an 2 C.

11
Friday, January 17, 2020
We recall that given a sequence fxn gn of real numbers, the limit superior of
fxn gn is de…ned as
lim sup xn := inf sup xk :
n!1 n k n

Exercise 22 Given a sequence fxn gn of real numbers and ` 2 R, prove that `

is the limit superior of the sequence fxn gn if and only if

(i) for every " > 0 there exists n" 2 N such that

xn `+" for all n n"

(ii) for every " > 0,

xn ` " for in…nitely many n.

State and prove a similar result for the case ` = 1.

Theorem 23 Given a power series

1
X
an z n ; z 2 C;
n=0

let R 2 [0; 1] be given by

1
:= lim sup jan j1=n :
R n!1

Then for jzj < R the series converges absolutely, while for jzj > R the series
oscillates.

Proof. If jzj < R, then jzj=R < 1. Fix " > 0 so small that (1=R + ")jzj < 1.
By the previous exercise, there exists N 2 N such that

jan j1=n 1=R + "

for all n N , and so

n
jan j (1=R + ")
for all n N . In turn,

jan z n j = jan jjzjn [(1=R + ") jzj]n

P
for all n N . Since (1=R +P") jzj < 1, the geometric series n=1 [(1=R + ") jzj]n
1
converges. Hence, so does n=1 jan z n j by the comparison test.
On the other hand, if jzj > R, …x " > 0 so small that (1=R ")jzj > 1. By
the previous exercise,
jan j1=n 1=R " > 0

12
for in…nitely many n, and so
n
jan j (1=R ")

for in…nitely many n. In turn,

jan z n j = jan jjzjn [(1=R ") jzj]n

Thus,
lim sup jan z n j lim sup[(1=R ") jzj]n = 1;
n!1 n!1
P1
since (1=R ") jzj > 1. It follows by Theorem 20, that the series n=1 an z n
oscillates.
The number R is called radius of convergence of the power series.
Pl
Exercise 24 Given fak gnk=1 and fbk gnk=1 in C, let Bl := k=1 bk , B0 := 0.
Prove that
n
X n
X1
ak bk = an Bn am Bm 1 (ak+1 ak )Bk :
k=m k=m
P1
Exercise 25 Assume that the series of complex numbers n=1 an converges.
Use the previous exercise to show that
1
X 1
X
lim an rn = an :
r!1
n=1 n=1

Exercise 26 Version of Abel’s with angles. Ahlfors.

Example 27 When jzj = R anything can happen as the two power series

X1 X1
1 n 1 n
z ; 2
z
n=1
n n=1
n

show. Note that for a > 0,

1=n
1 1 1 1 1 1
= = = ! = :
na na=n elog na=n e(a=n) log n e0 R

Exercise 28 Let fxn gn be a sequence of real numbers, with xn > 0 for all
n 2 N. Prove that
xn+1 p p xn+1
lim inf lim inf n
xn lim sup n
xn lim sup :
n!1 xn n!1 n!1 n!1 xn
Show that the inequality can be strict.

13
Remark 29 In view of the previous exercise, if there exists
xn+1
lim
n!1 xn
then there exists
p
lim n
xn
n!1
and the two limits are the same.

Next we show that a power series is di¤erentiable in B(0; R):

Theorem 30 Given a power series

1
X
an z n ; z 2 C;
n=0

let R be P
its radius of convergence and assume that R > 0. Then the function
1
f (z) := n=0 an z n is di¤ erentiable in the open set UR := fz 2 C : jzj < Rg
and
1
X
f 0 (z) = nan z n 1 :
n=0
P1
Moreover, the power series n=0 nan z n 1
has the same radius of convergence
R.

Proof. The fact that the two power series have the same radius of conver-
gence follows from the fact that
1=n
lim sup jnan j1=n = lim sup n1=n jan j1=n = lim sup elog n jan j1=n
n!1 n!1 n!1
1
= lim sup e(log n)=n jan j1=n = lim e(log n)=n lim sup jan j1=n = 1 :
n!1 n!1 n!1 R
Let z0 2 UR and …nd jz0 j < r < R. Let h 2 C be so small that jz0 + hj < r.
De…ne
1
X
g(z) := nan z n 1
n=0
and consider
1
X 1
X
f (z0 + h) f (z0 ) (z0 + h)n z0n
g(z0 ) = an nan z0n 1
h n=0
h n=0
N
X n
(z0 + h) z0n
= an nz0n 1

n=0
h
1
X 1
X
(z0 + h)n z0n
+ an nan z0n 1
h
n=N +1 n=N +1

=: I + II + III:

14
Using the facts that an bn = (b a)(an 1 + an 2
b+ + abn 2
+ bn 1
), that
jz0 j < r, and that jz0 + hj < r we have that
j(z0 + h)n z0n j jhjnrn 1
:
In turn,
1
X
jIIj jan jnrn 1
:
n=N +1
Since g has the same radius of convergence than f and r < R we have that
the right-hand side is the tail of a convergent series and thus goes to zero as
N ! 1. Hence, given " > 0 we can …nd N" 2 N such that jIIj " for all
N N" .
Similarly, since jz0 j < R and g(z0 ) converges, by taking N" larger, if neces-
sary, we have that jIIIj " for all N_ N" .
Fix N = N" . Since I is the di¤erence quotient of a …nite number of di¤eren-
tiable functions, we can …nd " > 0 such that jIj " for all h 2 C with jhj ".
This concludes the proof.
By repeated applications of the previous theorem we obtain the following:
P1
Corollary 31 A power series f (z) = n=0 an z n is in…nitely di¤ erentiable in
UR := fz 2 C : jzj < Rg, where R is its radius of convergence. Moreover,
the higher derivatives f (k) are power series obtained by pointwise di¤ erentiation
and with the same radius of convergence. To be precise,
1
X
f (k) (z) = n(n 1) (n k + 1)an z n k
; z 2 UR :
n=k

Moreover,
1
f (k) (0) =
ak ; k 2 N0 :
k!
P1
Remark 32 Similarly, if we consider f (z) = n=0 an (z z0 )n , then f is
1
in…nitely di¤ erentiable in UR := fz 2 C : jz z0 j < Rg, with f (k) (z0 ) = k! ak ,
k 2 N0 .
Using power series we can de…ne ez , cos z, and sin z as follows
X1 1
X 1
X
1 n 1 2n 1
ez := z ; cos z := ( 1)n z ; sin z := ( 1)n z 2n+1 :
n=0
n! n=0
(2n)! n=0
(2n + 1)!
Using Remark 29, we compute
1 1
(n+1)! n!(n+1) 1
1 = 1 = ! 0;
n! n!
n+1
1 1
( 1)n+1 (2n+2)! (2n)!(2n+2)(2n+1) 1
1 = 1 = ! 0;
( 1)n (2n)! (2n)!
(2n + 2)(2n + 1)
1 1
( 1)n+1 (2n+3)! (2n+1)!(2n+3)(2n+2) 1
1 = 1 = ! 0;
( 1)n (2n+1)! (2n+1)!
(2n + 3)(2n + 2)

15
and so all three series have radius of convergence R = 1, so they converge in
C. For cos z we used the fact that z 2n = (z 2 )n and for sin z we pulled out z and
used the same trick.
Note that if we di¤erentiate ez , by Theorem 30,
X1 1
X
n n 1 k
(ez )0 = z 1
= z = ez ;
n=1
n! k!
k=1

which is the same property of the real exponential function. Similarly, by The-
orem 30,
(cos z)0 = sin z; (sin z)0 = cos z:
Consider the function g(z) = ez ea z
, where a 2 C is …xed. By the product
rule, Exercise 8, and Theorem 30,

g 0 (z) = ez ea z
ez ea z
=0

and so by Corollary 14, g is constant. Since e0 = 1, taking z = 0 gives g(z) ea .

Hence,
ez ea z = ea for all z 2 C:
Taking a = z + w we get

ez ew = ez+w for all z; w 2 C: (17)

Taking w = z gives ez e z
= 1 so ez 6= 0 for all z and
1 z
=e : (18)
ez
1 n 1 n
Observe also that by (4), n! z = n! z = z n and so ez = ez . In turn, by (3)
and (17),
jez j2 = ez ez = ez ez = ez+z = e2 Re z : (19)
Note that
eiz + e iz eiz e iz
cos z = ; sin z = : (20)
2 2i
These are called Euler formulas for cos z and sin z. From these formulas and
(17) we obtain

e2iz + e 2iz
+ 2eiz e iz
e2iz + e 2iz
2eiz e iz
cos2 z + sin2 z = =1
4 4
and
eiz = cos z + i sin z; (21)
which is what we used in (6).

Exercise 33 Let x + iy 2 C. Prove that

j cos zj2 = cos2 x + cosh2 y; j sin zj2 = sin2 x + sinh2 y;

16
Monday, January 20, 2020
MLK Day, no classes.
Wednesday, January 22, 2020
Next we study the periodicity of ez . We say that a function f : C ! C is
periodic with period w 2 C if

f (z + w) = f (z) for all z 2 C:

Given w 2 C, assume that

ez = ez+w
for all z 2 C. Then by multiplying both sides by e z and using (17) we get
1 = ew . Taking the modulus on both sides and using 19 we get

1 = e2 Re w ;

which implies that Re w = 0. Thus, w = i for some 2 R. In turn, by (21),

1 = ei = cos + i sin ;

and so
= 2 k; k 2 Z:
This shows that the exponential function is periodic with period 2 i. This is
one of the main di¤erences with the real exponential. In particular, this implies
that ez is not one-to-one in C. Thus, we cannot de…ne the complex logarithmic
function as the inverse of the complex exponential function.

De…nition 34 Given a connected open set U C, a branch of the logarithm

is a continuous function f : U ! C such that

z = ef (z) for all z 2 U:

We sometimes write f = logU .

Remark 35 Note that since ez 6= 0 for all z, in order for a branch of the
logarithm to exist in U , we must have 0 2
= U.

Exercise 36 Let

W := C n fz 2 C : z = x + 0i; x 0g:

For every z 2 W , write z = rei , r = jzj, < < , and de…ne

f (z) := log r + i :

(i) Prove that f is branch of the logarithm in W .

(ii) Prove that for all z 2 B(0; 1) with 1 + z 2 W ,
1
X zn
f (1 + z) = ( 1)n :
n=1
n

17
(iii) Prove that in general
f (z1 z2 ) 6= f (z1 ) + f (z2 ):

The branch of the logarithm constructed in the previous exercise is called

the principal branch of the logarithm.
Proposition 37 Let U C be an open connected set and let f : U ! C be a
branch of of the logarithm. Then f is di¤ erentiable in U , with
1
f 0 (z) = for all z 2 U:
z
Moreover, every other branch of the logarithm in U has the form
g(z) = f (z) + 2k i
for some k 2 Z.
Proof. The di¤erentiability of f follows from Exercise 9. By the chain rule
(see Exercise 8) and the de…nition of f ,
1 = ef (z) f 0 (z) = zf 0 (z) for all z 2 U;
which implies that f 0 (z) = z1 .
Given k 2 Z, consider the function g(z) := f (z) + 2k i, z 2 U . Then the
periodicity of the exponential
eg(z) = ef (z)+2k i
= ef (z) = z;
which shows that g is a branch of log z.
Conversely, assume that g : U ! C is another branch of log z. Then the
function
1
h(z) := (f (z) g(z)); z 2 U;
2 i
is continuous and since
1
e2 ih(z) = ef (z) g(z) = ef (z) e g(z) = z = 1;
z
by (18), we have that 2 ih(z) = 2k i for some k 2 Z (depending on z). This
shows that h(U ) Z, but since h is continuous and U is connected, h must be
constant, and thus there is k0 2 Z such that h(z) = k0 for all z 2 U , which
completes the proof.
If U C is an open connected set and f : U ! C is a branch of the logarithm
in U , then for every a 2 C we de…ne a branch of z a as
g(z) := eaf (z) ; z 2 U:
In view of the previous theorem, g is di¤erentiable in U , since composition of
di¤erentiable functions, and every other branch is given by
h(z) = eaf (z)+a2k i
= eaf (z) ea2k i
= g(z)ea2k i :

18
4 Riemann-Stieltjes integrals
In what follows, given an interval [a; b] R, a partition of [a; b] is a …nite set
P := ft0 ; : : : ; tn g [a; b], where
a = t0 < t1 < < tn = b:
De…nition 38 Let g : [a; b] ! C be a function. The pointwise variation of g
on the interval [a; b] is
( n )
X
Var g := sup jg(tk ) g(tk 1 )j ;
k=1

where the supremum is taken over all partitions P := ft0 ; : : : ; tn g of [a; b], n 2 N.
A function g : [a; b] ! C has …nite or bounded pointwise variation if Var g < 1.
The space of all functions g : [a; b] ! C of bounded pointwise variation is
denoted by BV ([a; b]; C).
To highlight the dependence on the interval [a; b], we will sometimes write
Var[a;b] g.
Given a function g : [a; b] ! C, we say that g is piecewise C 1 , if g is
continuous, and there exists a partition P := ft0 : : : ; tn g [a; b] such that
g : [tk 1 ; tk ] ! C is of class C 1 for every k = 1; : : : ; n.
Exercise 39 Let g : [a; b] ! C be piecewise C 1 . Prove that
Z b
Var g = jg 0 (t)j dt:
a

Exercise 40 Let g : [a; b] ! C be Lipschitz continuous. Prove that g 2

BV ([a; b]; C).
Exercise 41 Let g : [a; b] ! R be a monotone function. Prove that
Var g = sup g inf g:
[a;b] [a;b]

Exercise 42 Let f; g 2 BV ([a; b] ; C). Prove the following.

(i) f g 2 BV ([a; b] ; C).
(ii) f g 2 BV ([a; b] ; C).
f
(iii) If jg(t)j c > 0 for all t 2 [a; b] and for some c > 0, then g 2
BV ([a; b] ; C).
Exercise 43 Let g : [a; b] ! C, and let c 2 [a; b]. Prove that
Var[a;c] g + Var[c;b] g = Var[a;b] g:

19
Exercise 44 Prove that g 7! Var g is a seminorm in BV ([a; b]; C).

Theorem 45 Let g 2 BV ([a; b]; C) and let f : [a; b] ! C be a continuous

function. Then there exists ` 2 C with the property that for every " > 0 there
exists " > 0 such that if P = ft0 ; : : : ; tn g is a partition of [a; b] with tk tk 1
" for all k = 1; : : : ; n, then

n
X
` f (sk )(g(tk ) g(tk 1 )) ";
k=1

for every sk 2 [tk 1 ; tk ], k = 1; : : : ; n.

The number ` is called the Riemann-Stieltjes integral of f with respect to g

over [a; b] and is denoted
Z b
`= f dg:
a

Exercise 46 Let g : [a; b] ! C be piecewise C 1 and let f : [a; b] ! C be a

continuous function. Prove that
Z b Z b
f dg = f (t)g 0 (t) dt:
a a

Exercise 47 Let g 2 BV ([a; b]; C), let f : [a; b] ! C be a continuous functions,

and P = ft0 ; : : : ; tn g be a partition of [a; b] with a = t0 and b = tn . Prove that
Z b n Z
X tk
f dg = f dg:
a k=1 tk 1

Exercise 48 Let g 2 BV ([a; b]; C), let f : [a; b] ! C be a continuous functions.

Prove that Z b
f dg max jf j Var g:
a [a;b]

Exercise 49 Let g 2 BV ([a; b]; C), let f1 ; f2 : [a; b] ! C be continuous func-

tions, and let ; 2 C. Prove that
Z b Z b Z b
( f1 + f2 ) dg = f1 dg + f2 dg:
a a a

Exercise 50 Let g1 ; g2 2 BV ([a; b]; C), let f : [a; b] ! C be continuous func-

tions, and let ; 2 C. Prove that
Z b Z b Z b
f d( g1 + g2 ) = f dg1 + f dg2 :
a a a

20
Exercise 51 Let g 2 BV ([a; b]; C), let f : [a; b] ! C be a continuous functions,
and P = ft0 ; : : : ; tn g be a partition of [a; b] with a = t0 and b = tn . Prove that
Z b n Z
X tk
f dg = f dg:
a k=1 tk 1

Monday, January 20, 2020

We turn to the proof of Theorem 45.
1
Proof of Theorem 45. Since f is uniformly continuous, given " = m we
can …nd m > 0 such that
1
jf (z) f (w)j (22)
m
for all z; w 2 [a; b] with jz wj m . By an induction argument, we can assume
that m m+1 for all m 2 N. For each m let Pm be the family of all partitions
P = ft0 ; : : : ; tn g of [a; b] with tk tk 1 m for all k = 1; : : : ; n. Note that
Pm+1 Pm for every m. Let

Em := fS(P ) : P = ft0 ; : : : ; tn g 2 Pm ; sk 2 [tk 1 ; tk ]; k = 1; : : : ; ng

where
n
X
S(P ) := f (sk )(g(tk ) g(tk 1 ));
k=1

and let Cm = E m . Since Pm+1 Pm , we have that Em+1 Em , and so

Cm+1 Cm .
Next we claim that
2
diam Cm Var g: (23)
m
To see this, let P; Q 2 Pm . Let P = ft0 ; : : : ; tn g and assume …rst that Q is
obtained from P by adding a point c and let k0 be such that tk0 1 < c < tk0 .
Then
X
S(Q) = f ( k )(g(tk ) g(tk 1 )) + f ( 0 )(g(c) g(tk0 1 ))
k6=k0
00
+ f( )(g(tk0 ) g(c));
0 0
where tk 1 k tk , tk0 1 c, c tk0 . In turn, by (22),
X
jS(Q) S(P )j jf ( k ) f (sk )jjg(tk ) g(tk 1 )j + jf ( 0 ) f (sk0 )jjg(c) g(tk0 1 )j
k6=k0

+ jf ( 00 ) f (sk0 )jjg(tk0 ) g(c)j

1 X 1 1
jg(tk ) g(tk 1 )j + jg(c) g(tk0 1 )j + jg(tk0 ) g(c)j
m m m
k6=k0
1
Var g:
m

21
1
With a similar proof, we can show that if P Q, then jS(Q) S(P )j m Var g.
Finally, if P; Q 2 Pm , let R 2 Pm be such that P; Q R, then
1 1
jS(Q) S(P )j jS(Q) S(R)j + jS(R) S(P )j Var g + Var g:
m m
By taking the supremum over all such partitions we conclude that diam Em
2
m Var g, and in turn, (23) follows.
It now follows from Cantor’s theorem that there exists a unique ` 2 C such
that
1
\
f`g = Cm :
m=1
2
Given " > 0 let m be so large that Var g < m and take " :=
" m. Since
` 2 Cm , we have that Cm B(`; "), which proves the theorem.

5 Line Integrals
De…nition 52 Given two functions ' : [a; b] ! C and : [c; d] ! C, we say
that they are equivalent if there exists a continuous, strictly increasing, onto
function h : [a; b] ! [c; d] such that

' (t) = (h (t))

for all t 2 [a; b]. We write ' and we call ' and parametric representations
and the function h a parameter change.
1
Note that in view of a theorem real analysis, h : [c; d] ! [a; b] is also
continuous.

Exercise 53 Prove that is an equivalence relation.

De…nition 54 An oriented curve is an equivalence class of parametric rep-

resentations.

Remark 55 The de…nition of a curve is a restrictive, although it is what we

will need it in this course. More generally, given two intervals I; J R, and
two functions ' : I ! C and : J ! C, we say that they are equivalent if there
exists a continuous, strictly increasing, onto function h : I ! J such that

' (t) = (h (t))

for all t 2 I. We write ' and we call ' and parametric representations
and the function h a parameter change.

Given an oriented curve with parametric representation ' : [a; b] ! C the

multiplicity of a point z 2 C is the (possibly in…nite) number of points t 2 [a; b]
such that ' (t) = z. Since every parameter change h : [a; b] ! [c; d] is bijective,

22
the multiplicity of a point does not depend on the particular parametric repre-
sentation. The range of is the set of points of C with positive multiplicity,
that is, '([a; b]).
A point in the range of with multiplicity one is called a simple point. If
every point of the range is simple, then is called a simple arc.
Given an oriented curve with parametric representation ' : [a; b] ! C, the
oriented curve 1 with parametric representation '1 : [a; b] ! C given by

'1 (t) := '( t + b + a)

is called the curve opposite to .

De…nition 56 Given two functions ' : [a; b] ! C and : [c; d] ! C of class

C k , k 2 N0 , we say that they are equivalent if there exists a strictly increasing,
onto function h : [a; b] ! [c; d] with h and h 1 of class C k such that

' (t) = (h (t))

for all t 2 [a; b]. We write ' k and we call ' and parametric repre-
sentations of class C k and the function h a parameter change of class C k . An
oriented curve of class C k is an equivalence class of parametric representations
of class C k .

Similarly we can de…ne C 1 oriented curves, Lipschitz oriented curves, ana-

lytic oriented curves, and so on.
Given a continuous curve, the points '(a) and '(b) are called endpoints of
the curve. If ' (a) = ' (b), then the oriented curve is called a closed oriented
curve. A closed curve is called simple if every point of the range is simple, with
the exception of ' (a), which has multiplicity two.
The following theorem will be used in the sequel.

Theorem 57 (Jordan’s curve theorem) Given a continuous closed simple

oriented curve in C with range , the set C n consists of two connected
components.

The bounded connected component of C n is called the interior of .

We are ready to de…ne the notion of length of a curve.

Exercise 58 Let be an oriented curve in C. Let ' : [a; b] ! C and : [c; d] !

C be two parametric representations of . Prove that Var[a;b] ' = Var[c;d] .

We are now ready to de…ne the length of a curve.

De…nition 59 Let be an oriented curve in C and let ' : [a; b] ! C be a

parametric representation of . We de…ne the length of as

L ( ) := Var ':

We say that the curve is recti…able if L ( ) < 1.

23
Theorem 60 Given a recti…able oriented curve in C with range and a
continuous function f : ! C, let ' : [a; b] ! C and : [c; d] ! C be two
parametric representations of . Then
Z b Z d
f ' d' = f d :
a c

24
Monday, January 27, 2020
No class
Wednesday, January 29, 2020
No class
Friday, January 31, 2020
2 hours
Proof. Since ' and are equivalent, there exists h : [c; d] ! [a; b] continu-
ous, strictly increasing, with h(c) = a and h(d) = b, such that
'(h(s)) = (s) for all s 2 [c; d]: (24)
By Theorem 45 for every " > 0 there exists " > 0 such that if P = ft0 ; : : : ; tn g
is a partition of [a; b] with tk tk 1 " for all k = 1; : : : ; n, then
Z b Xn
f ' d' f ('(t0k ))('(tk ) '(tk 1 )) "; (25)
a k=1

for every t0k 2 [tk 1 ; tk ], k = 1; : : : ; n. Similarly, there exists " > 0 such that if
Q = fs0 ; : : : ; sm g is a partition of [c; d] with sl sl 1 " for all l = 1; : : : ; m,
then Z d Xm
f d f ( (s0l ))( (sl ) (sl 1 )) "; (26)
c l=1
for every s0l
2 [sl 1 ; sl ], l = 1; : : : ; m. Since h is uniformly continuous, there
exists " > 0 such that
jh(s) h(s0 )j "

for all s; s0 2 [c; d] with js s0 j " . Let Q = fs0 ; : : : ; sm g be a partition of [c; d]

with sl sl 1 minf " ; " g for all l = 1; : : : ; m. Then P = fh(s0 ); : : : ; h(sm )g
is a partition of [a; b] with h(sk ) h(sk 1 ) " . Hence, (25) holds for this
partition,On the other hand, by (24), '(h(sl )) = (sl ) and so
m
X m
X
f ('(h(s0l )))('(h(sl )) '(h(sl 1 ))) = f ( (s0l ))( (sl ) (sl 1 )):
l=1 l=1

Hence, by (25) and (26),

Z b Z d Z b m
X
f ' d' f d f ' d' f ('(h(s0l )))('(h(sl )) '(h(sl 1 )))
a c a l=1
Z d m
X
+ f d f ( (s0l ))( (sl ) (sl 1 )) 2":
c l=1

Letting " ! 0+ gives the result.

Given a recti…able oriented curve in C parametrized by ' : [a; b] ! C and
a continuous function f : '([a; b]) ! C, the line integral of f over is de…ned
as Z Z b
f dz := f ' d':
a

25
In view of the previous theorem, the integral does not depend on the particular
representation of the curve.
Note that all the properties in the exercises in the previous section continue
to hold for line integrals.

De…nition 61 Let U C be an open set and let f : U ! C. We say that f

has a primitive in U if there exists a holomorphic function F : U ! C such that
F0 = f.

Remark 62 The function f (z) = az n , where a 2 C and n 2 N0 has a primitive

a
given by F (z) = n+1 z n+1 + c.

Theorem 63 (Fundamental theorem of calculus) Let U C be an open

set, let f : U ! C be a continuous function, which has a primitive F in U .
Then for every z1 ; z2 2 U and for every recti…able continuous oriented curve
starting at z1 and ending at z2 and with range in U ,
Z
f dz = F (z2 ) F (z1 ):

We begin with a preliminary result.

Lemma 64 Let U C be an open set, let f : U ! C be a continuous function,

and let be a recti…able continuous oriented curve with range in U . Then for
every " > 0 there exists a polygonal path " with the same endpoints of and
range in U such that Z Z
f dz f dz "
"

Proof. Step 1: Assume …rst that U = B(z0 ; r). Let ' : [a; b] ! C
be a parametric representation of . Since '([a; b]) is compact, we have that
dist('([a; b]); @U ) = > 0. Hence, '([a; b]) B(z0 ; r ) =: K. Since f is
uniformly continuous on K, given " > 0, there exists " > 0 such that

jf (z) f (w)j " (27)

for all z; w 2 K with jz wj ".

Since ' : [a; b] ! C is uniformly continuous, there exists " > 0 such that

j'(t) '(s)j " (28)

for all s; t 2 [a; b] with js tj " . Moreover, by Theorem 45, there exists
" > 0 such that if P = ft0 ; : : : ; tn g is a partition of [a; b] with tk tk 1 "
for all k = 1; : : : ; n, then
Z n
X
f dz f ('(sk ))('(tk ) '(tk 1 )) "; (29)
k=1

26
for every sk 2 [tk 1 ; tk ], k = 1; : : : ; n. Consider a partition P = ft0 ; : : : ; tn g is
a partition of [a; b] with tk tk 1 minf " ; " g. Let '" be the polygonal path
joining '(t0 ), . . . , '(tn ). To be precise
1
'" (t) := [(t tk 1 )'(tk )+(tk t)'(tk 1 )]; t 2 [tk 1 ; tk ]; k = 1; : : : ; n:
tk tk 1

Note that
'(tk ) '(tk 1)
'0" (t) = ; t 2 (tk 1 ; tk ); k = 1; : : : ; n:
tk tk 1
Hence, by Exercise 46,
Z Z b n
X Z tk
0 '(tk ) '(tk 1)
f dz = f ('" (t))'" (t) dt = f ('" (t)) dt:
" a tk tk 1 tk 1
k=1

Hence, by (29),
Z Z Z n
X
f dz f dz f dz f ('(sk ))('(tk ) '(tk 1 ))
" k=1
n
X n
X Z tk
'(tk ) '(tk 1)
+ f ('(sk ))('(tk ) '(tk 1 )) f ('" (t)) dt
tk tk 1 tk 1
k=1 k=1
Xn Z tk n
X Z tk
'(tk ) '(tk 1) '(tk ) '(tk 1)
"+ f ('" (sk )) dt f ('" (t)) dt
tk tk 1 tk 1
tk tk 1 tk 1
k=1 k=1
n
X Z tk
j'(tk ) '(tk 1 )j
"+ jf ('(sk )) f ('" (t))j dt:
tk tk 1 tk 1
k=1

For t 2 [tk 1 ; tk ] we have

1
'(sk ) '" (t) = [(t tk 1 )('(sk ) '(tk )) + (tk t)('(sk ) '(tk 1 )]
tk tk 1

and so by (28),
1
j'(sk ) '" (t)j [(t tk 1 )j'(sk ) '(tk )j+(tk t)j'(sk ) '(tk 1 )j] "
tk tk 1

In turn, by (27), jf ('(sk )) f ('" (t))j ". Using this inequality we have that
Z Z Xn Xn
j'(tk ) '(tk 1 )j
f dz f dz "+ "(tk tk 1 ) " + " j'(tk ) '(tk 1 )j
"
tk tk 1
k=1 k=1
" + "L( ):
Step 2: For a generic open set, since '([a; b]) is compact, as before dist('([a; b]); @U ) >
0. Let 0 < < dist('([a; b]); @U ). Since ' is uniformly continuous, there exists
> 0 such that
j'(t) '(s)j <

27
for all s; t 2 [a; b] with js tj . Consider a partition P = ft0 ; : : : ; tn g of
[a; b] with tk tk 1 for all k = 1; : : : ; n. It follows that '([tk 1 ; tk ])
B('(tk 1 ); ) and so we may apply the previous step to the curve k parame-
trized by ' : [tk 1 ; tk ] ! C to …nd a polygonal path k with endpoints '(tk 1 )
and '(tk ) such that Z Z
f dz f dz "=n:
k k

By joining 1 , . . . , n we get a polygonal path joining '(a) and '(b). The result
now follows from the previous inequality and Exercise 47.
We turn to the proof of the fundamental theorem of calculus.
Proof. In view of the previous lemma, for every " > 0 there exists a
polygonal path " with endpoints z1 and z2 such that
Z Z
f dz f dz ":
"

Let '" : [a; b] ! C be a parametric representation of " . By Exercise 46,

Z Z b Z b Z b
f dz = f ('" (t))'0" (t) dt = F 0 ('" (t))'0" (t) dt = (F '" )0 (t) dt
" a a a
=F '" (b) F '" (a) = F (z2 ) F (z1 ):
Hence, Z
f dz (F (z2 ) F (z1 )) ":

Letting " ! 0+ completes the proof.

Corollary 65 Let U C be an open set, let f : U ! C be a continuous
function, which has a primitive F in U . Then
Z
f dz = 0

for every closed recti…able continuous oriented curve with range in U .

Exercise 66 Given a recti…able oriented curve in C with range and a con-
tinuous function f : ! C, let 1 be the curve opposite to . Prove that
Z Z
f dz = f dz:
1

6 Cauchy’s Theorem in a Ball

Theorem 67 (Goursat) Let U C be an open set and let f : U ! C be a
holomorphic function. Then for every closed triangle T U ,
Z
f dz = 0:
@T

28
Proof. Set T0 := T , bisect each side of T0 and connect the middle points.
This creates four triangles T1;1 , T1;2 , T1;3 , and T1;4 . By choosing an orientation
for these triangles consistent with the one of T0 and by canceling the sides which
are integrated in two opposite directions (see Exercises 47 and 66), we get
Z Z Z Z Z
f dz = f dz + f dz + f dz + f dz:
@T0 @T1;1 @T1;2 @T1;3 @T1;4

Hence, for some j 2 f1; 2; 3; 4g,

Z Z
f dz 4 f dz :
@T0 @T1;j

Let T1 := T1;j . Note that L (@T1 ) = 21 L(@T0 ) and diam T1 = 12 diam T0 . We

now bisect the sides of T1 and connect the middle points. Inductively we obtain
a decreasing sequence of closed triangles Tn such that
Z Z
f dz 4n f dz ; (30)
@T0 @Tn

L (@Tn ) = 21n L(@T0 ) and diam Tn = 21n diam T0 . By Cantor’s theorem there
exists z0 2 Tn for all n. Since f is di¤erentiable, we can write
f (z) = f (z0 ) + f 0 (z0 )(z z0 ) + R(z);
where
R(z)
= 0: lim
z z0 z!z0

Since a constant function and a linear function az have a primitive, by the

fundamental theorem of calculus,
Z Z
f dz = R dz:
@Tn @Tn

Since z0 2 Tn and z 2 @Tn , we have

jR(z)j "n jz z0 j "n diam Tn ;
where "n ! 0+ . Hence,
Z Z
f dz = R dz "n (diam Tn )L (@Tn )
@Tn @Tn
1
"n L(@T0 ) diam T0 :
4n
Using (30) we get
Z Z
n
f dz 4 f dz "n L(@T0 ) diam T0 ! 0
@T0 @Tn

as n ! 1.

29
Exercise 68 Let U C be an open set, let z0 2 U , and let f : U ! C be
a continuous function, which is holomorphic in U n fz0 g. Prove that for every
closed triangle T U , Z
f dz = 0:
@T
Hint: Consider …rst the case in which z0 is a vertex of T .

Saturday, February 1, 2020

Make-up class.
As a corollary we get

Corollary 69 Let U C be an open set and let f : U ! C be a holomorphic

function. Then for every closed rectangle R U ,
Z
f dz = 0:
@R

Proof. Divide R into two triangles and one side of the triangles is in common
and are integrated in two opposite directions.

Theorem 70 Let B C be an open ball and let f : B ! C be holomorphic.

Then f has a primitive in B.

Proof. Without loss of generality assume that B is centered at the origin.

Given z = x + iy 2 B, with x; y 2 R, we connect the origin to x + 0i and then
x + 0i to z and let z be this polygonal path in B. We choose the orientation
starting at 0 and ending at z. De…ne
Z
F (z) := fd :
z

We claim that F 0 = f . Let z + h 2 B. Then

Z Z
F (z + h) F (z) = fd fd :
z+h z

Using Goursat’s theorem for triangles and rectangles we are left with the seg-
ment Sz;h going from z to z + h. Given 2 Sz;h write f ( ) = f (z) + r( ), where
by continuity
lim r( ) = 0:
!z

Then
Z Z Z
F (z + h) F (z) = f d = f (z) 1d + rd
Sz;h Sz;h Sz;h
Z
= f (z)h + rd ;
Sz;h

30
where we used the fact that the constant 1 has a primitive. Hence,
Z
F (z + h) F (z) 1
f (z) = rd
h h Sz;h

and Z
1 jhj
rd max jrj = max jrj ! 0
h Sz;h Sz;h jhj Sz;h

as h ! 0.

Remark 71 In the previous proof we only used the fact that f is continuous
and for every closed triangle T B,
Z
f = 0: (31)
@T

Hence, if we assume that f : B ! C is a continuous function which is holomor-

phic in B n fz0 g for some z0 2 B, then by Exercise 68, (31) holds, and so f has
a primitive in B.

Corollary 72 (Cauchy) Let B be an open ball, let f : B ! C be holomorphic.

Then Z
f dz = 0

for every closed oriented curve with range in B.

Proof. This follows from the previous theorem and Corollary 72.

31
Remark 73 In view of Remark 71, Corollary 72 continues to hold if we assume
that f : B ! C is a continuous function which is holomorphic in B n fz0 g for
some z0 2 B.

Exercise 74 Let z0 = x0 + iy0 2 B(0; 1), let U C be the open set obtained
from B(0; 1) by removing the segment fx0 + yi : y y0 g, and let f : U ! C be
holomorphic. Prove that f has a primitive in U .

Exercise 75 Let U C be a star-shaped set and let f : U ! C be holomorphic.

Prove that Z
f dz = 0

for every closed oriented curve with range in U .

We are now ready to prove Cauchy’s integral formula.

Theorem 76 (Cauchy’s integral formula) Let U C be an open set, let

f : U ! C be holomorphic. Then for every open ball B with B U and every
z 2 B, Z
1 f( )
f (z) = d ;
2 i @B z
where @B is oriented counterclockwise.

Proof. Fix z 2 B and consider the closed curve ;" given in the picture
below, where " is the radius of the small circle centered at z and is the width
of the corridor. Since the function g( ) := f ( z) is holomorphic in U n fzg, by
considering V := B 0 n S 0 , where B 0 is a concentric ball contained in U and
containing B, S is the segment obtained when " ! 0 and ! 0, and S 0 is a
slightly larger segment we can apply Exercise 74, to obtain that g has a primitive
in V . Since the range of ;" is contained in V , it follows from Corollary 65 that
Z
f( )
d = 0:
;"
z

If we let ! 0+ and use the fact that g is continuous, we get that the two seg-
ments converge to a segment which is integrated in opposite directions. Hence,
we obtain Z Z
f( ) f( )
d d = 0:
@B z @B(z;") z
Write
f( ) f( ) f (z) f (z)
= + :
z z z
f ( ) f (z)
Since f is holomorphic, z ! f 0 (z) as ! z and so

f( ) f (z)
M
z

32
Figure 1: Figure 1: Keyhole contour

for all 2 @B(z; "). It follows that

Z Z Z
f( ) f( ) f (z) 1
d = d + f (z) d
@B(z;") z @B(z;") z @B(z;") z
= I + II:
Then jIj M (2 ") ! 0 as " ! 0+ . On the other hand, if we use the parame-
trization '(t) = z + "eit , t 2 [0; 2 ]. Then
Z Z 2
1 i"eit
d = dt = 2 i:
@B(z;") z 0 "eit
Hence, Z
f( )
d = f (z)2 i;
@B z
which proves the result.
Exercise 77 Use Remark 73 to give an alternative proof of the previous theo-
rem, which does not make use of ;" .
Exercise 78 Use contour integration to show that for 2 R,
Z
2 2
e = e x e 2 ix dx:
R

Exercise 79 Use contour integration to show that for 2 R,

Z 1
1 cos x
= dx:
2 0 x2

33
Exercise 80 Use contour integration to show that
Z 1
sin x
= dx:
2 0 x

34
Monday, February 3, 2020
Corollary 81 Let U C be an open set and let f : U ! C be holomorphic.
Then f is analytic and for every open ball B with B U , every z 2 B, and
every k 2 N, Z
k! f( )
f (k) (z) = d ; (32)
2 i @B ( z)k+1
where @B is oriented counterclockwise.
Proof. Let B = B(z0 ; r). Fix z 2 B. For 2 @B write
1 1 1 1
= = z z0 :
z z0 (z z0 ) z0 1 z0

Then
z0 jz z0 j
z
= =: < 1
z0 r
and so we can use geometric power series to write
1
X n
1 z z0
z z0 = :
1 z0 n=0
z0

Note that the series converges uniformly for all 2 @B, and so (using Lebesgue
dominated convergence theorem or any equivalent theorem for Riemann inte-
gration) we can interchange the integral and the series in Cauchy’s formula to
get
Z Z
1 f( ) 1 f( ) 1
f (z) = d = d
2 i @B z 2 i @B z0 1 z zz00
Z 1
f ( ) X z z0
n
1
= d
2 i @B z0 n=0 z0
X1 Z X1
1 f( )
= (z z0 )n n+1
d =: an (z z0 )n :
n=0
2 i @B ( z 0 ) n=0

The formula for the derivatives now follows by di¤erentiating the power series.
To see this, we use Corollary 31 to get
1
X
f (k) (z) = n(n 1) (n k + 1)an (z z0 )n k

n=k
1 Z
1 X n k f( )
= n(n 1) (n k + 1)(z z0 ) d
2 i @B ( z0 )n+1
n=k
1 Z
1 X f( ) (z z0 )n k
= n(n 1) (n k + 1) d
2 i ( z0 )k+1 ( z0 )n k
n=k @B
Z X1
1 f( ) (z z0 )n k
= n(n 1) (n k + 1) d :
2 i @B ( z0 )k+1 ( z0 )n k
n=k

35
z z0
Let w = z0 . Then
1
X 1
d(k) X n
n(n 1) (n k + 1)wn k
= w
dwk n=0
n=k
d(k) 1 k 1
= (1 w) = k!(1 w)
dwk
and so (using again the uniform convergence of the power series and its deriva-
tives)
Z 1
X
(k) 1 f( ) (z z0 )n k
f (z) = n(n 1) (n k + 1) d
2 i @B ( z0 )k+1 ( z0 )n k
n=k
Z Z
1 f( ) 1 k! f( )
= k! k+1
d = d ;
2 i @B ( z0 )k+1 z z0 2 i @B ( z)k+1
1 z0

which completes the proof.

Remark 82 Note that we have proved that for every open ball B(z0 ; r) with
B(z0 ; r) U , we can write
1
X
f (z) = an (z z0 )n ; z 2 B(z0 ; r);
n=0

where
f (n) (z0 )
:
an :=
n!
Moreover, if we denote by R the radius of convergence R of the power series,
then
R dist(z0 ; @U ) = supfr > 0 : B(z0 ; r) U g:
Hence, if U = C then R = 1.

Corollary 83 Let U C be an open set and let f : U ! C be holomorphic.

Given a closed ball B(z0 ; r) U , let M maxB(z0 ;r) jf j. Then for every n 2 N,

n!M
jf (n) (z0 )j :
rn
Proof. In view of (32),
Z Z
n! jf ( )j n!M 1
jf (n) (z0 )j d d
2 @B(z0 ;r) j z0 jn+1 2 @B(z0 ;r) j z0 jn+1
n!M n!M
= 2 r= n ;
2 rn+1 r
which concludes the proof.

36
De…nition 84 Given an open connected set U C and a holomorphic function
f : U ! C, a point z0 2 @U is called a regular point if there exist r > 0 and a
holomorphic function g : B(z0 ; r) ! C such that f = g on U \ B(z0 ; r). A point
z0 2 @U is called a singular point if it is not a regular point. We say that @U
is the natural boundary of f if every point on @U is a singular point.
Exercise 85 Let f : B(z0 ; r) ! C be holomorphic and assume that the power
series
1
X
f (z) = an (z z0 )n
n=0
has radius of convergence exactly r. Then there is at least one singular point on
@B(z0 ; r).
Exercise 86 Given the function
1
X n
f (z) = z2 ;
n=0

…nd its natural boundary.

Next we discuss some important consequences of Cauchy’s formula.
Corollary 87 (Liouville) Let f : C ! C be holomorphic and bounded. Then
f is constant.
Proof. Let M > 0 be such that jf (z)j M for all z 2 C. By the previous
corollary,
M
jf 0 (z)j
r
for every r > 0. Hence, letting r ! 1 we get f 0 (z) = 0 for all z. We can now
apply Corollary 14.
Theorem 88 (Fundamental theorem of algebra) Every polynomial P : C !
C of degree n 1 has precisely n roots in C.
Proof. Step 1: Write
P (z) = an z n + + a1 z + a0 ;
where an 6= 0. We claim that P has at least one root. Assume by contradiction
that this is not the case, that is, that P (z) 6= 0 for all z 2 C. Then the function
1=P is well-de…ned and holomorphic. Let’s prove that it is bounded. We have
P (z) an 1 a0
n
= an + + + ! an + 0
z z zn
as jzj ! 1. Hence, taking " = 21 jan j > 0 we can …nd R > 0 such that

1 P (z) 3
jan j jan j for all jzj R:
2 zn 2

37
In particular,
1 2 2
for all jzj R:
jP (z)j jan jjzjn jan jRn

Since jP1 j is continuous on the compact set B(0; R), there exists M 0 such
1
that jP (z)j M for all jzj R, which, together with the previous inequality,
proves the claim. It now follows from Liouville’s theorem that P1 is constant,
which is a contradiction since P has degree at least one.
Wednesday, February 5, 2020
Proof. Step 2: In view of the previous step there exists w1 2 C such that
P (w1 ) = 0. Let z = (z w1 ) + w1 . Then

P (z) = an [(z w1 ) + w1 ]n + + a1 [(z w1 ) + w1 ] + a0 :

Using the binomial theorem

Xk
k j k
(a + b)k = a b j

j=0
j

with a = z w1 and b = w1 , we can rewrite P (z) as

P (z) = bn (z w1 )n + + b1 (z w1 ) + b0 ;

where bn = an . Since P (w1 ) = 0, we get that b0 = 0. Hence,

P (z) = (z w1 )[bn (z w1 )n 1
+ + b1 ] =: (z w1 )P1 (z);

where P1 is a polynomial of degree n 1. If n 2, we can apply the previous

step to P1 to …nd a second root w2 .
Inductively, we can …nd w1 ; : : : ; wn 2 C such that

P (z) = an (z w1 ) (z wn ) for all z 2 C:

This concludes the proof.

Another corollary of Cauchy’s theorem is the following.

Corollary 89 (Morera) Let B C be an open ball, let f : B ! C a continu-

ous function such that for every closed triangle T B,
Z
f = 0:
@T

Then f is holomorphic in B.

Proof. In view of Remark 71 we have that f has a primitive F : B ! C.

Hence, F is holomorphic. In turn, by the previous corollary, F is in…nitely
di¤erentiable. Since F 0 = f , it follows that f is also holomorphic.

38
Let’s see how to use Morera’s theorem. Let U C be an open set. De…ne
U + := fz = x + iy 2 U : y > 0g;
U := fz = x + iy 2 U : y < 0g;
S := fz = x + i0 2 U g;
+
so that U = U [ U [ S.
Theorem 90 Let U C be an open set, let f + : U + [ S ! C be a continuous
function which is holomorphic in U + and let f : U [ S ! C be a continuous
function which is holomorphic in U + . Assume that f + = f in S. Then the
function f : U ! C, de…ned by
f + (z) if z 2 U + [ S;
f (z) :=
f (z) if z 2 U ;
is holomorphic in U .
Proof. We only need to prove di¤erentiability at points in S. Fix z0 2 S
and let B(z0 ; r) U . Since f is continuous, we can use Morera’s theorem to
prove that f is holomorphic in B(z0 ; r). Let T B(z0 ; r) be a closed triangle.
If
R T does not intersect S, then it is contained either in U + or in U and so
@T
f dz = 0 by Exercise 75 since f and f are holomorphic in U + and U ,
+ +

respectively. If T U + and one of its sides lies in S, for " > 0 small consider
the
R triangle T" := T \ fz = x + yi : y "g. Then again by Exercise 75,
+
@T"
f dz = 0. Since f is continuous, letting " ! 0 and using the Lebesgue
dominated convergence R theorem (or Arzelá’s convergence theorem for Riemann’s
integration) we get @T f dz = 0. The case in which T U and one of its
sides lies in S is similar.
If T has a vertex in S and is contained in U + (or U ) we either raise (lower)
T so that it is contained in U + (U ) and reason as above.
If the interior of T intersects S, we split T using S into three triangles whose
interior is contained in U + or U and which have one side or a vertex R in S. We
then apply the previous cases and Exercise 66 to conclude that @T f dz = 0.
Hence, the hypotheses of Morera’s theorem are satis…ed and so f is holomorphic
in B(z0 ; r).
We are ready to prove Schwarz’s re‡ection principle
Theorem 91 (Schwarz re‡ection principle) Let U C be an open set which
is symmetric with respect to the real line, that is,
z2U if and only if z 2 U:
and let f : U [ S ! C be a continuous function which is holomorphic in U +
+ +

and real-valued on S. Then the function f : U ! C, de…ned by

f + (z) if z 2 U + [ S;
f (z) :=
f + (z) if z 2 U ;
is holomorphic in U .

39
Proof. Given z0 2 U , we have that z0 2 U + . By Corollary 81 we can
write
X1
+
f (w) = an (w z0 )n
n=0
+
for all w 2 B(z0 ; r) U and for some r > 0. By symmetry B(z0 ; r) U and
for every z 2 B(z0 ; r) we have that z 2 B(z0 ; r) and so
1
X
f + (z) = an (z z0 )n :
n=0

Taking the conjugate in the partial sums and then passing to the limit we have
that
1
X 1
X
f (z) = f + (z) = an (z z0 )n = an (z z0 )n :
n=0 n=0
P1 P1
Since the radius of convergence of n=0 an (z z0 )n is the same as n=0 an n
,
we conclude that f is holomorphic in B(z0 ; r).
To conclude observe that since f + is real-valued on S,
f + (x) = f + (x)
for all x 2 S. Hence, f is continuous at points of S. Thus, by the previous
theorem we conclude that f is holomorphic in U .

7 Cauchy’s Theorem, General Case

In this section we extend Corollary 72 to simply connected domains.
In what follows, given the unit square Q = [0; 1] [0; 1], we consider the
oriented closed simple curve obtained by moving along @Q counterclockwise
starting from (0; 0). Denote by '0 : [0; 4] ! @Q the parametric representation
obtained by using arclength.
Theorem 92 Let U C be an open set, let h : Q ! U be Lipschitz continuous,
let be the Lipschitz continuous oriented closed curve parametrized by h '0 :
[0; 4] ! U , and let f : U ! C be holomorphic. Then
Z
f dz = 0:

Friday, February 7, 2020

Proof. Assume by contradiction that
Z
f dz = c 6= 0:

By replacing f with f =c, without loss of generality, we may assume that c = 1.

Divide Q into four squares Q1;1 , Q1;2 , Q1;3 , Q1;4 of side-length 12 and para-
metrize their boundaries as we did for @Q. Let '1;1 , '1;2 , '1;3 , '1;4 be the

40
corresponding parametric representations and let 1;1 , 1;2 , 1;3 , 1;4 be the
oriented closed curve parametrized by h '1;k : [0; 4=21 ] ! U , k = 1; : : : ; 4,
respectively. Using Exercise 66 we have that
Z Z Z Z
1= f dz + f dz + f dz + f dz
1;1 1;2 1;3 1;4

and thus there exists k1 2 f1; : : : ; 4g such that

Z
1
f dz :
1;k
4

Let Q1 := Q1;k1 and 1 := 1;k1 . We now divide Q1 into four squares Q2;1 , Q2;2 ,
1
Q2;3 , Q2;4 of side-length 16 . Proceeding as before we …nd k2 2 f1; : : : ; 4g such
that Z
1
f dz :
2;k2
16
Inductively we obtain a decreasing sequence of closed squares Qn of side-length
1
2n such that Z
1
f dz n
: (33)
n
4
where n is the oriented closed curve parametrized by h 'n : [0; 24n ] ! U and
'n : [0; 24n ] ! @Qn . By Cantor’s theorem there exists (x0 ; y0 ) 2 Qn for all n.
Let z0 = h((x0 ; y0 )). Since f is di¤erentiable, we can write

f (z) = f (z0 ) + f 0 (z0 )(z z0 ) + R(z);

where
R(z)
lim = 0: (34)
z!z0 z z0
Since a constant function and a linear function az have a primitive, by the
fundamental theorem of calculus,
Z Z
f dz = R dz:
n n

Let n be the range of n . If z 2 n = h('n ([0; 24n ])), we can …nd (x; y) 2 @Qn
such that z = h(x; y). Hence, if L > 0 is the Lipschitz constant of h, we have
that
p
p 2
2
jz z0 j = jh(x; y) h(x0 ; y0 )j L (x x0 ) + (y y0 ) 2 L diam Qn = L n :
2
In turn, by (34), p
2
jR(z)j "n jz z0 j "n L ;
2n

41
where "n ! 0+ . Hence,
Z Z p
2
f dz = R dz "n L L( n)
n n
2n
4L2
"n ;
4n
where we used the fact that
Z 24n Z 4 Z 4
2n 2n 4L
L ( n) = j(h 'n )0 (s)j ds L j'0n (s)j ds = L 1 ds =
0 0 0 2n
Using (34) we get Z
1 4L2
f dz "n
4n n
4n
as n ! 1, which is a contradiction.
Next we consider the case in which h is only continuous.
Theorem 93 Let U C be an open set, let h : Q ! U be continuous, let be
the oriented closed curve parametrized by h '0 : [0; 4] ! U , and let f : U ! C
be holomorphic. If is recti…able, then
Z
f dz = 0:

Proof. Since Q is compact and h is continuous, h(Q) is compact. Hence,

d := dist(h(Q); @U ) > 0. For every n consider a partition t0 = 0 < t1 < <
1
tn = 1 with tk tk 1 n for every k = 1; : : : ; n (for example n = n and tk =
k=n, k = 0; : : : ; n). We construct hn : Q ! U by de…ning hn (tj ; tk ) := h(tj ; tk )
for each j; k = 0; : : : ; n and by interpolating linearly in each subrectangle
hn (rtj + (1 r)tj 1 ; stk + (1 s)tk 1) := (1 r)(1 s)hn (tj 1 ; tk 1 )
+ r(1 s)hn (tj ; tk 1) + (1 r)shn (tj 1 ; tk ) + rshn (tj ; tk )
for r; s 2 [0; 1]. Then hn : Q ! C is Lipschitz continuous. Using the uniform
continuity of h we have that hn ! h uniformly in Q as n ! 1. In particular,
dist(hn (Q); @U ) d=2 for all n su¢ ciently large. Hence, hn : Q ! U for n
large. By the previous theorem
Z
f dz = 0:
n

Since n is parametrized by hn '0 : [0; 4] ! U we have that hn '0 ! h '0

uniformly, and since f is continuous and h '0 has …nite length, it follows that
(Exercise, see the proof of Lemma 64)
Z Z
0 = lim f dz = f dz;
n!1
n

which concludes the proof.

42
Corollary 94 Let U C be an open set, let h : Q ! U be continuous and
such that h(s; 0) = h(s; 1) for all s 2 [0; 1], let be the oriented closed curve
parametrized by h '0 : [0; 4] ! U , and let f : U ! C be holomorphic. Assume
that the curves 1 and 2 parametrized by h '0 : [1; 2] ! U and h '0 : [3; 4] !
U are recti…able, then Z Z
f dz + f dz = 0:
1 2

Proof. Since h(s; 0) = h(s; 1) for all s 2 [0; 1], in the previous proof we
will have hn (s; 0) = hn (s; 1) for all s 2 [0; 1]. Hence, the Lipschitz curves
parametrized by h '0 : [0; 1] ! U and h '0 : [2; 3] ! U are one the opposite
of the other and so their corresponding integrals will cancel each other. In turn,
Z Z
f dz + f dz = 0:
1;n 2 ;n

Letting n ! 1 will give the desired result.

De…nition 95 Given a set E C, two continuous oriented closed curves 1

and 2 with range in E and parametric representations '1 : [a; b] ! C and
'2 : [a; b] ! C, respectively, are homotopic in E if there exists a continuous
function h : [0; 1] [a; b] ! C such that h ([0; 1] [a; b]) E,

h (0; t) = '1 (t) for all t 2 [a; b] ; h (1; t) = '2 (t) for all t 2 [a; b] ;
h (s; a) = h (s; b) for all s 2 [0; 1] :

The function h is called a homotopy in E between the two curves.

Roughly speaking, two curves are homotopic in E if it is possible to deform

the …rst continuously until it becomes the second without leaving the set E.

De…nition 96 A set E C is simply connected if it is pathwise connected and

if every continuous closed curve with range in E is homotopic in E to a point
in E (that is, to a curve with parametric representation a constant function).

Example 97 A star-shaped set is simply connected. Indeed, let E C be star-

shaped with respect to some point z0 2 E and consider a continuous closed curve
with parametric representation ' : [a; b] ! C such that ' ([a; b]) E. Then
the function
h (s; t) := s'(t) + (1 s) z0
is an homotopy between and the point z0 .

43
Monday, February 10, 2020

Theorem 98 (Cauchy) Let U C be an open set, let 1 and 2 be two ori-

ented closed recti…able curves which are homotopic in U and let f : U ! C be
holomorphic. Then Z Z
f dz = f dz:
1 2

In particular, if U is simply connected, then

Z
f dz = 0

for every recti…able closed oriented curve with range in U .

Proof. Let '1 : [0; 1] ! U and '2 : [0; 1] ! U be parametric representations

of 1 and 2 , respectively, and let h : [0; 1] [0; 1] be a corresponding homotopy.
Then h '0 is composed of four curves: …rst s 2 [0; 1] ! h(s; 0) followed by 1 ,
then the opposite of s 2 [0; 1] ! h(s; 1) and …nally the opposite of 2 . Since
the …rst and the third of these four curves are the opposite to each other, the
corresponding integrals will cancel out. Hence, in view of Corollary 94,
Z Z
f dz + f dz = 0:
1 2

The result now follows from 66.

Exercise 99 Let U C be a simply connected open set and let f : U ! C be

holomorphic. Prove that f has a primitive in U .

Using the previous exercise we can show that in a simply connected open
set which does not contain the origin there is a branch of the logarithm. More
generally, we have the following important result.

Corollary 100 Let U C be a simply connected open set and let f : U ! C

be a holomorphic function such that f (z) 6= 0 for all z 2 U . Then there exists a
holomorphic function g : U ! C such that

f (z) = eg(z) for all z 2 U:

If z0 2 U and f (z0 ) = ew0 for w0 2 C, then we can choose g in such a way that
g(z0 ) = w0 .

Proof. Fix z0 2 U and use polar coordinates to write f (z0 ) = rei . Taking
w0 := log r + i , we have that f (z0 ) = ew0 . Since f (z) 6= 0 for all z 2 U , the
function f 0 =f is well-de…ned and holomorphic in U . By the previous exercise,
f 0 =f has a primitive F1 , that is, F10 = f 0 =f in U . By adding a constant, we can
assume that F1 (z0 ) = w0 . Then h(z) := eF1 (z) is holomorphic in U and never

44
vanishes (since the exponential never does). In turn, f =h is holomorphic. Let’s
compute its derivative
0
f f 0 (z)h(z) f (z)h0 (z) f 0 (z)eF1 (z) f (z)F10 (z)eF1 (z)
(z) = 2
=
h h (z) e2F1 (z)
0
f 0 (z)eF1 (z) f (z) ff (z)
(z) F1 (z)
e
= = 0:
e2F1 (z)
Since U is connected, it follows from Corollary 14 that f =h is a constant function.
Hence, there is c 2 C n f0g such that

f (z) = ch(z) = ceF1 (z) :

Taking z = z0 we get

ew0 = f (z0 ) = ceF1 (z0 ) = cew0

and so c = 1. This completes the proof.

Exercise 101 Let U C be a simply connected open set with 0 2 = U . Prove

that in U there exists a branch logU of the logarithm. Prove also that if 1 2 U ,
then we can assume that logU r = log r whenever r is a real number su¢ ciently
close to 1.

Exercise 102 Prove that the previous exercise continues to hold if in place of
U simply connected we assume that
Z
f ds = 0

for every holomorphic function f : U ! C and for every closed oriented Lip-
schitz continuous curve with range contained in U .

Remark 103 In view of Exercise 101, if U C is a simply connected open set

= U and a 2 C, then in U there is a branch of z a , de…ned as usual by
with 0 2

z a := ea logU z :

De…nition 104 Given a set E C, two continuous oriented curves, with para-
metric representations ' : [a; b] ! C and : [a; b] ! C such that ' ([a; b]) E,
([a; b]) E, '(a) = (a) = , '(b) = (b) = are …xed-endpoint homo-
topic in E if there exists a continuous function h : [0; 1] [a; b] ! C such that
h ([0; 1] [a; b]) E,

h (0; t) = '(t) for all t 2 [a; b] ; h (1; t) = (t) for all t 2 [a; b] ;
h (s; a) = ; h (s; b) = for all s 2 [0; 1] :

The function h is called a …xed-endpoint homotopy in E between the two curves.

45
Exercise 105 Let U C be an open set, let 1 and 1 be two oriented recti-
…able continuous curves with the same endpoints and which are …xed-endpoint
homotopic in U , and let f : U ! C be holomorphic. Prove that
Z Z
f dz = f dz:
1 2

8 Harmonic Functions
Given an open set RN , a function u : ! R of class C 2 is called harmonic
in if it satis…es
u(x) = 0 for all x 2 ;
where we recall that is the Laplace operator de…ned by
N
X @2
:= :
@x2k
k=1

As a consequence of Cauchy’s integral formula we have the following important

result.
Theorem 106 Let U C be an open set, let f : U ! C be a holomorphic
function. Then the real-valued functions
u(x; y) := Re f (x + iy); v(x; y) := Im f (x + iy)
are harmonic in := f(x; y) 2 R2 : x + iy 2 U g.
Proof. In what follows given a function g : U ! C we de…ne Rg : !R
and Ig : ! R via
Rg (x; y) = Re g(x + iy); Ig (x; y) := Im g(x + iy):
Recall that by (9),
@u @v
(x; y) = (x; y) = Re f 0 (x + iy);
@x @y
@u @v
(x; y) = (x; y) = Im f 0 (x + iy):
@y @x
This shows that Rf 0 = @u @x , If =
0
@u
@y . By Corollary 81, the function f is
analytic. In particular, it is of class C 1 (U ). In particular, f 0 is holomorphic,
and so we can apply Theorem 13 to f 0 to conclude that Rf 0 = @u @x , If =
0
@u
@y
are di¤erentiable, with
@ @u @ @u
(x; y) = (x; y) = Re f 00 (x + iy); (35)
@x @x @y @y
@ @u @ @u
(x; y) = (x; y) = Im f 00 (x + iy):
@y @x @x @y

46
This implies that all second order partial derivatives of u exist and since f 00 is
continuous, so are they. Thus, u 2 C 2 ( ). Moreover, from the …rst equation in
(35) we get that u is harmonic.
@v @v
We can repeat a similar argument for v since Rf 0 = @y , If 0 = @x or use the
2
Cauchy-Riemann equations, to obtain that v 2 C ( ) and is harmonic.
We also have the converse of this theorem.

Theorem 107 Let R2 be an open set and let u; v : ! R be two harmonic

functions satisfying the Cauchy–Riemann equations
@u @v @u @v
= ; = in : (36)
@x @y @y @x
Then the function f : U ! C de…ned by

f (z) = u(x; y) + iv(x; y); z = x + iy 2 U;

where U := fz = x + iy : (x; y) 2 g, is holomorphic in U .

Proof. This follows from Theorem 15.

An interesting problem is, given an open set R2 and an harmonic
function u : ! R, to …nd another harmonic function v : ! R in such a way
that the Cauchy–Riemann equations hold in . If such a function v exists, it is
called complex conjugate of u.

Exercise 108 Let = R2 nf(0; 0)g. Prove that the function u(x; y) := log(x2 +
y 2 ), (x; y) 2 , is harmonic but does not have a complex conjugate v.

Theorem 109 Let R2 be simply connected and let u : ! R be an

harmonic function. Then u admits a complex conjugate v : ! R.

Proof. De…ne
@u @u
g(z) = (x; y) (x; y)i; z = x + iy 2 U;
@x @y

where as before U := fz = x + iy : (x; y) 2 g. Since u is of class C 2 and

harmonic ,

@ @u @ @u
(x; y) = (x; y) in ;
@x @x @y @y
@ @u @ @u
(x; y) = (x; y) in ;
@y @x @x @y

and so @u
@x and
@u
@y satisfy the Cauchy–Riemann equations. In turn, by the pre-
vious theorem the function g is holomorphic in U . Since is simply connected,
so is U , and so we can apply Exercise 99 to conclude that g has a primitive,
that is, there exists a holomorphic function f : U ! C such that f 0 = g.

47
Let
u1 (x; y) := Re f (x + iy); v(x; y) := Im f (x + iy):
By (9),
@u1 @v @u
(x; y) = (x; y) = Re f 0 (x + iy) = Re g(x + iy) = (x; y);
@x @y @x
@u1 @v @u
(x; y) = (x; y) = Im f 0 (x + iy) = Im g(x + iy) = (x; y)
@y @x @y
and so
@u1 @u @u1 @u
(x; y) = (x; y); (x; y) = (x; y):
@x @x @y @y
Since U is connected, this implies that u u1 must be constant. Since v is a
complex conjugate of u1 , it follows that it is also a complex conjugate to u, and
the proof is complete.
Wednesday, February 12, 2020
As a corollary of Cauchy’s integral form we obtain the mean value theorem.
Theorem 110 (Mean value theorem) Let R2 be an open set and let
u : ! R be an harmonic function. Then for every closed ball B((x0 ; y0 ); r)
we have Z 2
u(x0 ; y0 ) = u(x0 + r cos ; y0 + r sin ) d :
0

Proof. Let z0 = x0 + iy0 . By applying the previous theorem in a larger

open ball B containing z0 we can …nd a function v which is conjugate to u in
B. In turn, the function
f (z) = u(x; y) + iv(x; y); z = x + iy 2 B;
is holomorphic and so by Cauchy’s formula,
Z
1 f (z)
f (z0 ) = dz:
2 i @B(z0 ;r) z z0

Taking as parametric representation of @B(z0 ; r) the function '( ) = z0 + rei ,

2 [0; 2 ], we get
Z 2 Z 2
1 f (z0 + rei ) i 1
f (z0 ) = rie d = f (z0 + rei ) d
2 i 0 rei 2 0
where we used the fact that '0 ( ) = riei . In particular, taking the real part on
both sides
Z 2 Z 2
1
Re f (z0 ) = Re f (z0 + rei ) d = (Re f )(z0 + rei ) d ; (37)
2 0 0

which gives the result.

Using this formula, one can show as in Corollary 81 that u is analytic in .
We leave this as an exercise.

48
9 Zeros and Isolated Singularities
In this section we study zeros and isolated singularities of holomorphic functions.
We begin by showing that zeros of holomorphic functions are isolated.

Theorem 111 Let U C be an open connected set and let f : U ! C be

holomorphic. Assume that there exists a sequence fzk gk in U with zk 6= zm for
k 6= m such that zk ! z0 2 U as n ! 1 and f (zk ) = 0 for all k. Then f = 0.

Proof. Since f is analytic by Corollary 81, there exists r > 0 such that
B(z0 ; r) U such that
1
X
f (z) = an (z z0 )n
n=0

for all z 2 B(z0 ; r). If f 6= 0 in B(z0 ; r), at least one of an must be di¤erent
from 0. Let m be the …rst integer such that am 6= 0. Let m 2 N be the smallest
integer such that am 6= 0. Then as in the proof of Theorem 111 we can write
1
X 1
X
f (z) = an (z z0 )n = (z z0 )m an (z z0 )n m

n=m n=m
1
X
= (z z0 )m ak+m (z z0 )k
k=0
=: (z z0 )m g(z):

Now
1
X
g(z) = am + ak+m (z z0 )k ;
k=1

where the power series is convergent. Hence g(z) ! am 6= 0 as z ! z0 . Hence,

taking " = 21 jam j, there exists 0 < < r such that

1
jg(z) am j jam j
2
1
for all z with jz z0 j , and so jg(z)j jam j jg(z) am j 2 jam j, and in
turn,
1
jam jjz z0 jm
jf (z)j
2
for all z with jz z0 j . Since zk ! z0 we have that jzk z0 j for all k large.
In particular, there are in…nitely many zk such that zk 6= z0 and jzk z0 j .
But
1
0 = jf (zk )j jam jjzk z0 jm > 0;
2
which is a contradiction. This shows that f = 0 in B(z0 ; r).
Let
V := fz 2 U : f (z) = 0g :

49
The set V is open by de…nition and B(z0 ; r) V . The set V is also closed in
U , since if wk 2 V and wk ! w0 2 U , then either wk = w0 for some k and so
w0 2 V or wk 6= w0 for all k, in which case the sequence must have in…nitely
many distinct elements. Hence, by the previous argument we can …nd a ball
centered at w0 where f is zero. This shows that w0 2 V . Hence, V is closed in
U . Hence, U = V [ (U n V ), with U n V open. Since U is connected, it follows
that U n V must be empty.
Observe that in the previous proof we actually showed that each zero of a
holomorphic function f is isolated and has …nite multiplicity, unless f = 0.

Corollary 112 Let U C be an open connected set and let f : U ! C be

holomorphic and not identically zero. Assume that there exists z0 2 U such that
f (z0 ) = 0. Then there exists m 2 N such that

f (z) = (z z0 )m g(z);

where g : U ! C is holomorphic and g(z0 ) 6= 0. Moreover, there exists r > 0

such that f (z) 6= 0 for all z 2 B(z0 ; r) n fz0 g U

Proof. Writing f as a power series centered at z0 ,

1
X
f (z) = an (z z0 )n ;
n=0

If an = 0 for all n 2 N0 , then f = 0 by Theorem 111. Let m 2 N be the smallest

integer such that am 6= 0. Then as in the proof of Theorem 111 we can write
1
X 1
X
f (z) = an (z z0 )n = (z z0 )m an (z z0 )n m

n=m n=m
1
X
= (z z0 )m ak+m (z z0 )k
k=0
=: (z z0 )m g(z):

Then g(z0 ) = am + 0 + + 0 = am 6= 0. The function g is holomorphic in

B(z0 ; R), where R is its radius of convergence. On the other hand, in U nB(z0 ; R)
the function
f (z)
g(z) :=
(z z0 )m
is holomorphic, since quotient of two holomorphic functions.
The last statement follows from Theorem 111.
The number m is called multiplicity of z0 . We say that f has a zero of order
m or of multiplicity m.

Example 113 Consider the function

1+z
f (z) = cos ; z 2 B(0; 1):
1 z

50
1+z
The function f is holomorphic and has in…nitely many zeros when 1 z = 2 +n ,
that is, 1 + z = ( 2 + n )(1 z), or
1+ 2 +n
z= !1
1+ 2 +n
as n ! 1. Note that 1 2 @B(0; 1), and so this does not contradict Theorem
111.
Corollary 114 Let U C be an open connected set and let f : U ! C be
holomorphic. Assume that there exists z0 2 U such that f (n) (z0 ) = 0 for all
n 2 N0 . Then f = 0.
Proof. Writing f as a power series centered at z0 we get that f = 0 in
B(z0 ; r) U . But then we can apply the previous theorem to conclude that
f = 0 in U .
Corollary 115 Let U C be an open connected set and let f; g : U ! C be
holomorphic. Assume that there exists a sequence fzk gk in U with zk 6= zm for
k 6= m such that zk ! z0 2 U as n ! 1 and f (zk ) = g(zk ) for all k. Then
f = g in U .
Next we study isolated singularities.
De…nition 116 Let U C be an open set and let f : U ! C be a holomorphic
function. We say that z0 2 C n U is a point singularity or isolated singularity
of f if there exists r > 0 such that B(z0 ; r) n fz0 g U .
Example 117 If we take U = C n f0g then the holomorphic function f (z) = z
has an isolated singularity at 0. In this case we can extend f to 0 as a holomor-
phic function by setting f (0) := 0. This is called a removable singularity. The
functions f (z) = z1 and g(z) = e1=z have an isolated singularity at z = 0.
We will show that isolated singularities are of three types;
1. removable singularities;
2. poles;
3. essential singularities
De…nition 118 Let U C be an open set, let f : U ! C be a holomorphic
function, and let z0 2 C n U be an isolated singularity of f . We say that z0 is a
removable singularity if we can de…ne f at z0 in such a way that the resulting
function is homomorphic in U [ fz0 g.
Theorem 119 Let U C be an open set, let f : U ! C be a holomorphic
function, and let z0 2 C n U be an isolated singularity of f . Then z0 is a
removable singularity if and only if
lim (z z0 )f (z) = 0: (38)
z!z0

In particular, if f is bounded near z0 , then z0 is a removable singularity.

51
Friday, February 14, 2020
Proof. If z0 is a removable singularity for f then f is continuous at z0 and
so
lim (z z0 )f (z) = 0f (z0 ) = 0:
z!z0

Conversely, assume that (38) holds. De…ne g : U [ fz0 g ! C via

(z z0 )f (z) if z 6= z0 ;
g(z) :=
0 if z = z0 :

In view of (38), the function g is holomorphic in U and continuous at z0 . In

view of Remark 71, g has a primitive G in B(z0 ; r) U [ fz0 g, and so G is
holomorphic. By Corollary 81, G is analytic. Since G0 = g, we have that g is
holomorphic. Since g(z0 ) = 0, by Corollary 112, there exists m 2 N such that

g(z) = (z z0 )m h(z);

where h : U [ fz0 g ! C is holomorphic and h(z0 ) 6= 0. Set f1 (z) = (z

z0 )m 1 h(z). Then f1 is holomorphic in U [ fz0 g. Since B(z0 ; r) n fz0 g is
connected, it follows that f and f1 must coincide in B(z0 ; r) n fz0 g by Corollary
115. Thus, f1 extends f to z0 as an holomorphic function.

De…nition 120 Let U C be an open set, let f : U ! C be a holomorphic

function, and let z0 2 C n U be an isolated singularity of f . We say that z0 is a
pole if
lim jf (z)j = 1: (39)
z!z0

Theorem 121 Let U C be an open set, let f : U ! C be a holomorphic

function, and let z0 2 C n U be a pole of f . Then there exist m 2 N, r > 0, and
a holomorphic function g : B(z0 ; r) ! C such that B(z0 ; r) U n fz0 g, g(z) 6= 0
for all z 2 B(z0 ; r) and
g(z)
f (z) = for all z 2 B(z0 ; r) n fz0 g:
(z z0 )m
Proof. By the de…nition of limit there exists r > 0 such that B(z0 ; r)
U n fz0 g and jf (z)j 1 for all z 2 B(z0 ; r) n fz0 g. Hence, the function f1 is
well-de…ned and holomorphic in B(z0 ; r) U n fz0 g. Moreover, by (39),
1
lim = 0:
z!z0 f (z)
Thus, if we de…ne
1
f (z) if z 6= z0 ;
h(z) :=
0 if z = z0 :
Then h is holomorphic in B(z0 ; r) by the previous theorem. Since h(z0 ) = 0, by
by Corollary 112, there exists m 2 N such that

h(z) = (z z0 )m q(z);

52
where q : B(z0 ; r) ! C is holomorphic and q(z0 ) 6= 0. By continuity and taking
r smaller, if necessary, we can assume that q(z) 6= 0 for all z 2 B(z0 ; r). Then
1
= (z z0 )m q(z)
f (z)
for all z 2 B(z0 ; r) n fz0 g, that is,
1 g(z)
f (z) = =: ;
(z z0 )m q(z) (z z0 )m
where g(z) := 1=q(z).
The number m is called multiplicity of z0 . We say that f has a pole of order
m or of multiplicity m. When m = 1, we say that f has a simple pole at z0 .
Theorem 122 Let U C be an open set, let f : U ! C be a holomorphic
function, and let z0 2 C n U be a pole of f or order m. Then there exist b1 ,
. . . , bm 2 C, r > 0, and a holomorphic function h : B(z0 ; r) ! C such that
B(z0 ; r) U n fz0 g, and
b1 bm
f (z) = + + + h(z) for all z 2 B(z0 ; r) n fz0 g: (40)
z z0 (z z0 )m
Proof. By the previous theorem, there exist m 2 N, r > 0, and a holomor-
phic function g : B(z0 ; r) ! C such that B(z0 ; r) U n fz0 g, g(z) 6= 0 for all
z 2 B(z0 ; r) and
g(z)
f (z) = for all z 2 B(z0 ; r) n fz0 g:
(z z0 )m
Since g is analytic, by taking r smaller, if necessary, we can write
1
X
g(z) = a0 + a1 (z z0 ) + + am 1 (z z0 )m 1
+ an (z z0 )n
n=m

and so
1
X
g(z) a0 a1 am 1
f (z) = = + + + + an (z z0 )n m
:
(z z0 )m (z z0 )m (z z0 )m 1 (z z0 ) n=m

It su¢ ces to de…ne

1
X
h(z) := an (z z0 )n m
;
n=m
which is holomorphic.
The sum
b1 bm
+ +
z z0 (z z0 )m
is called the principal part of f at the pole z0 and the number b1 is the residue
of f at z0 . We write
resz0 f = b1 :

53
Theorem 123 Let U C be an open set, let f : U ! C be a holomorphic
function, and let z0 2 C n U be a pole of f or order m. Then

1 dm 1
resz0 f = lim ((z z0 )m f (z)):
z!z0 (m 1)! dz m 1

In particular, if f has a simple pole at z0 , then

resz0 f = lim (z z0 )f (z):

z!z0

Proof. By (40),

(z z0 )m f (z) = b1 (z z0 )m 1
+ b2 (z z0 )m 2
+ + bm + (z z0 )m h(z):

Hence
dm 1
dm 1
((z z0 )m f (z)) = b1 (m 1)! + 0 + +0+ ((z z0 )m h(z)):
dz m 1 dz m 1

To conclude observe that

dm 1
lim ((z z0 )m h(z)) = 0
z!z0 dz m 1

since we are di¤erentiating m 1 times and so by the product rule each term
dm 1
in dz m 1 ((z z0 )m h(z)) will have some power of z z0 .
Next we prove the residue formula. We begin with a simple case.

Theorem 124 (Residue formula) Let U C be an open set, let z0 2 U , and

let f : U n fz0 g ! C be a holomorphic function having a pole at z0 . Then for
closed ball B U having z0 in its interior,
Z
f dz = 2 i resz0 f:
@B

Proof. Consider the closed curve ;" given in Figure 1, where " is the
radius of the small circle centered at z0 and is the width of the corridor. Since
the function f is holomorphic in U n fz0 g, by considering V := B n S , where S
is the segment obtained when " ! 0 and ! 0, we can apply Exercise 74, to
obtain that f has a primitive in V . Since the range of ;" is contained in V , it
follows from Corollary 65 that
Z
f dz = 0:
;"

If we let ! 0+ and use the fact that f is continuous, we get that the two seg-
ments converge to a segment which is integrated in opposite directions. Hence,
we obtain Z Z
f dz f dz = 0: (41)
@B @B(z0 ;")

54
Thus to prove the theorem it su¢ ces to show that
Z
f dz = 2 i resz0 f: (42)
@B(z0 ;")

By Theorem 122 there exist b1 , . . . , bm 2 C, r > 0, and a holomorphic function

h : B(z0 ; r) ! C such that B(z0 ; r) U n fz0 g, and

b1 bm
f (z) = + + + h(z) for all z 2 B(z0 ; r) n fz0 g:
z z0 (z z0 )m

Hence,
Z Z Z Z
b1 bm
f dz = dz + + dz + h dz:
@B(z0 ;") @B(z0 ;") z z0 @B(z0 ;") (z z0 )m @B(z0 ;")
(43)
By Cauchy’s integral formula applied to the constant function b1 we have that
Z
1 b1
b1 = dz; (44)
2 i @B(z0 ;") z z0

while by Corollary 81 applied to the constant functions bk ,

Z
dk 1 (k 1)! bk
0 = k 1 (bk ) = dz: (45)
dz 2 i @B(z0 ;") (z z0 )k

Since h is holomorphic in B(z0 ; r), taking " < r we have that

Z
h dz = 0 (46)
@B(z0 ;")

by Corollary 72. Formula (42) follows by combining (43)–(46).

Remark 125 Note that since @B and @B(z0 ; ") are homotopic in U , we could
have used Theorem 98 to obtain (41).

55
Monday, February 17, 2020

Exercise 126 Let U C be an open set, let z1 ; : : : ; zn 2 U , and let f : U n

fz1 ; : : : ; zn g ! C be a holomorphic function having poles at z1 ; : : : ; zn . Prove
that for every closed ball B U having z1 ; : : : ; zn in its interior,
Z Xn
f dz = 2 i reszk f:
@B k=1

Exercise 127 (Residue formula) Let U C be an open set, let z1 ; : : : ; zn 2

U , and let f : U n fz1 ; : : : ; zn g ! C be a holomorphic function having poles
at z1 ; : : : ; zn . Prove that for every continuous recti…able closed simple curve
homotopic to 0 in U and having z1 ; : : : ; zn in its interior,
Z n
X
f dz = 2 i reszk f:
k=1

Note that in the previous exercise we are using Jordan’s curve theorem (see
Theorem 57).
The calculus of residues can be used to compute many interesting improper
integrals.

Example 128 Let’s prove that for 0 < a < 1,

Z
eax
x
dx = :
R 1+e sin( a)
Consider the function
eaz
f (z) = :
1 + ez
Note that 1 + ez = 0 for z = i + 2i k, k 2 Z. Given ` > 0 consider the
rectangle R` = fz = x + iy : x 2 ( `; `); 0 < y < 2 g and let ` be the oriented
closed curve which parametrizes @R` using arclength and going counterclockwise
starting from ` + 0iy. The only point at which the denominator vanishes in
R` is i. Note that
z i z i
(z i)f (z) = eaz z
= eaz z i
:
1+e e e
d z
Since dz e = ez , we have that

ez e i
i
lim =e = 1
z! i z i
and so
lim (z i)f (z) = ea i :
z! i

In turn, by (40),
res i f = ea i :

56
It follows by the residue formula that
Z n
X
f dz = 2 i res i f = 2 iea i : (47)
` k=1

Set Z Z
` `
eax
I` := f (x) dx = dx: (48)
` ` 1 + ex
On the other hand, to parametrize the top we consider curve `;3 parametrized
by '3 (t) = 3` + 2 t + 2 i, where t 2 [2` + 2 ; 4` + 2 ]. Then by the change
of variables s = 3` + 2 t,
Z Z 4`++2 Z `
f dz = f ('3 (t))'03 (t) dt = f (s + 2 i) ds (49)
`;3 2`+2 `
Z ` Z `
eas e2 ia eas e2 ia
= ds = ds = e2 ia
I` :
` 1 + es+2 i ` 1 + es

Next to parametrize the right vertical side we consider curve `;2 parametrized
by '2 (t) = ` + i(t 2`), where t 2 [2`; 2` + 2 ]. Then by the change of variables
s = t 2`,
Z Z 2`++2 Z 2
f dz = f ('(t))'0 (t) dt = if (` + is) ds
`;2 2` 0
Z 2 Z 2
ea(`+is) ea` eais
= `+is
ds = ` ds:
0 1+e e 0 e + eis
`

`
Since je + eis j jeis j e `
=1 e `
, we have
Z Z 2
1 jeais j
f dz ds (50)
`;2
e`(1 a)
0 je ` + eis j
1 2
`
!0
e`(1 a) 1 e
as ` ! 1. A similar computation holds for the left vertical side, whose integral
can be bound in modulus by ce `a . It follows from (47)–(50) that
Z Z
a i 2 ia eax
2 ie = lim f dz = (1 e ) x
dx;
`!1
` R 1+e

that is,
Z
eax 2 iea i 2 iea i 2 i
dx = = = = ;
R 1 + ex 1 e2 ia e2 ia 1 e ia e ia sin( a)

where we used (6).

57
Exercise 129 Use the calculus of residues to prove that
Z
1
2
dx = :
R 1+x

Exercise 130 Use the calculus of residues to prove that for all 2 R,
Z
e 2 ix 1
dx = :
R cosh( x) cosh( )

We now the notion of meromorphic functions. Consider the extended complex

plane C1 obtained by adding to C a point not in C called 1,

C1 := C [ f1g:

Given an open set U C and z0 2 U , if a holomorphic function f : U nfz0 g ! C

has a pole at z0 , we can extend f to z0 by setting

f (z0 ) := 1;

so that f : U ! C1 .

De…nition 131 Let U C and let f : U ! C1 . We say that f is meromorphic

if there exists a sequence fzn gn of complex numbers such that the set fzn :
n 2 Ng has no accumulation points in U , f has poles at zn for every n, and
f : U n fzn : n 2 Ng ! C is holomorphic.

Let U C be an open set which contains C n B(0; R) for some R > 0 and
let f : U ! C be a holomorphic function. We say that f has a removable
singularity, a pole, or an essential singularity at in…nity if the function F (z) :=
f (1=z) has a removable singularity, a pole, or an essential singularity at 0,
respectively, In the …rst case we say that f is holomorphic at in…nity. We say
that f is meromorphic in the extended complex plane if it is meromorphic in
the complex plane and either has a pole at in…nity or is holomorphic at in…nity.

Exercise 132 Prove that a holomorphic function f : C ! C has a removable

singularity at in…nity i¤ it is constant.

Exercise 133 Prove that a holomorphic function f : C ! C has a pole at

in…nity of order m i¤ it is polynomial of degree m.

Exercise 134 Characterize those rational functions which have a removable

singularity at in…nity.

Exercise 135 Characterize those rational functions which have a pole of order
m at in…nity.

58
Next we prove the argument principle. We have seen that in general for a
branch logV of the logarithm, the formula

logV (z1 z2 ) = logV z1 + logV z2 :

Hence, we cannot expect the formula

logV (f1 f2 ) = logV f1 + logV f2

to holds for holomorphic functions f1 ; f2 : U ! V . However, the formula holds

for derivatives since
(f1 f2 )0 f 0 f2 + f1 f20 f0 f0
= 1 = 1 + 2:
f1 f2 f1 f2 f1 f2
More generally, !0
n
Y
fk n
k=1
X f0 k
n = : (51)
Y fk
fk k=1

k=1

Wednesday, February 19, 2020

We will use this observation to prove the argument principle. Given a set
E, we denote by card E its cardinality.

Theorem 136 (Argument principle) Let U C be an open set and let f :

U ! C1 be a meromorphic function. Then for every for closed ball B U
such that f has no poles or zeros on @B, we have
Z
1 f0
dz = (number of zeros of f in B) minus (number of poles of f in B);
2 i @B f
where the zeros and poles are counted with multiplicity.

Proof. Let z1 ; : : : ; zn be the zeros of f inside B and let p1 ; : : : ; p` be the poles

of f inside B. For every k = 1; : : : ; n, let mk be the order of zk . By Corollary
112 we can …nd rk > 0 and a holomorphic function gk : B(zk ; rk ) ! C such
that gk 6= 0 in B(zk ; rk ) B and

f (z) = (z zk )mk gk (z) for all z 2 B(zk ; rk ):

It follows from (51) that

f 0 (z) mk g 0 (z)
= + k :
f (z) z zk gk (z)
g0 f0
The function gkk is holomorphic in B(zk ; rk ). This shows that f has a simple
pole with residue mk at zk , that is, reszk f 0 =f = mk .

59
Similarly, for every k = 1; : : : ; `, let nk be the order of pk . by Theorem 121
we can …nd tk > 0 and a holomorphic function hk : B(pk ; tk ) ! C such that
hk 6= 0 in B(pk ; tk ) B and
hk (z)
f (z) = for all z 2 B(pk ; tk ): (52)
(z pk )nk
Since
d 1 1
= ;
dz z pk (z pk )2
we have
d 1 1
dz z pk (z pk )2 1
1 = 1 = ;
z pk z pk
z pk
and so, using (51) and (52) we get
f 0 (z) nk h0 (z)
= + k :
f (z) z pk hk (z)
h0 0
The function hkk is holomorphic in B(pk ; tk ). This shows that ff has a simple
pole with residue nk at pk , that is, respk f 0 =f = nk .
The conclusion now follows by applying the residue formula (Theorem 124)
to f 0 =f .
Exercise 137 Let U C be an open set and let f : U ! C1 be a meromor-
phic function. Prove that for every continuous recti…able closed simple curve
homotopic to 0 in U and whose range contains no zero or pole of f , we have
Z 0
1 f
dz = (number of zeros of f in the interior of ) minus
2 i f
(number of poles of f in the interior of );
where the zeros and poles are counted with multiplicity.
Next we discuss the last type of isolated singularities.
De…nition 138 Let U C be an open set, let f : U ! C be a holomorphic
function, and let z0 2 C n U be an isolated singularity of f . We say that z0 is
an essential singularity for f if z0 is not a removable singularity or a pole.
Example 139 The function f (z) = e1=z has an essential singularity at 0. In-
deed, if we take z = iy we have that
jf (iy)j = je1=(iy) j = je i=y
j = 1;
so z is not a pole. On the other hand,
lim xe1=x = 1
x!0+

and so by Theorem 119, z = 0 is not a removable singularity.

60
10 The Maximum Modulus Principle
In this section we prove some important theorems of holomorphic functions.

Theorem 140 (Rouché) Let U C be an open set and let f : U ! C and

g : U ! C be holomorphic functions. Assume that there exists a closed ball
B U such that
jf (z)j > jg(z)j for all z 2 @B: (53)
Then f and f + g have the same number of zeros inside B.

Proof. For t 2 [0; 1] consider the function

ft (z) := f (z) + tg(z); z 2 U:

Then f0 = f and f1 = f + g. Moreover ft is holomorphic in U . Let nt 2 N0 be

the number of zeros of ft inside B counted with multiplicity. The hypothesis
(53) guarantees that ft has no zeros on @B. Hence, by the argument principle
Z
1 ft0
nt = dz:
2 i @B ft

Again by (53) we have that the function

ft0 (z) f 0 (z) + tg 0 (z)

g(t; z) = = ; t 2 [0; 1]; z 2 @B
ft (z) f (z) + tg(z)

is continuous in the compact set [0; 1] @B. Hence, it is bounded. Using the
Lebesgue dominated convergence theorem (or Ascoli’s convergence theorem for
Riemann integrals), we have that nt is a continuous function of t. But since it is
integer-valued and [0; 1] is connected, it follows that nt must be constant. This
concludes the proof.
Using Rouché’s theorem we can prove that non-constant holomorphic func-
tions are open.

Theorem 141 (Open mapping) Let U C be an open set and let f : U ! C

be a non-constant holomorphic function. Then for every V U open, f (V ) is
open.

Proof. Let z0 2 V and let w0 = f (z0 ). We must …nd " > 0 such that
B(w0 ; ") f (V ). Since the zeros of f w0 are isolated by Theorem 111 (or
Corollary 112), there exists > 0 such that B(z0 ; ) V and f w0 6= 0 on
@B(z0 ; ). By uniform continuity, we can …nd " > 0 such that

jf (z) w0 j > " for all z 2 @B(z0 ; ):

Let w 2 B(w0 ; ") and de…ne

g(z) := f (z) w = (f (z) w0 ) + (w0 w0 ) =: F (z) + G(z):

61
By the previous inequality we have that jF (z)j > jG(z)j for all z 2 @B(z0 ; ).
Hence, by Rouché’s theorem F and F + G = g have the same number of zeros in
B(z0 ; ). Since F has one zero in B(z0 ; ), so must g. Hence, there is z 2 B(z0 ; )
such that f (z) = w. This shows that B(w0 ; ") f (B(z0 ; )) f (V ). This
concludes the proof.

Corollary 142 Let U C be open and let f : U ! C be injective and holo-

morphic. Then f 1 : f (U ) ! C is holomorphic and

1 0 1
(f ) (w) = ; w 2 f (U ):
f 0 (f 1 (w))

Proof. By the open mapping theorem, f 1 is continuous and f (U ) is open.

Hence, we can apply Exercise 9 to conclude that f 1 is di¤erentiable.

Theorem 143 (Maximum modulus principle) Let U C be an open con-

nected set and let f : U ! C be a nonconstant holomorphic function. Then jf j
cannot attains a maximum in U .

Proof. Assume that jf j assumes a maximum at some point z0 2 U . Let

B(z0 ; r) U . By the open mapping theorem, f (B(z0 ; r)) is open and so there
exists B(f (z0 ); ) f (B(z0 ; r)). This implies that there exists points in U with
modulus bigger that jf (z0 )j, which is a contradiction.

Exercise 144 Let U C be an open connected set and let f : U ! C be a

non-constant holomorphic function such that f (z) 6= 0 for all z 2 U . Prove that
jf j cannot attain its minimum on U .

Corollary 145 Let U C be an open bounded set and let f : U ! C be a

continuous function which is holomorphic in U . Then

sup jf j max jf j:
U @U

Proof. Since jf j is continuous on the compact set U , it admits a maximum.

By the maximum principle, this maximum must be attained at the boundary
of U .
The previous corollary fails in general in unbounded domains.

Example 146 Let U := fz = x + iy : x > 0; y > 0g be the …rst quadrant

2
and let f (z) = e iz . Then f is holomorphic in U and continuous on U . If
2
z = x 0, then jf (x)j = je ix j = 1, while if z = iy with y 0, then jf (iy)j =
2 p
jeiy j = 1. However, f is unbounded. To see this take z = r i = rei =4 . Then
f (z) = er ! 1 as r ! 1.

Friday, February 21, 2020

62
11 Essential Singularities
Next we study the behavior of a holomorphic function near an essential singu-
larity.

Theorem 147 (Casorati–Weierstrass) Let z0 2 C, r > 0, and let f : B(z0 ; r)n

fz0 g ! C be a holomorphic function having an essential singularity at z0 . Then
f (B(z0 ; r) n fz0 g) is dense in C.

Proof. Assume by contradiction that f (B(z0 ; r) n fz0 g) is not dense in C.

Then there exist w0 2 C and > 0 such that

jf (z) w0 j for all z 2 B(z0 ; r) n fz0 g:

It follows that the function

1
g(z) := ; z 2 B(z0 ; r) n fz0 g;
f (z) w0

is well-de…ned and holomorphic. Moreover, it is bounded by 1= . Hence, by

Theorem 119 it has a removable singularity at z0 . Extend g to z0 as a holomor-
phic function. There are now two cases. If g(z0 ) 6= 0, then g 6= 0 in B(z0 ; r),
and so f w0 has a removable singularity at z0 , which is a contradiction. If
g(z0 ) = 0, then
1
lim = 0;
z!z0 f (z) w0
which implies that
lim jf (z) w0 j = 1;
z!z0

and so f has a pole at z0 , which is again a contradiction. This concludes the

proof.
There is actually a much stronger result.

Theorem 148 (Picard Big Theorem) Let z0 2 C, r > 0, and let f : B(z0 ; r)n
fz0 g ! C be a holomorphic function having an essential singularity at z0 . Then
f takes all possible values of C with at most a single exception.

Exercise 149 Prove that

X̀ 1
1 1 X 2z
cot( z) = lim = + :
`!1 z+k z n=1 z 2 n2
k= `

The proof relies on several preliminary results. We begin with another im-
portant theorem.

Theorem 150 (Bloch) Let U C be an open set which contains B(0; 1) and
f : U ! C be a holomorphicp function such that f 0 (0) = 1. Then f (B(0; 1))
contains a ball of radius 23 2.

63
We begin with some lemmas.
Exercise 151 Let V C be an open bounded set, let f : V ! C be a continuous
function such that f : V ! C is open. Let w0 2 V be such that
R := min jf (z) f (w0 )j > 0:
z2@V

Prove that f (V ) contains B(f (w0 ); R).

Lemma 152 Let U C be an open set which contains B(z0 ; r) and f : U ! C
be a holomorphic function which is non-constant in B(z0 ; r) and such that
jf 0 (z)j 2jf 0 (z0 )j
for all z 2 B(z0 ; r): (54)
p
Then f (B(z0 ; r)) contains B(f (z0 ); r0 ), where r0 = (3 2 2)jf 0 (z0 )jr.
Proof. Without loss of generality we may assume that z0 = 0 and f (0) = 0.
De…ne g(z) = f (z) f 0 (0)z. By the fundamental theorem of calculus,
Z
g(z) = [f 0 ( ) f 0 (0)] d :
[0;z]

Consider the parametric representation '(t) = tz, t 2 [0; 1]. Then

Z 1
jg(z)j jzj jf 0 (tz) f 0 (0)j dt: (55)
0

Let w 2 B(0; r). By Cauchy’s formula applied to the holomorphic function f 0 ,

Z Z
0 1 f 0( ) 0 1 f 0( )
f (w) = d ; f (0) = d :
2 i @B(0;r) w 2 i @B(0;r)
Subtracting these identities gives
Z
1 1 1 0
f 0 (w) f 0 (0) = f ( )d
2 i @B(0;r) w
Z
1 w
= f 0( ) d
2 i @B(0;r) ( w)
and so using the the parametric representation ( ) = rei and the fact that
j wj j j jwj = r jwj, we get
1
jf 0 (w) f 0 (0)j jwj sup jf 0 j :
@B(0;r) r jwj
Taking w = tz and using this inequality in (55) gives
Z 1 Z 1
tjzj
jg(z)j jzj jf 0 (tz) f 0 (0)j dt jzj sup jf 0 j dt
0 @B(0;r) 0 r tjzj
Z 1 2
1 1 jzj
jzj2 sup jf 0 j t dt = sup jf 0 j (56)
@B(0;r) r jzj 0 2 r jzj @B(0;r)
jzj2
jf 0 (0)j;
r jzj

64
where in the last inequality we used (54). Now let 0 < < r and take z with
jzj = . Then
jg(z)j = jf (z) f 0 (0)zj jf 0 (0)j jf (z)j
Combining this inequality with (56) gives
2
jf 0 (0)j jf 0 (0)j jf (z)j;
r
or, equivalently,
2
jf (z)j jf 0 (0)j =: jf 0 (0)jh( ):
r

We have
2
d r2 4r + 2 2
h0 ( ) = = 2 0
d r (r )
p p p
2 2 2
for r 2 + 1 and r 1 2 , so h has a maximum at 0 =r 1 2 .
Hence,
p
jf (z)j jf 0 (0)jh( 0 ) = jf 0 (0)jr 3 2 2 for all z 2 @B(0; 0 ):

We now apply the previous exercise with w0 = 0 and V = B(0; 0 ) to obtain

that
f (B(0; r)) f (B(0; 0 )) B(0; R);
p
where R := min@B(0; 0 ) jf j jf 0 (0)jr 3 2 2 = r0 . This concludes the proof.

We now turn to the proof of Bloch’s theorem.

Proof. Step 1: Let U C be an open set which contains B(0; 1) and
f : U ! C be a holomorphic function which is non-constant in B(0; 1). Since
the function
g(z) = jf 0 (z)j(1 jzj)
is continuous in B(0; 1), it assumes a maximumpat some point z0 . We claim
that f (B(0; 1)) B(f (z0 ); r0 ), where r0 := ( 23 2)g(z0 ).
To see this, take t = 21 (1 jz0 j). Then

g(z0 ) = jf 0 (z0 )j(1 jz0 j) = 2tjf 0 (z0 )j: (57)

Moreover, B(z0 ; t) B(0; 1), since if z 2 B(z0 ; t), then

1 1 1
jzj jz z0 j + jz0 j < t + jz0 j = (1 jz0 j) + jz0 j = + jz0 j 1:
2 2 2
Note that the previous inequality also implies that

1 jzj t: (58)

65
Indeed, the previous inequality can be written 1 t + jzj = 21 (1 jz0 j) + jzj,or,
equivalently, 12 + 12 jz0 j jzj, which is what we just proved.
Using (57) and (58) and the fact that g has a maximum at z0 , we have

jf 0 (z)j(1 jzj) = g(z) g(z0 ) = 2tjf 0 (z0 )j (1 jzj)jf 0 (z0 )j;

which gives jf 0 (z)j jf 0 (z0 )j. It now follows from the previous lemma and the
fact that B(z0 ; t) B(0; 1), that

f (B(0; 1)) f (B(z0 ; t)) B(f (z0 ); r0 );

p p
where r0 = (3 2 2)jf 0 (z0 )jt = ( 32 2)g(z0 ), again by (57).
Step 2: To conclude the proof of p the theorem, observe that if f 0 (0) = 1,
3
then g(0) = 1 g(z0 ) and so r0 2 2.

66
Monday, February 24, 2020
Corollary 153 Let U C be an open set and let f : U ! C be a holomorphic
function. If z0 2 U is such that f 0 (z0 ) 6= 0, then f (U ) contains balls of every
1
radius 12 rjf 0 (z0 )j, where 0 < r < dist(z0 ; @U ).

Proof. Assume that z0 = 0. If 0 < r < dist(0; @U ), then B(0; r) U.

Consider the function
f (rz) 1
g(z) := 0 ; z 2 U:
rf (0) r
Since B(0; 1) 1r U and g 0 (0) = 1, by Bloch’s theorem g(B(0; 1)) contains a ball
p
of radius 32 1
2 > 12 1
. In turn, f (B(0; r)) contains a ball of radius 12 rjf 0 (0)j.

Corollary 154 Let f : C ! C be a non-constant entire function. Then f (C)

contains balls of every radius.
Exercise 155 Let f : C ! C. Prove that f f : C ! C has a …xed point unless
f is of the form f (z) = z + w for all z 2 C and for some w 2 C.
In this subsection we prove the following theorem.
Theorem 156 (Picard Little Theorem) Every non-constant entire function
f : C ! C takes every value except at most one.
We begin with some preliminary results.
Lemma 157 Let U C be a simply connected open set and let f : U ! C be
a holomorphic function which does not take value 1 and 1. Then there exists
a holomorphic function h : U ! C such that
f (z) = cos h(z); z 2 U:
Proof. Since f does not take values 1 andp 1, 1 f 2 is never equal to 0 and
so by by Remark 103 there exists a branch of 1 f 2 , that is a holomorphic
function g : U ! C such that g 2 = 1 f 2 in U . Write 1 = f 2 + g 2 = (f + ig)(f
ig). Then f +ig has no zeros in U and so by Corollary 100, f +ig = eih for some
holomorphic function h : U ! C. In turn, 1 = (f + ig)(f ig) = eih (f ig)
and so f ig = e ih . Using Euler’s formula (20) we get
eih + e ih
f= = cos h in U;
2
which concludes the proof.
Lemma 158 Let U C be a simply connected open set and let f : U ! C be
a holomorphic function which does not take value 0 and 1. Then there exists a
holomorphic function g : U ! C such that
1
f (z) = [1 + cos( cos( g(z)))]; z 2 U:
2
Moreover, g(U ) does not contain any ball of radius 1.

67
Proof. The function 2f 1 does not take the values 1 and 1, and so by
the previous lemma, there exists a holomorphic function h : U ! C such that
2f 1 = cos( h) in U . Note that by periodicity, the function h does not take
any integer values. In particular, it does not take the values 1 and 1. Hence,
by the previous lemma again, there exists a holomorphic function g : U ! C
such that we can write h = cos( g).
To prove the second part of the statement, consider the set
p
E = fk i 1 log(n + n2 1) : k 2 Z; n 2 Ng:

We claim that g(U ) \ E = ;. To see this, let w 2 E. Then by Euler’s formula

(20),

ei w
+e i w
1 i k p
log(n+ n2 1) i k
p
log(n+ n2 1)
cos( w) = = (e e +e e )
2 2
1 1 p
= ( 1)k p + n + n2 1
2 n + n2 1
1
= ( 1)k 2n = ( 1)k n:
2
Hence, cos( cos( w)) = cos( ( 1)k n) 2 f 1; 1g. In turn. 12 [1+cos( cos( w))] 2
f0; 1g. Since f does not take values 0 and 1, g cannot take value w. This proves
the claim.
The points in E are the verticespof a rectangular grid. Considerpthe rec-
tangle of vertices kp+ i 1 log(n + n2 1), k + 1 + i 1 log(n p + n
2 1),
k + i 1 log(n + 1 + (n + 1)2 1), and k + 1 + i 1 log(n + 1 + (n + 1)2 1).
The base has length 1 and the height has length
p p
log(n + 1 + (n + 1)2 1) log(n + n2 1)
q
p 1
n + 1 + (n + 1) 2 1 1 + n + 1 + n2
= log p = log q
n + n2 1 1 + 1 n12
r
1 2 p
< log(1 + + 1 + ) log(2 + 3) 1:317 < ;
n n
where we factor out n and used the monotonicity of the logarithm. Hence, the
height of the rectangle is less than 1. Thus for every w 2 C we can …nd z 2 E
such that j Re w Re zj 21 , j Im w Im zj < 21 , which implies that jw zj < 1.
This shows that every ball of radius 1 intersects E. Since g(U ) does not intersect
E, it cannot intersect any ball of radius 1.
Wednesday, February 26, 2020
We are now ready to prove Picard’s little theorem.
Proof of Theorem 156. Assume by contradiction that there exist a; b 2 C
with a 6= b such that f : C ! C does not takes value a and b. The the function
f (z) a
h(z) = ; z 2 C;
b a

68
does not take the values 0 and 1. Hence, by the previous lemma there exists an
entire function g : C ! C such that
1
h(z) = [1 + cos( cos( g(z)))]:
2
Moreover, g(C) does not contain any ball of radius 1. However, since g is not
constant, by Corollary 154 we have a contradiction.
Another important theorem is the following.

Theorem 159 (Schottky) Let U C be an open set which contains B(0; 1),
let > 0, 0 < r < 1, and let f : U ! C be a holomorphic function which does
not take values 0 and 1 and such that jf (0)j . Then

jf (z)j exp( exp( (3 + + 12r=(1 r)))) for all z 2 B(0; r): (59)

Proof. Since U contains B(0; 1), we can …nd R > 1 such that contains
B(0; 1) B(0; R) U . In the remaining of the proof we take U = B(0; R), so
that U is simply connected. As in the proof of Lemma 158, since f does not
take the values 0 and 1, the function 2f 1 does not take the values 1 and 1
and so by Lemma 157 there exists a holomorphic function h : U ! C such that
2f 1 = cos( h) in U . By periodicity, we can add to h any integer multiple of
2. Hence, without loss of generality, we may assume that

1 Re h(0) 1:

By Exercise 33, for every w = x + iy we have that

jyj cosh y j cos wj (60)

and so
j Im h(0)j j cos( h(0))j = j2f (0) 1j 2jf (0)j + 1:
Hence,
2 1
jh(0)j 1+ jf (0)j + < 2 + jf (0)j: (61)

Since 2f 1 does not take the values 1 and 1, the function h omits all integer
values. In particular, it omits the values 1 and 1 and so by Lemma 158 there
exists a holomorphic function g : U ! C such that h = cos( g). Moreover,
g(U ) does not contain any ball of radius 1.
Reasoning as in the …rst part of the proof, by periodicity we can add to g
any integer multiple of 2 and so we can assume that 1 Re g(0) 1. By (60)
and (61),
j Im g(0)j j cos( g(0))j = jh(0)j 2 + jf (0)j;
and so
2 2
jg(0)j 1+ jf (0)j + 3 + jf (0)j: (62)

69
If jzj r < 1, then dist(z; @B(0; 1)) 1 r. On one hand, g(U ) does not
contain any ball of radius 1. On the other hand, by Corollary 153, if g 0 (z) 6= 0,
1
then g(U ) contains balls of every radius 12 (1 r)jg 0 (z)j. Hence,

1
(1 r)jg 0 (z)j < 1
12
for all z 2 B(0; r). By the fundamental theorem of calculus,
Z
g(z) g(0) = g0 ( ) d
[0;z]

and so by the previous inequality, (62), and the fact that jf (0)j ,

jg(z)j jg(0)j + 12jzj=(1 r) 3+ + 12r=(1 r): (63)

Since j cos wj ejwj and 21 j1 + cos wj ejwj , it follows that

1
jf (z)j j1 + cos( cos( g(z)))j exp( j cos( g(z))j)
2
exp( exp( jg(z)j)) exp( exp( (3 + + 12r=(1 r))));

where in the last inequality we used (63).

The beauty of Schottky’s theorem is that the right-hand side of (59) depends
only on and r. Hence, we have a universal bound.

12 Sequences of Holomorphic Functions

Theorem 160 Let U C be an open set and let fn : U ! C be holomorphic
functions which converge uniformly on compact sets of U to a function f : U !
C. Then f is holomorphic and ffn0 gn converges uniformly to f 0 on compact sets
of U .

Proof. By Goursat’s theorem,

Z
fn = 0
@T

for every n and for every closed triangle T U . Letting n ! 1 and using
uniform convergence we get Z
f =0
@T
and so by the previous corollary f is holomorphic in every open ball contained
in U , which implies that f is holomorphic in U .
To prove the second part of the statement, we use (32) to get
Z
1 fn ( )
fn0 (z) = d ;
2 i @B(z0 ;r) ( z)2

70
for every B(z0 ; r) U and every z 2 B(z0 ; r). If z 2 B(z0 ; ), where 0 < < r,
since j zj j z0 j jz0 zj r ,
Z
f( ) fn ( ) 2 r
jf 0 (z) fn0 (z)j = d kf fn kC(@B(z0 ;r))
@B(z0 ;r) ( z)2 (r )2

and so there is uniform convergence in B(z0 ; ). Since any compact set K U

can be covered by a …nite number of these balls, we have uniform convergence
of ffn0 gn on compact sets of U .

De…nition 161 A metric space (X; d) is separable if there exists a countable

subset that is dense in X.

De…nition 162 Let (X; dX ) and (Y; dY ) be metric spaces. A family F of func-
tions f : X ! Y is said to be equicontinuous at a point x0 2 X if for every
" > 0 there exists = (x0 ; ") > 0 such that

dY (f (x); f (x0 )) "

for all f 2 F and for all x 2 X with d (x; x0 ) . The family F of functions
f : X ! Y is said to be uniformly equicontinuous if for every " > 0 there exists
> 0 such that
dY (f (x); f (y)) "
for all f 2 F and for all x; y 2 X with d(x; y) .

Theorem 163 (Ascoli–Arzelà) Let (X; d) be a separable metric space and let
F Cb (X) be a family of functions. Assume that F is bounded and equicon-
tinuous at every point x 2 X. Then every sequence in F has a subsequence that
converges pointwise to a function g 2 Cb (X) and uniformly on every compact
subset of X.

Friday, February 28, 2020

Theorem 164 (Montel) Let U C be an open set and let F be a family of

holomorphic functions de…ned on U . Assume that for every K U there exists
a constant MK > 0 such that

jf (z)j MK

for all f 2 F and for all z 2 K. Then the family F is equicontinuous on K

and for every sequence in F there is a subsequence which converges uniformly
on compact sets to a holomorphic function f : U ! C.

Proof. Fix a compact set K U and let dK := dist(K; @U ) > 0 and let
0 < r < 31 dK . Then for z 2 K, B(z; 3r) U . Hence, for z; w 2 K with
jz wj < r we can apply the Cauchy’s theorem to get
Z
1 f( ) f( )
f (z) f (w) = d :
2 i @B(w;2r) z w

71
For 2 @B(w; 2r) we have j wj = 2r and j zj j wj jz wj 2r r.
Then
1 1 z w jz wj
= :
z w ( z)( w) 2r2
Hence,
2MK jz wj
jf (z) f (w)j (4 r)
2 2r2
for all z; w 2 K with jz wj < r and for all f 2 F. This shows that the family
F is equicontinuous in K. We can now apply the Ascoli–Arzelà to get that for
every sequence in F there a subsequence converging uniformly on compact sets
to a continuous function. By the previous theorem, the function is holomorphic.

Exercise 165 Let U RN be an open connected set and let f : U ! R be

an analytic function such that f is constant in a ball B U . Prove that f is
constant in U .

Theorem 166 (Hurwitz) Let U C be an open set, let fn : U ! C be a

sequence of functions converging uniformly on compact set to a holomorphic
function f : U ! C. Assume that there exists B(z0 ; r) U such that f (z) 6= 0
for all z 2 @B(z0 ; r). Then there exists n1 such that fn and f have the same
number of zeros in B(z0 ; r) for all n n1 .

Proof. By continuity

:= min jf j > 0:
@B(z0 ;r)

In turn, by uniform convergence on compact sets, there is n such that jfn (z)j
=2 for all z 2 @B(z0 ; r) and all n n . It follows that

1 1 jf (z) fn (z)j 2
= 2
jf (z) fn (z)j;
fn (z) f (z) jf (z)jjfn (z)j

and so f1=fn g converges uniformly to 1=f on @B(z0 ; r). Moreover, since fn0 ! f 0
f0 0
uniformly on compact sets by Theorem 160 it follows that fnn ! ff uniformly
on @B(z0 ; r), and so
Z Z
fn0 (z) f 0 (z)
lim dz = dz:
n!1 @B(z ;r) fn (z) @B(z0 ;r) f (z)
0

R f0
But by the argument principle (see Theorem 136) the integrals @B(z0 ;r) fnn dz
R 0
and @B(z0 ;r) ff dz are the numbers of zeros of fn and f inside B(z0 ; r), and
these numbers are …nite. Since the limit exists, for n large these values must
coincide.
The following corollary will be useful to prove the Riemann mapping theo-
rem.

72
Theorem 167 Let U C be an open connected set and let fn : U ! C be
a sequence of injective holomorphic functions converging uniformly on compact
set to a holomorphic function f : U ! C. Then either f is injective or constant.

Proof. Let z0 2 U . De…ne gn (z) = fn (z) fn (z0 ) and g(z) := f (z) f (z0 ).
Assume that there exists z1 6= z0 such that f (z1 ) = f (z0 ). Then g has a zero at
z1 . If g is not constant, then since the zeros of g are isolated, we can …nd r > 0
such that B(z1 ; r) U and g(z) 6= 0 for all z 2 B(z1 ; r) n fz1 g. In particular,
we are in a position to apply Hurwitz theorem to conclude that for all n large
all functions gn have a zero in B(z1 ; r). But by taking r > 0 we can assume
that z0 2 = B(z1 ; r). Since the functions fn are injective, they cannot have a zero
at z1 , which is a contradiction.
An important application of Schottky’s theorem is a sharpened version of
Montel’s theorem. In what follows, given an open set and fn : U ! C, we say
that the sequence ffn gn converges uniformly to 1 on compact sets if for every
compact set K U and every M > 0 there exists nK;M such that

jfn (z)j M for all z 2 K

and all n nK;M .

Theorem 168 Let U C be an open connected set and let F be the family
of holomorphic functions f : U ! C which do not take the values 0 and 1.
Then for every sequence ffn gn in F there is a subsequence ffnk gk such that
ffnk gk converges uniformly on compact sets either to a holomorphic function
f : U ! C or to 1.

Proof. Step 1: Let z0 2 U and > 0 and let

Fz0 ; := ff 2 F : jf (z0 )j g:

We claim that there exist > 0 and M > 0 such that

jf (z)j M

for all z 2 B(z0 ; ) and all f 2 Fz0 ; . To see this, let r > 0 be so small that
B(z0 ; 2r) U . By a dilation and a translation, without loss of generality, we
may assume that z0 = 0 and 2r = 1. Then by Schottky’s theorem with r = 1=2,

jf (z)j exp( exp( (3 + + 12)))

for all z 2 B(0; 1=2) and all f 2 Fz0 ;r .

73
Monday, March 2, 2020
Proof. Step 2: Fix z1 2 U and let

Fz1 ;1 := ff 2 F : jf (z1 )j 1g:

Consider the set V := fz 2 U : Fz1 ;1 is equibounded in a neighborhood of zg.

The set V is open, since if w 2 V , then there are B(w; r) U and L > 0 such
that jf (z)j L for all z 2 B(w; r) and all f 2 Fz1 ;1 . But since B(z; r jz wj)
B(w; r), it follows that w is an interior point of V , and so V is open. Moreover,
V is nonempty in view of Step 1. We claim that V = U . If not, then using the
previous step there exists z2 2 @V \ U and a sequence of functions ffn gn in
Fz1 ;1 such that
lim jfn (z2 )j = 1: (64)
n!1

De…ne gn := 1=fn . Then gn is holomorphic in U and does not take values 0 and
1. Hence, gn 2 F. In view of (64),

lim gn (z2 ) = 0 (65)

n!1

and so there is > 0 such that jgn (z2 )j for all n. In turn, by Step 1, the
sequence fgn gn is equibounded in a neighborhood B(z2 ; r) of z2 . It follows by
Montel’s theorem (Theorem 164) that there exist a subsequence fgnk gk and a
holomorphic function g : B(z2 ; r) ! C such that gnk ! g uniformly on compact
sets of B(z2 ; r). In view of (65), g(z2 ) = 0, but since gn does not vanish in U , it
follows from Hurwitz’s theorem (see Theorem 166) that g 0 in B(z2 ; r). This
implies that limn!1 jfn (z)j = 1 for all z 2 B(z2 ; r). But since z2 2 @V \U , this
implies that there exist points z 2 B(z2 ; r) \ V such that limn!1 jfn (z)j = 1,
which is a contradiction by the de…nition of V . Hence, the claim holds and so
V = U.
Step 3: Let ffn gn be a sequence of functions in F. If there exists countably
many n such that fn 2 Fz1 ;1 , say fnk 2 Fz1 ;1 , then by the previous step,
the sequence ffnk gk is locally bounded on compact sets, and thus by Montel’s
theorem there exists a further subsequence converging uniformly on compact
set to a holomorphic function. On the other hand, if only …nitely many fn
belong to Fz1 ;1 , then jfn (z1 )j > 1 for all n su¢ ciently large. In turn, f1n 2 Fz1 ;1
for all n su¢ ciently large. By the previous step and Montel’s theorem, there
exists a subsequence ffnk gk and a holomorphic function g : U ! C such that
f1=fnk gk converges uniformly on compact set to g. If g never vanishes, then
ffnk gk converges uniformly to the holomorphic function 1=g : U ! C. If g
vanishes at some point, then by Hurwitz’s theorem, g 0 (since 1=fnk never
vanishes). In turn, ffnk gk converges uniformly on compact set to 1.

13 Picard’s Big Theorem

In this section we prove Picard’s big theorem.

74
Theorem 169 (Picard Big Theorem) Let z0 2 C, r > 0, and let f : B(z0 ; r)n
fz0 g ! C be a holomorphic function having an essential singularity at z0 . Then
f takes all possible values of C with at most a single exception.

Proof. Without loss of generality we assume that z0 = 0, that r = 1. As-

sume by contradiction that f does not assume two values a and b. By composing
f with a linear function, we can assume that f does not take values 0 and 1.
Consider the sequences of functions

fn (z) := f (z=n); z 2 B(0; 1) n f0g:

In view of the previous theorem, taking K = @B(0; 1=2), we can …nd a sub-
sequence ffnk gk such that ffnk gk is equibounded in @B(0; 1=2) or f1=fnk gk is
equibounded in @B(0; 1=2). In the …rst case, there exists M > 0 such that

jf (z=nk )j M for all z 2 @B(0; 1=2)

and all k. In turn,

jf (w)j M for all w 2 @B(0; 1=(2nk ))

and all k. It follows by the maximum modulus principle that

jf (w)j M for all 1=(2nk + 1) < jzj < 1=(2nk )

and for all k. But this implies that f is bounded in a neighborhood of z0 , and so
it has a removable singularity at z0 by Theorem 119, which is a contradiction.
Similarly, if f1=fnk gk is equibounded in @B(0; 1=2), then 1=f is bounded in
a neighborhood of z0 , which implies that 1=f has a removable at z0 , again, by
Theorem 119, that is, there exists
1
lim = ` 2 C:
z!z0 f (z)
If ` 6= 0 then f has a removable singularity at z0 , while if ` = 0, then f has a
pole at z0 . This is again a contradiction.
Wednesday, March 4, 2020

14 Entire Functions
We begin by reviewing in…nite products.

14.1 In…nite Products

De…nition 170 Given a sequence fzn gn of complex numbers, we say that the
in…nite product
Y1
(1 + zn )
n=1

75
converges if there exists
k
Y
lim (1 + zn ) = ` 2 C:
k!1
n=1

The following theorem gives a necessary condition for the convergence of an

in…nite product.
1
X
Theorem 171 Given a sequence fzn gn of complex numbers, if the series jzn j
n=1
1
Y
converges, then the in…nite product (1+zn ) converges. Moreover, the product
n=1
converges to 0 if and only if 1 + an = 0 for some n.

Proof. By Theorem 20, limn!1 zn = 0, and so there exists n1 2 N such

that jzn j < 21 for all n n1 . By Exercise 36, for z 2 W \ B(0; 1),
1
X zn
logW (1 + z) = ( 1)n ; (66)
n=1
n

where W = C n fz 2 C : z = x + 0i; x 0g and logW is the principal branch of

the logarithm. In particular, if jzj < 12 ,

X1 1
X
jzjn jzj
j logW (1 + z)j jzjn = 2jzj: (67)
n=1
n n=1
1 jzj

For k n1 we use (66) to write

k k n
!
Y Y X
logW (1+zn )
(1 + zn ) = e = exp logW (1 + zn ) :
n=n1 n=n1 n=n1

X1
By (67), j logW (1 + zn )j 2jzn j and since jzn j converges, by the com-
P1 n=1
parison test, the series n=n1 j logW (1 + zn )j converges. Hence, the series
P1
n=n1 log W (1 + zn ) converges absolutely. In particular, there exists
n
X
lim logW (1 + zn ) = ` 2 C:
k!1
n=n1

By the continuity of the exponential function, there exists

k n
!
Y X
lim (1 + zn ) = lim exp logW (1 + zn ) = e` :
k!1 k!1
n=n1 n=n1

76
In turn,
k
Y n1
Y k
Y n1
Y
(1 + zn ) = (1 + zn ) (1 + zn ) ! (1 + zn )e` :
n=1 n=1 n=n1 n=1

This concludes the …rst part of the proof.

Yk
If 1 + zm = 0 for some m, then (1 + zn ) = 0 for all k m and so the
n=1
in…nite product converges to zero. On the other hand, if 1 + zn 6= 0 for all n,
k
Y n1
Y
then by the previous part we have that (1 + zn ) ! (1 + zn )e` =: `1 . Since
n=1 n=1
e` 6= 0, it follows that `1 6= 0.
As a corollary of the previous theorem we have the following result.

Theorem 172 Let U C be an open set and let fn : U ! C be holomorphic

functions, n 2 N. Assume that for each n 2 N there exists an > 0, such that

jfn (z) 1j an for all z 2 U: (68)

1
X 1
Y
If an converges, then the in…nite product fn (z) converges uniformly to a
n=1 n=1
holomorphic function P : U ! C. Moreover, if fn (z) 6= 0 for all z 2 U and all
n 2 N, then P (z) 6= 0 for all z 2 U and
1
P 0 (z) X fn0 (z)
= for all z 2 U:
P (z) f (z)
n=1 n

Proof. Let n1 2 N be such that an < 21 for all n n1 . In view of (67) and
(68),
j logW fn (z)j = j logW (1 + (fn (z) 1))j 2jfn (z) 1j 2an
for all n n1 . Taking the supremum over all z 2 U gives

sup j logW fn (z)j 2an

U

and so the series

1
X 1
X
sup j logW fn (z)j 2an = R :< 1:
n=n1 U n=n1
P1
This implies that the series of functions n=n1 logW fn converges uniformly in
Pk
U and that n=n1 logW fn (z) 2 B(0; R) for all k n1 and all z 2 U . Since
w 7! ew is continuous, it follows that
k k
!
Y X
gk (z) := fn (z) = exp logW fn (z)
n=n1 n=n1

77
converges uniformly in U to some function g : U ! C, with
1
!
X
g(z) = exp logW fn (z) : (69)
n=n1

By Theorem 160, g is holomorphic and gk0 ! g 0 uniformly on compact sets of

U.
De…ne
n1
Y
P (z) := g(z)h(z); h(z) := fn (z);
n=1
k
Y
Pk (z) := fn (z) = gk (z)h(z)
n=1

Then
sup jPk (z) P (z)j = sup jh(z) (gk (z) g(z))j
U U
= sup jh(z)j jgk (z) g(z)j
U
L sup jgk (z) g(z)j ! 0
U

as k ! 1, where we used the fact that jh(z)j L for all z 2 U by (68), with
n1
Y
L := (1 + an ):
n=1

Next, assume that fn (z) 6= 0 for all z 2 U and all n 2 N and …x a compact
set K U . Since g is the exponential of a holomorphic function g(z) 6= 0 for
all z 2 U . In particular, jg(z)j 0 for all z 2 K. Moreover, by assumption
h(z) 6= 0 for all z 2 U and so jh(z)j 1 for all z 2 K. This implies that
jf (z)j 1 0 =: 2 for all z 2 K. By uniform convergence we have that

1
jPk (z)j 1 for all z 2 K and all k k1 ; (70)
2
where k1 depends only on K. Since gk0 ! g 0 uniformly on compact sets and
Pk = hgk then Pk0 = h0 gk + hgk0 converges uniformly on compact sets to P 0 . In
turn, by (70), Pk0 =Pk ! P 0 =P uniformly in K. Using (51), we get
k
Pk0 (x) X fn0 (x) P 0 (z)
= !
Pk (x) n=1 fn (x) P (z)
uniformly in K. In particular,
1
P 0 (z) X fn0 (x)
=
P (z) f (x)
n=1 n

for all z 2 K. Since this holds for every compact set K U , this concludes the
proof.

78
Exercise 173 Prove that
1
Y
sin( z) z2
=z 1
n=1
n2

Hint: Use Exercise 149.

14.2 Entire Functions of Finite Order

We begin by proving Jensen’s formula.

Theorem 174 (Jensen formula) Let U C be an open set containing 0 and

let f : U ! C be a holomorphic function such that f (0) 6= 0. Then for every for
closed ball B(0; r) U such that f has no zeros on @B(0; r), we have
n
X Z 2
jzk j 1
log jf (0)j = log + log jf (rei )j d ; (71)
r 2 0
k=1

where z1 ; : : : ; zn are the zeros (if any) of f inside B(0; r) counted with multi-
P0
plicities. Here, if n = 0, we take k=1 := 0.

Proof. Step 1: Assume …rst that f has no zeros inside B(0; r). We claim
that Z 2
1
log jf (0)j = log jf (rei )j d : (72)
2 0
Consider an open ball B U containing B(0; r). Since B is simply connected,
by Corollary 100 there exists a holomorphic function g : B ! C such that

f (z) = eg(z) for all z 2 B:

Taking the modulus on both sides we have

jf (z)j = jeg(z) j = jeRe g(z)+i Im g(z) j = jeRe g(z) ei Im g(z) j

= jeRe g(z) jjei Im g(z) j = eRe g(z)

and so log jf (z)j = Re g(z). We now apply the mean value formula (37) (see
Theorem 110) to Re g, to get (72).

79
Monday, March 16, 2020
No class.
Wednesday, March 16, 2020
Online teaching.
Proof. Step 2: Next assume that f (z) = z w0 for some w0 2 B(0; r)nf0g.
We claim that
Z 2
jw0 j 1
log jw0 j = log + log jrei w0 j d : (73)
r 2 0
jw0 j
Writing log r = log jw0 j log r and

log jrei w0 j = log(rjei w0 =rj) = log r + log jei w0 =rj;

we have that formula (73) is equivalent to

Z 2 Z 2
0= log jei 0 j d = log je is
0 j ds
0 0
Z 2 Z 2
is is is is
= log je e e 0 j ds = log(je jj1 eis 0 j) ds (74)
0 0
Z 2
= log j1 eis 0 j ds;
0

where j 0 j < 1 and we have made the change of variables = s. Since the
holomorphic function h(z) = 1 z 0 does not vanish in B(0; 1), we can apply
Step 1 together with the fact that h(0) = 1, to get
Z 2
1
0 = log jh(0)j = log j1 eis 0 j ds;
2 0

which proves (73) in view of (74).

Step 3: Let f1 : U ! C and f2 : U ! C be holomorphic function such that
f1 (0) 6= 0 and f2 (0) 6= 0, and f1 and f2 have no zeros on @B(0; r). We claim that
if f1 and f2 satisfy Jensen’s formula (71), then so does their product f1 f2 . Let
z1 ; : : : ; zn1 and w1 ; : : : ; wn2 be the zeros of f1 and f2 inside B(0; r), respectively.
Then f1 f2 has zeros z1 ; : : : ; zn1 and w1 ; : : : ; wn2 . Moreover,

log j(f1 f2 )(0)j = log(jf1 (0)jjf2 (0)j) = log jf1 (0)j + log jf2 (0)j
Xn1 Z 2
jzk j 1
= log + log jf1 (rei )j d
r 2 0
k=1
X n2 Z 2
jwk j 1
+ log + log jf2 (rei )j d
r 2 0
k=1
Xn1 Xn2 Z 2
jzk j jwk j 1
= log + log + log j(f1 f2 )(rei )j d :
r r 2 0
k=1 k=1

80
Step 4: We are …nally ready to prove the general case. Let f : U ! C be a
holomorphic function such that f (0) 6= 0 and f has no zeros on @B(0; r). Let
z1 ; : : : ; zn be the zeros of f inside B(0; r) counted with multiplicities. Since the
zeros are counted with their multiplicity and are isolated, by Corollary 112 the
function
f (z)
q(z) =
(z z1 ) (z zn )
is de…ned in U , holomorphic, and does not vanish in B(0; r). Hence, Jensen’s
formula (71) holds for q by Step 1. On the other hand, by Step 2 it holds for
each function z 7! z zk . Since

f (z) = q(z)(z z1 ) (z zn );

the conclusion follows from Step 3 and an induction argument.

We now de…ne functions of …nite order.

De…nition 175 Given an entire function f : C ! C and a > 0, we say that

f has an order of growth less than or equal a if there exist constants A; B > 0
such that a
jf (z)j AeBjzj for all z 2 C: (75)
We de…ne the order of growth of f as af = inf a, where the in…mum is taken
over all a > 0 such that f has an order of growth less than or equal to a.
2
The function f (z) = ez has order of growth 2.

Theorem 176 Let f : C ! C be an entire function that has an order of growth

less than or equal to a > 0. For every r > 0 let n(r) be the number of zeros
counted with their multiplicity inside B(0; r). Then

n(r) Cra for all r 1 (76)

and for some constant C > 0. Moreover, if fzn gn are the zeros of f di¤ erent
from zero and counted with their multiplicity, then for every b > a,
X 1
< 1: (77)
n
jzn jb

When needed, we write nf for n to highlight the dependence on f .

Proof. Step 1: We …rst show that if f (0) 6= 0 and if f does not vanish on
@B(0; r), then
Z r Z 2
n(s) 1
ds = log jf (rei )j d log jf (0)j:
0 s 2 0
In view of Jensen’s formula, it is enough to show that
Z r Xn
n(s) jzk j
ds = log ;
0 s r
k=1

81
where z1 ; : : : ; zn are the zeros of f inside B(0; r) counted with their multiplicity.
To see this, observe that
n
X n Z
X r
jzk j 1
log = ds:
r jzk j s
k=1 k=1

Write
n
X
n(s) = (jzk j;1) (s):
k=1

Then
Xn Z r X n Z r Z r n
X Z r
1 1 1 n(s)
ds = (jzk j;1) (s) ds = (jzk j;1) (s) ds = ds;
jzk j s 0 s 0 k=1 s 0 s
k=1 k=1

which completes the proof of this step.

Step 2: To prove (76), we …rst assume that f (0) 6= 0. Take r > 0 such that
f does not vanish on @B(0; 2r). Since n is increasing,
Z 2r Z 2r
2r 1 n(s)
n(r) log 2 = n(r) log = n(r) ds ds:
r r s r s
Z 2r Z 2
n(s) 1
ds = log jf (2rei )j d log jf (0)j;
0 s 2 0

where we used the previous step with r replaced by 2r. On the other hand by
(75),
Z 2 Z 2
1 1 a a
log jf (2rei )j d log(AeB2 r
)d
2 0 2 0
Z 2
1 a a
= [log A + log(eB2 r )] d
2 0
= log A + B2a ra :

Combining these inequalities gives

n(r) log 2 log A + B2a ra :

Taking r 1 and C = (log A + B2a )= log 2, we obtain (76) for all r such
that f does not vanish on @B(0; 2r). Fix r 1. Since the number of zeros in
B(0; 2r + 1) is …nite, we have that f does not vanish on @B(0; 2r + 2s) for all but
…nitely many s 2 (0; 1). Consider a sequence sk ! 0+ such f does not vanish
on @B(0; 2r + 2sk ). By what we just proved and the fact that n is increasing,

n(r) n(r + sk ) C(r + sk )a

for all k. It su¢ ces to send k ! 1.

82
Step 3: Next we prove (76), in the case f (0) = 0. Assume that ` is the
multiplicity of 0. Then the function g(z) := f (z)=z ` is holomorphic, ng di¤ers
from nf by `. Moreover, for jzj 1,
jf (z)j a
jg(z)j AeBjzj :
jzj`
On the other hand, since g is holomorphic, there exists A1 > 0 such that
a
jg(z)j A1 A1 eBjzj
for all jzj 1. Hence, by replacing A with maxfA; A1 g, we have that g also has
an order of growth less than or equal to a. By applying Step 2 to g we get
ng (r) Cra for all r su¢ ciently large,
say for r 1 and for some constant C 1. In turn,
nf (r) = ng (r) + ` Cra + ` (C + `)ra :
Step 4: We prove (77). If the number of zeros is …nite, there is nothing to
prove. Thus, we assume that there are in…nitely many zeros. Then by (76),
X 1
X X 1
X X X 1
1 1 1 1
= = nf (2j+1 ) jb
jzn jb j=0
jzn jb j=0 2j
2jb j=0
2
jzn j 1 2j jzn j<2j+1 jzn j<2j+1
1
X 1
X
2(j+1)a 1
C = C2a < 1:
j=0
2jb j=0
2j(b a)

Since there are only …nitely many zeros in B(0; 1), (77)
The next example shows that we cannot take b to be the order of growth of
f.
Example 177 Let f (z) = sin( z). By Euler’s identity
ei z
e i z
f (z) = :
2i
Hence,
jzj
jf (z)j e ;
so f has an order of growth less than or equal to 1. Taking z = ix gives
x x
e e
f (ix) = ;
2i
which shows that the order of growth is 1. Note that f (n) = sin( n) = 0 and
X1
1
= 1:
n=1
n

Friday, March 20, 2020

83
14.3 Weierstrass Theorem
In this section we show that given a sequence fzn gn of complex numbers whose
moduli converge to in…nity, we can construct an entire function which vanishes
exactly at each zn .

Theorem 178 (Weierstrass) Let fzn gn be a sequence of complex numbers

such that jzn j ! 1 as n ! 1. Then there exists an entire function f : C ! C
such that f (zn ) = 0 for all n and f 6= 0 otherwise. Moreover, any other entire
function with the same property is of the form f (z)eg(z) , where g : C ! C is an
entire function.

The natural choice of f would be

1
Y
f (z) = (1 z=zn ):
n=1

However, in general the in…nite product will not converge.

Proof. Step 1: De…ne
1 1 n
E0 (z) = 1 z; En (z) = (1 z) exp(z + z 2 + + z ): (78)
2 n
We claim that if jzj 1=2, then

j1 En (z)j 2ejzjn+1 :

By Exercise 36, for z 2 W \ B(0; 1),

1
X zk
logW (1 z) = ;
k
k=1

where W = C n fz 2 C : z = x + 0i; x 0g and logW is the principal branch of

the logarithm. Writing 1 z = elogW (1 z) , we have

1 1 n
En (z) = exp logW (1 z) + z + z 2 + + z (79)
2 n
1
!
X zk
= exp =: ew :
k
k=n+1

1
In particular, if jzj 2,

1
X 1
X 1
X X1
zk n+1 jzjk n 1
n+1 j n+1 1
jwj = jzj jzj jzj jzj j
= 2jzjn+1 1:
k k j=0 j=0
2
k=n+1 k=n+1
(80)

84
Hence,
1
X 1
X 1
X
wk jwjk jwjk 1
j1 En (z)j = j1 ew j = = jwj
k! k! k!
k=1 k=1 k=1
1
X 1
jwj = jwj(e 1) 2(e 1)jzjn+1 ;
k!
k=1

which proves the claim for z 2 W \ B(0; 1=2). For z 2 B(0; 1=2) we can use the
fact that En and jzjn+1 are continuous functions.
Step 2: We are now ready to construct the function f . Since jzn j ! 1, by
relabelling the sequence, we can assume that
jz1 j jz2 j jzn j jzn+1 j
for all n. If 0 is one of the numbers zn with multiplicity ` we de…ne
1
Y
f (z) = z ` En (z=zn );
n=1

while if zero is not, we take ` = 0 and set z 0 := 1 in the previous de…nition.

Fix r > 0 and consider z 2 B(0; r). Let n1 > 1 be such that jzn j 2r for all
n n1 . Then jz=zn j 1=2 and so by the previous step
j1 En (z=zn )j 2ejz=zn jn+1 2e=2n+1 :
P1 e
Since the series n=n1 2n converges, by Theorem 172, the in…nite product
1
Y
En (z=zn ) converges uniformly to a holomorphic function P : B(0; r) ! C.
n=n1
Moreover, since En (z=zn ) vanishes only at zn , we have that if En (z=zn ) 6= 0 for
all z 2 B(0; r) and all n n1 . Thus, again by Theorem 172, P (z) 6= 0 for all
z 2 B(0; r). Since
nY
1 1

f (z) = z ` En (z=zn )P (z);

n=1
we have that f is holomorphic in B(0; r). Moreover, since P 6= 0 in B(0; R),
En (z=zn ) vanishes only at zn , we have that f vanishes only at those zn , n =
1; : : : ; n1 1, which are inside B(0; r). By the arbitrariness of r > 0 this
concludes the …rst part of the proof.
Step 3: Let h : C ! C be an entire function such that h(zn ) = 0 for all
n and h(z) 6= 0 otherwise. If wk is a zero of h and f with multiplicity mk , by
Corollary 112 applied f and h we can write
h(z) = (z wk )mk h1 (z); f (z) = (z wk )mk f1 (z);
where h1 and f1 are holomorphic functions in some ball B(wk ; rk ) which do not
vanish in B(wk ; rk ). Hence,
h(z) h1 (z)
= for all z 2 B(wk ; rk ) n fwk g:
f (z) f1 (z)

85
This shows that h=f has a removable singularity at wk and does not vanishes
in B(wk ; rk ). By the arbitrariness of the zero wk and the fact that the zeros
are isolated, we have shown that h=f can be extended to C as a holomorphic
function which vanishes nowhere. We now apply Corollary 100 to write h=f = eg
for some entire function g : C ! C. This concludes the proof.
The functions En are called canonical factors and n the degree of the canon-
ical factor.

Corollary 179 Let f : C ! C be an entire function. Then the following hold:

(i) if f (z) 6= 0 for all z 2 C, then there exists an entire function g : C ! C

such that f (z) = eg(z) for all z 2 C,

(ii) if f has …nitely many zeros z1 ; : : : ; zn counted with their multiplicity, then
there exists an entire function g : C ! C such that

f (z) = (z z1 ) (z zn )eg(z)

for all z 2 C,
(iii) if f has in…nitely many zeros fzn gn counted with their multiplicity, then
there exists an entire function g : C ! C such that
1
Y
f (z) = z ` En (z=zn )eg(z)
n=1

for all z 2 C.

Proof. Item (i) is Corollary 100. Items (ii) and (iii) follow as in Step 3 of
the previous proof.
Note that Weierstrass theorem shows that any entire function with in…nitely
many zeros can be written as the product of the function constructed by Weier-
strass and an exponential function. Thus, it provides a way to represent entire
functions. This is why this theorem is called Weierstras representation theorem.
The next theorem shows that if f has …nite order of growth, then the function
g in the exponential is a polynomial.

Theorem 180 (Hadamard) Let f : C ! C be an entire function which has

growth order a and in…nitely many zeros zn . Then
Y
f (z) = ep(z) z ` En (z=zn );
n

where p is a polynomial of degree less than or equal to bac, ` 2 N0 is the order

of the zero of f at z = 0.

86
15 Prime Number Theorem
Throughout this section p denotes a prime number.

Theorem 181 (Prime Number Theorem) Given x 2 R, let (x) be the

number of prime numbers which are less than or equal to x. Then
x
(x) as x ! 1:
log x
Consider the
X1
1
(z) := ; z 2 C; Re z > 1: (81)
n=1
nz
This function is called the Riemann zeta function.

Lemma 182 The function converges absolutely and uniformly on compact

sets of U := fz 2 C : Re z > 1g. Moreover,
Y 1
(z) = z
; z 2 U:
p prim e
1 p

In particular, has no zeros in U .

Proof. We have

jnz j = jez log n j = e(Re z) log n = nRe z

and so if Re z 1 + ", with " > 0, then

X1 1
X
1 1
zj 1+"
< 1;
n=1
jn n=1
n

which implies that there is uniform and absolute convergence in the set fz 2
C : Re z 1 + "g. In particular, there is absolute convergence in U . Hence, we
can rearrange terms in the series.

87
Monday, March 23, 2020
Proof. Let fpn gn be the ordered sequence of prime numbers. For each
` 2 N let S` be the set of all natural numbers which are not divisible by p1 , . . . ,
p` . We claim that
Ỳ 1 X 1
(z) 1 z = : (82)
pl nz
l=1 n2S`

For ` = 1 we have p1 = 2 and so

X1 1
X X 1
1 1 1
(z) 1 = = ;
2z n=1
n z
n=1
(2n)z nz
n2S1

since we removed all the even natural numbers. Hence, the base case ` = 1 is
true. Next assume that the claim holds for ` and let’s prove it for ` + 1. By the
induction hypothesis,

Ỳ 1 X 1
(z) 1 = :
pzl nz
l=1 n2S`

1
Multiply both sides by 1 pz`+1 to get

`+1
Y X1
1 1 1
(z) 1 = 1
pzl pz`+1 nz
l=1 n2S`
X 1 X 1 X 1
= = ;
nz (p`+1 nz ) nz
n2S` n2S` n2S`+1

which proves the claim.

Letting ` ! 1 in (82) gives
1
Y X 1
1
(z) 1 = lim = 1;
pzl `!1 nz
l=1 n2S`
T1
where we used the fact that S`+1 S` and `=1 S` = f1g.
The last part of the statement follows from Theorem 172 and the fact that
1
1 p z 6= 0 for all z 2 U .

Exercise 183 Let z 2 C with Re z > 1. Prove that

Z 1 Z x
1 1 z 1 1
z
dt = ; z+1
dt = z
1 t z 1 n t n xz
for every n 2 N.

Lemma 184 The function z 7! (z) z 1 1 can be extended as an holomorphic

function to the half-plane fz 2 C : Re z > 0g.

88
Proof. By the previous exercise
X1 Z 1 X1 Z n+1
1 1 1 1 1
(z) = dt = dt
z 1 n=1 nz 1 tz n=1 n
nz tz
X1 Z n+1 Z x
z
= z+1
dsdt:
n=1 n n s

Note that
Z n+1 Z x Z n+1 Z n+1
z z
dsdt dsdt
n n sz+1 n n sz+1
1 1
jzj max = jzj Re z+1 :
s2[n;n+1] jsjRe z+1 n
P1 R n+1 R x z
Hence, the series n=1 n n sz+1
dsdt is absolutely convergent for every z 2
C with Re z > 0.
In view of the previous lemma, the Riemann zeta function can be extended
as a meromorphic function to fz 2 C : Re z > 0g with a simple pole in z = 1
and no other poles. Next we study the zeros of . The Riemann hypothesis is
the conjecture that all zeros of lie on the line Re z = 12 .
The following lemma shows that there are no zeros for Re z 1.

Lemma 185 The Riemann zeta has no zeros in fz 2 C : Re z = 1g.

Proof. Step 1: Let U := fz 2 C : Re z > 1g. Since has no zeros in U ,

using Lemma 182 and Theorem 172,
0
1 p z log p
0
(z) X 1 p z X (1 p z )2
X p z
log p
= 1 = 1 = ; (83)
(z) z 1 p z
p prim e 1 p p prim e 1 p z p prim e

where we used the fact that pz = ez log p and so (pz )0 = ez log p log p = pz log p.
Using the geometric series we have that
1
X
1 kz
z
= p :
1 p
k=0

Hence,
0 1
X X 1
X X
(z) (k+1)z nz
= p log p = p log p:
(z)
p prim e k=0 p prim e n=1

Step 2: Assume that (1 + iy) = 0 and consider the function

g(z) := ( (z))3 ( (z + iy))4 (z + 2iy):

89
c0
Note that has a simple pole at z = 1, so ( (z))3 (z 1)3 , while 1 (z) :=
(z + iy) has a zero of order n 2 N at z = 1 so ( (z + iy))4 c1 (z 1)4 , and
2 (z) := (z + 2iy) may have a zero of order m 2 N0 at z = 1, so (z + 2iy)
c2 (z 1)m . It follows that
c
g(z) (z 1)4n (z 1)m = c(z 1)4n+m 3
(z 1)3

as z ! 1. Thus g has a zero at z = 1 of order 4n + m 3 1. Hence,

g(z) = (z 1)4n+m 3
h(z);

where h is holomorphic near z = 1 and h(1) 6= 0. In turn, by (51),

g 0 (z) (4n + m 3)(z 1)4n+m 4

h0 (z)
= +
g(z) (z 1)4n+m 3 h(z)
4n + m 3 h0 (z)
= +
z 1 h(z)

and so
g 0 (z)
lim (z 1) = 4n + m 3 > 0: (84)
z!1 g(z)
On the other hand, for z 2 C with Re z > 1, by (51) and the previous step,

g 0 (z) ( (z))2 0 (z) ( (z + iy))3 0 (z + iy) 0

(z + 2iy)
=3 3
+4 4
+
g(z) ( (z)) ( (z + iy)) (z + 2iy)
0 0 0
(z) (z + iy) (z + 2iy)
=3 +4 +
(z) (z + iy) (z + 2iy)
X X 1 X X 1 1
X X
nz
= 3 p log p 4 p nz p nyi log p p kz
p 2nyi
log p
p prim e n=1 p prim e n=1 p prim e n=1
X X 1
nyi 2nyi nz
= (3 + 4p +p )p log p:
p prim e n=1

Taking z = x > 1 we have that

1
X X
g 0 (x) nyi 2nyi nx
Re = (Re(3 + 4p +p ))p log p:
g(x)
p prim e n=1
X X 1
nx
= (3 + 4 cos(ny) + cos(2ny))p log p:
p prim e n=1

Since cos(2 ) = 2 cos2 1 we have that

3+4 cos +cos(2 ) = 3+4 cos +2 cos2 1 = 2(1+2 cos +cos2 ) = 2(1+cos )2 :

90
Hence,
g 0 (x)
lim+ (x 1) Re 0;
x!1 g(x)
which contradicts (84).
Wednesday, March 25, 2020
The following theorem is of independent interest.

Theorem 186 Let f : [0; 1) ! C be bounded and locally integrable and let
Z 1
g(z) := f (t)e tz dt; Re z > 0:
0

Assume that for every z 2 C with Re z = 0 there exists rz > 0 such that g can
be extended holomorphically to B(z; rz ). Then the generalized Riemann integral
Z 1
f (t) dt (85)
0

is well-de…ned and equals g(0).

Proof. Using Corollary 115 and a compactness argument for every R >
1 we can …nd = (R) 2 (0; 21 ) and M = M (R) > 0 such that g can be
extended to a holomorphic function g in an open set UR containing the set
CR := B(0; R) \ fz 2 C : Re z g and jg(z)j M for every z 2 CR .
Consider the counterclockwise contour given by the intersection of @B(0; R)
and the segment Re z = , jzj R. Also denote by + and the parts of
in the right half-plane Re z 0 and in the left half-plane Re z 0, respectively.
Let + and be their ranges. Let T > 0 and consider the function
1 z
hT (z) := g(z)ezT + 2 ; z 2 UR n f0g:
z R
If g(0) 6= 0, the function hT has only one pole at 0 with residue res0 hT = g(0),
while if g(0) = 0, then hT is holomorphic in UR . It follows by the residue’s
formula
Z Z
1 z
2 ig(0) = 2 i res0 hT = hT dz = g(z)ezT + 2 dz (86)
z R
Z Z
zT 1 z 1 z
= g(z)e + 2 dz + g(z)ezT + 2 dz:
+
z R z R

If z belongs to the range of +, then by (85), we can write

Z T Z 1
tz
g(z) = f (t)e dt + f (t)e tz dt =: ST (z) + RT (z): (87)
0 T

Consider the function

1 z
qT (z) := ST (z)ezT + 2 ; z 2 B(0; R + 1) n f0g:
z R

91
Again by the residue’s formula
Z Z
1 z
2 iST (0) = 2 i res0 qT = qT dz = ST (z)ezT + 2 dz
@B(0;R) @B(0;R) z R
Z Z
1 z 1 z
= ST (z)ezT + 2 dz + ST (z)ezT + 2 dz
+
z R @B(0;R)n +
z R
(88)
Z Z
1 z 1 w
= ST (z)ezT + 2 dz + ST ( w)e wT
+ 2 dw
+
z R +
w R

where we have made the change of variable z = w. Subtracting (88) from

(86), and using (87) gives
Z
1 z
2 i(g(0) ST (0)) = RT (z)ezT + 2 dz
+
z R
Z
1 1 z
ST ( z) zT + dz (89)
+
e z R2
Z
1 z
+ g(z)ezT + 2 dz =: I + II + III:
z R

We now estimate I, II, and II. Let z = x + iy with x > 0. Since f is bounded,
say, jf (t)j L for all t 2 [0; 1), we have
Z 1 Z 1 t!1 Tx
tz tx 1 tx e
jRT (z)j jf (t)jje jdt C e dt = e = :
T T x t=T x

On the other hand, for z 2 @B(0; R), we have that

1 z z z
+ 2 = + 2 (90)
z R zz R
z z Re z x
= 2+ 2 = 2 = 2
R R R R
In turn, for z 2 +,

Tx
1 z e x 1
RT (z)ezT + 2 exT = 2:
z R x R2 R

Hence,
1
jIj R= : (91)
R2 R
Similarly,
Z T Z T t=T
1 tx eT x 1 eT x
jST ( z)j jf (t)jjetz jdt C etx dt = e = :
0 0 x t=0 x x

92
In turn, for z 2 +,

1 1 z eT x 1 x 1
ST ( z) + 2 = 2:
ezT z R T
x e R x 2 R

It follows that
R
= :jIIj (92)
R2 R
It remains to estimate III. Along the segment given by Re z = , jzj R
we have z = + iy and so

1 z 1 jzj 1 1
+ 2 + + :
z R jzj R2 R

Since jg(z)j M for all z 2 CR , In turn,

1 z 1 1
g(z)ezT + 2 Me T
+
z R R

and so
Z Z R
1 z 1 1
g(z)ezT + 2 dz Me T
+ 1 dy (93)
z R R R

T 2R
= Me +2 :

Friday, March 27, 2020

Proof. On the other hand, on n , we have x = Re z 0 and jzj = R.
Using (90) we have

1 z jxj
g(z)ezT + 2 M exT :
z R R2
p
Since x 0 we can parametrize these two arcs by '(x) = x+ i R2 x2 .
Then r
0 x2 R R 1
j' (x)j = 1 + 2 2
=p p
R x R 2 x 2 R 2 2 2
1
since R2 1> 4
2
. Hence,
Z Z 0
1 z jxj
g(z)ezT + 2 dz M exT dx
n z R R2
Z t=
M tT M 1 Tt
= e t dt = e (T t + 1)
R2 0 R2 T2 t=0
M 1 1 T
e (T + 1) :
R2 T2 T2

93
Together with (93) this shows that

2R M
jIIIj MT +2 + :
R2 T 2

Combining this inequality with (91) and (92), it follows from (89) that

T 2R M
j2 i(g(0) ST (0))j + + + Me +2 + :
R T R R2 T 2

We now choose R = 1" . This determines = (") and M = M ("). Since

T 2R M
lim + Me +2 + = 0;
T !1 T R2 T 2

taking T su¢ ciently large, we have that

j2 i(g(0) ST (0))j 2 " + ";

which proves that ST (0) ! g(0) as T ! 1.

De…ne X
(x) := log p; x 2 R:
p prim e x

Theorem 187 The generalized Riemann integral

Z 1
(x) x
dx
1 x2

converges. In turn,
(x)
lim = 1: (94)
x!1 x
Proof. Step 1: We claim that there exists a constant C > 0 such that

j (x)j Cx

for all x > 0 su¢ ciently large. For n 2 N, by the binomial theorem

2n 2n 2n (2n)!
22n = (1 + 1)2n = + + =
0 n n (n!)2
0 Q 1
n
Y1 Y Y
(2n k) p 2n p
= p = exp log @ A
p = exp log Q
n! p np
k=0 n<p 2n n<p 2n
0 1
X X
= exp @ log p log pA = e (2n) (n) ;
p 2n p n

94
where in the last inequality we used the fact that 2n
n is an integer (this can be
proved by induction). Taking logarithms on both sides gives

2n log 2 (2n) (n):

Hence for m 2 N,
m
X m
X
(2m ) = ( (2n ) (2n 1
)) log 2 2n = (2m+1 2) log 2 < 2m+1 log 2:
n=1 n=1

Given x > 1 …nd m 2 N such that 2m 1

x < 2m . Since is increasing,

(x) (2m ) 2m+1 log 2 x4 log 2;

which proves the claim.

R 1 Step 2: Observe that in view of the previous step, for Re z > 1 the integral
(z+1)t
0
e (et ) dt is well-de…ned. Indeed,
(z+1)t t(Re z+1)
je j=e :

Let pn be the n-th prime number. If pn < et < pn+1 , then

X
(et ) = log p = (pn );
p prim e et

or equivalently, (et ) = (pn ) for all log pn < t < log pn+1 . Also (et ) = 0 for
0 < t < log 2 = log p1 . Hence, for Re z > 1,
Z 1 X1 Z log pn+1 1
X Z log pn+1
(z+1)t t (z+1)t t
e (e ) dt = e (e ) dt = (pn ) e (z+1)t dt
0 n=1 log pn n=1 log pn
1
X e (z+1)t t=log pn+1
1
1
X h i
(z+1)
= (pn ) = (pn ) pn (z+1) pn+1
n=1
z+1 t=log pn z + 1 n=1
1
X 1
1 1 X (z+1)
= (pn )pn (z+1) (pk 1 )pk
z + 1 n=1 z+1
k=2
1
1 (z+1) 1 X (z+1)
= 2 log 2 + ( (pn ) (pn 1 ))pn
z+1 z + 1 n=2
1
1 (z+1) 1 X (z+1) (z + 1)
= 2 log 2 + p log pn = ;
z+1 z + 1 n=2 n z+1

where in the second to last equality we used the fact that (pn ) (pn 1) = log pn
and we set k = n + 1 and where
X log p
(z) := ; z 2 C; Re z > 1:
pz
p prim e

95
Monday, March 30, 2020
1
Proof. Step 3: We prove that the function z 7! (z) z 1 can be
extended as a meromorphic function to the half-plane fz 2 C : Re z > 1=2g and
is holomorphic for all z 2 C with Re z 1. Using the identity
1 1 1
= + z z
pz 1 pz p (p 1)

by (83) we can write

0
(z) X log p X log p X log p
= z
= z
+
(z) p 1 p pz (pz 1)
p prim e p prim e p prim e
X log p
= (z) + :
pz (pz 1)
p prim e

Note that for Re z > 12 , and p > 4,

1 Re z
jpz 1j jpz j 1 pRe z 1 p
2
and so
log p 2 log p
:
pz (pz 1) p2 Re z
Since the series
X1
log n
n=1
n2 Re z
P log p 1
converges, the series p prim e pz (pz 1) is absolutely convergent for Re z > 2 .
0
(z)
Moreover, by Lemma 184, (z) is a meromorphic function for Re z > 0. Hence,
0
(z) X log p
(z) =
(z) pz (pz 1)
p prim e

1
can be extended as a meromorphic function to Re z > 2 with poles at z = 1
and at the zeros of .
Step 4: Consider the continuous bounded function
t
f (t) = e (et ) 1:

By Step 2 for Re z > 0, we have that

Z 1 Z 1 Z 1
tz (z + 1) 1
f (t)e dt = e t(z+1) (et ) dt e tz
dt = :
0 0 0 z+1 z

It follows from Step 3 that (z+1)

z+1
1
z can be extended to a meromorphic function
g for Re z > 21 , which is holomorphic for Re z 0. Hence, we are in a position

96
R1
to apply Theorem 186 to conclude that the integral 0 f (t) dt is well-de…ned
and Z 1 Z 1
(e t (et ) 1) dt = f (t)dt = g(0):
0 0
By considering the change of variables x = e , that is log x = t, so that x1 dx = dt
t

we have that
Z 1 Z 1
(x) x
dx = (e t (et ) 1) dt = g(0);
1 x2 0

which proves the …rst part of the statement.

Step 5: We prove (94). Assume by contradiction that
(x)
lim sup > 1:
x!1 x
There there exists an increasing sequence xn ! 1 such that (xn ) > (1 + ")xn
for all n 2 N and for some 0 < " < 1. Since is increasing, if x > xn ,
(x) (xn ) > (1 + ")xn , and so
Z (1+")xn Z (1+")xn
(x) x (1 + ")xn x
2
dx dx
xn x xn x2
Z (1+")
(1 + ") s
= ds > 0
1 s2
where we made the change of variables x = xn s so dx = xn ds.
On the other hand, since
Z T
(x) x
lim dx = ` 2 R
T !1 1 x2
there exists T" > 0 such that
Z S Z (1+")
(x) x (1 + ") s
2
dx < ds
T x 1 s2

for all S; T T" . Hence, by taking n so large that xn T" we obtain a

contradiction.
Similarly, if
(x)
lim inf < 1;
x!1 x
There there exists an increasing sequence yn ! 1 such that (yn ) < (1 ")yn
for all n 2 N and for some 0 < " < 1. Since is increasing, if yn > x,
(x) (yn ) (1 ")yn , and so
Z yn Z yn
(x) x (1 ")xn x
2
dx dx
(1 ")yn x (1 ")yn x2
Z 1
(1 ") s
= ds < 0
1 " s2

97
where we made the change of variables x = yn s. On the other hand, there exists
S" > 0 such that
Z S Z 1
(x) x (1 ") s
dx < ds
T x2 1 " s2

for all S; T S" . Hence, by taking n so large that (1 ")yn S" we obtain a
contradiction. This shows that
(x)
lim inf 1;
x!1 x
which would complete the proof.
We turn to the proof of the prime number theorem.
Proof of Theorem 181. For every " 2 (0; 1) and x > 1 we have
X X
(x) = log p log x = (x) log x
p prim e x p prim e x

while
X X
(x) = log p log p
p prim e x x1 " p prim e x
X X
log x1 "
= (1 ") log x
x1 " p prim e x x1 " p prim e x

= (1 ") log x( (x) (x1 "

))
1 "
(1 ") log x( (x) x ):

Hence,
(x) (x) (x) log x
x (1 ") x C :
log x x log x x"
Letting x ! 1 gives

(x) (x) (x)

lim inf x lim (1 ") lim sup x :
x!1
log x
x!1 x x!1 log x

It su¢ ces to let " ! 1 .

Wednesday, April 1, 2020

16 Conformal Mappings
De…nition 188 Given two open set U; V C, a bijective holomorphic function
f : U ! V is called a conformal map. If such a map exists, the sets U and V
are said to be conformally equivalent.

98
We have seen in Corollary 142 that the inverse function of a injective holo-
morphic function is also holomorphic. Hence, the inverse of a conformal mapping
is still a conformal mapping.

Exercise 189 Consider the upper half-plane

H := fz 2 C : Im z > 0g

and let
i z
f (z) = ; z 2 H:
i+z
Prove that f : H ! B(0; 1) is a conformal map.

Mappings of the form

az + b
z 7! ;
cz + d
where a; b; c; d 2 C are called fractional linear transformations.

Example 190 Given n 2 N, the function f (z) = z n is a conformal mapping

from the sector S = fz 2 C : 0 < arg z < =ng to the upper half-plane H.
Its inverse is f 1 (w) = w1=n , de…ned in terms of the principal branch of the
logarithm.

Exercise 191 Let 0 < < 2. Prove that the sector S = fz 2 C : 0 < arg z <
g and the upper half-plane are conformally equivalent.

The Riemann mapping theorem proves that any simply connected open set
which is not the entire space is conformally equivalent to the open unit ball. To
prove the Riemann mapping theorem we will need the following auxiliary result.

Theorem 192 (Schwarz’s lemma) Let f : B(0; 1) ! C be a holomorphic

function such that f (0) = 0 and jf (z)j 1 for all z 2 B(0; 1). Then jf (z)j jzj
for all z 2 B(0; 1) and jf 0 (0)j 1. Moreover, if jf (z0 )j = jz0 j for some z0 2
B(0; 1) or jf 0 (0)j = 1, then f (z) = az for all z 2 B(0; 1) and for some a 2 C
with jaj = 1.

Proof. Since f (0) = 0, we can write

1
X
f (z) = an z n ; z 2 B(0; 1):
n=1

Hence the function

1
X
h(z) := an z n 1
; z 2 B(0; 1)
n=1

is analytic in B(0; 1), since the radius of convergence is the same. In turn,
f (z)
z 2 B(0; 1); z 6= 0;
g(z) := z
0
f (0) z = 0;

99
is holomorphic (since g = h near 0). For every r 2 (0; 1) and every z 2 @B(0; r),
jg(z)j 1=r, and so by the maximum modulus principle jg(z)j 1=r for all
z 2 B(0; r). Letting r ! 1+ , it follows that jg(z)j 1 in B(0; 1). Moreover, if
jg(z0 )j = 1 for some z0 2 B(0; 1), then g must be constant, which shows that
f (z) = az for all z 2 B(0; 1) and for some a 2 C with jaj = 1.

Exercise 193 Let z; 2 C be such that 1 z 6= 0.

(i) Prove that

z
<1
1 z
if jzj < 1 and j j < 1 and that

z
=1
1 z

if jzj = 1 or j j < 1.
(ii) Given 2 B(0; 1), the function : B(0; 1) ! B(0; 1) given by
z
(z) := :
1 z
Prove that is a bijection.

We turn to the proof of the Riemann mapping theorem.

Theorem 194 (Riemann mapping) Let U C be an open simply connected

set. Then U is comformally equivalent to a sphere.

Proof. Step 1: Since U is strictly contained in C there exists 2 C n U .

Hence the function z 7! z never vanishes on the simply connected set U and
so by Exercise 101 we may de…ne the holomorphic function f (z) := logU (z ).
Since ef (z) = z , we have that f is injective. Fix z0 2 U . We claim that

f (z) 6= f (z0 ) + 2 i for all z 2 U: (95)

Indeed, if f (z) = f (z0 ) + 2 i then by taking the exponential on both sides we

get
z = ef (z) = ef (z0 )+2 i = ef (z0 ) e2 i = (z0 )1;
which implies that z = z0 and in turn that f (z) = f (z0 ). This contradicts the
fact that f (z) = f (z0 ) + 2 i. Hence, the claim (95) holds.
We claim that there exists r > 0 small such that B(f (z0 )+2 i; r)\f (U ) = ;.
Indeed, if not then taking r = n1 we could …nd zn 2 U such that f (zn ) !
f (z0 ) + 2 i. Again by exponentiation

zn = ef (zn ) ! ef (z0 )+2 i

= ef (z0 ) e2 i
= (z0 )1;

100
which implies that zn ! z0 , and in turn that f (z0 ) = f (z0 ) + 2 i, which
contradicts (95). It follows that the function
1
F (z) := ; z2U
f (z) (f (z0 ) + 2 i)

is holomorphic. Moreover, since jf (z) (f (z0 ) + 2 i)j r > 0 for all z 2 U , we

have that F is bounded. By a translation and a rescaling we can assume that

F : U ! B(0; 1)

and that 0 2 F (U ). By the open mapping theorem the set F (U ) is open. Since
F : U ! F (U ) is a homeomorphism and U is simply connected, it follows that
F (U ) is also simply connected.
Since U and F (U ) are conformally equivalent, it su¢ ces to prove that F (U )
and B(0; 1) are conformally equivalent.
Friday, April 4, 2020
Proof. Step 2: In view of Step 1, by replacing U with F (U ), without loss
of generality we may assume that U B(0; 1) and that 0 2 U . Let

G := fg : U ! B(0; 1) holomorphic, injective, g(0) = 0g:

The family G is nonempty since the identity belongs to G. Since, jg(z)j 1 for
all z 2 U and 0 2 U , by (32),
Z
0 1 jg( )j 2 r
jg (0)j 2
d
2 @B(0;r) 2 r2

for all g 2 G and for r > 0 such that B(0; r) U . Let

s := supfjg 0 (0)j : g 2 Gg:

Consider a sequence fgn gn in G such that jgn0 (0)j ! s. Since the family G
is equibounded, by Montel’s theorem there exists a subsequence fgnk gk which
converges uniformly on compact sets to a holomorphic function g : U ! C.
By uniform convergence, g(0) = 0 and g : U ! B(0; 1). By Theorem 70,
jgn0 (0)j ! jg 0 (0)j = s. Since s 1 (since the identity has derivative with
modulus one), the function g cannot be constant and thus by Theorem 167 it
must be injective. Since g(U ) is open, it follows that g : U ! B(0; 1). It follows
that g belongs to G.
Step 3: It remains to show that g is onto. Assume by contradiction that
there is 2 B(0; 1) n g(U ). Consider the di¤eomorphism : B(0; 1) ! B(0; 1)
given by
z
(z) := :
1 z
Note that interchanges 0 with , since ( ) = 0 and (0) = . The
set V := ( g)(U ) B(0; 1) is open and simply connected and 0 does not

101
belong to V since 2 B(0; 1) n g(U ) and ( ) = 0. Hence, by Exercise 101
the function h1 : V ! C given by
1 p
h1 (w) := e 2 logV w
= w

is holomorphic and injective and h1 : V ! B(0; 1). It follows that the function

g1 := h1 ( ) h1 g

is injective, holomorphic, and

g1 (0) = h1 ( ) (h1 ( (g(0)))) = h1 ( ) (h1 ( (0)))

= h1 ( ) (h1 ( )) = 0:

Hence, g1 2 G.
1
Next consider the function h2 (w) := w2 and := 1
h2 h1 ( ) . Then

1 1
g1 := h2 h1 ( ) h1 ( ) h1 g
1 1
= h2 h1 g= g=g

and
g 0 (0) = ( g1 )0 (0) = 0
(0)g10 (0)
and so
s = jg 0 (0)j = j 0 (0)jjg10 (0)j:
The function : B(0; 1) ! C satis…es all the hypotheses of Schwarz’s lemma,
but it is not injective since h2 is not injective. Hence j 0 (0)j < 1, which implies
that jg10 (0)j > s and contradicts the maximality of s. Hence, g is onto and the
proof is complete.

Remark 195 In view of Exercise 102, the Riemann mapping theorem continues
to hold is instead of assuming U simply connected, we assume that
Z
f ds = 0

for every holomorphic function f : U ! C and for every closed oriented Lip-
schitz continuous curve with range contained in U .

An important consequence of the Riemann mapping theorem is the following

characterization of simply connected open sets.

Theorem 196 Let U C be an open connected set. Then the following are
equivalent:

(i) U is homeomorphic to an open ball,

(ii) U is simply connected,

102
R
(iii) f dz = 0 for every holomorphic function f : U ! C and for every
recti…able closed oriented curve with range contained in U .

Proof. Assume that U is homeomorphic to an open ball, say B(0; 1). Then
there exists an invertible function : U ! B(0; 1), which is continuous to-
gether with its inverse and consider a continuous closed curve, with parametric
representation ' : [a; b] ! C such that ' ([a; b]) U . De…ne the function
h : [a; b] [0; 1] ! C by
1
h (t; s) = (s (' (t))) :

Then h ([a; b] [0; 1]) U,

1
h (t; 0) = (0) for all t 2 [a; b] ; h (t; 1) = ' (t) for all t 2 [a; b] ;
1 1
h (a; s) = (s (' (a))) = (s (' (b))) = h (b; s) for all s 2 [0; 1] :

Hence, U is simply connected. Hence (ii) holds.

Conversely, assume that U is simply connected. Then by the Riemann map-
ping theorem U is homeomorphic to a ball. This shows that (i) and (ii) are
equivalent.
To show that (ii) and (iii) are equivalent, note that if U is simply connected,
then (iii) holds in view of Theorem 98. Conversely, if (iii) holds then by Remark
195, U is homeomorphic to a ball and so it is simply connected by the equivalence
between (i) and (ii).
Next we study the behavior of conformal mappings at the boundary.

De…nition 197 A set E C is locally connected if for every " > 0 there exists
> 0 such that for all z; w 2 E with 0 < jz wj < there exists a compact
connected set F E such that z; w 2 F and diam F < ".

The range of a continuous curve is locally connected.

103
Monday, April 6, 2020

Exercise 198 Let E1 ; : : : ; En be locally connected. Prove that their union is

locally connected.

Exercise 199 Let

1
[ i i
E = fx + iy : jxj < 1; 0 < y < 1g n ; +1 :
n=1
n n

Prove that @E is not locally connected.

Theorem 200 Let U C be an open bounded simply connected set and and let
f map conformally B(0; 1) onto U . Then the following conditions are equivalent

(i) f can be extended continuously to B(0; 1),

(ii) @U is the range of an oriented closed curve,
(iii) @U is locally connected,
(iv) C n U is locally connected.

In general the extension of f to @B(0; 1) will not be injective.

Example 201 An example of a simply connected domain whose boundary is not

the range of an oriented simple closed curve is U = B(0; 1) n fx : 0 x < 1g.

Indeed, we have the following result:

Theorem 202 (Carathéodory) Let U C be an open bounded simply con-

nected set and let f map conformally B(0; 1) onto U . Then f has a continuous
and injective extension to B(0; 1) if and only if @U is the range of an oriented
simple closed curve.

17 Runge’s Theorem
Next we proof another important theorem. There is a more general statement
but we will prove …rst a simpler version.

Theorem 203 (Runge) Let U C be an open set, let K U be a compact

set with C n K connected, and let f : U ! C be a holomorphic function. Then
there exists a sequence of polynomials pn : C ! C such that pn ! f uniformly
in K.

Exercise 204 Let K C be a compact set.

(i) Let B be an open ball such that K B and let z1 2 C n B. Let f (z) :=
1
z z1 . Prove that there exists a sequence of polynomials which converges
to f uniformly in K.

104
(ii) Assume that C n K is connected and let z0 2 C n K. Let g(z) := z 1z0 .
Prove that there exists a sequence of polynomials which converges to g
uniformly in K.

Lemma 205 Let U C be an open set and let f : U ! C be a holomorphic

function. Then there exist …nitely many oriented segments 1 , . . . , n with
range in U n K such that
n
X Z
1 f( )
f (z) = d
2 i k
z
k=1

for all z 2 K.

Proof. Let d := dist(K; @U ) and partition C into squares of side-length less

than p12 d. Let Q1 ; : : : ; Q` be the closed cubes which intersects K with @Qk
oriented counterclockwise. Since K \ Qk 6= ; and Qk has diameter less than d,
each Qk is contained in U . Let 1 , . . . , n be the oriented sides of these cubes
which do not belong to two adjacent squares. Then each k does not intersect
K since otherwise k would belong to two adjacent cubes intersecting K. Let
z 2 K and assume that z is not on the boundary of one of the cubes. Then
there exists a unique j such that z 2 Qj . It follows by Cauchy’s theorem and
Theorem 98 that Z
1 f( )
f (z) = d :
2 i @Qj z
On the other hand for all k 6= j,
Z
1 f( )
d = 0:
2 i @Qj z

Hence, if we sum these equalities we get

n
X Z n
X Z
1 f( ) 1 f( )
f (z) = d = d ;
2 i @Qk z 2 i k
z
k=1 k=1

where in the second equality we used the fact that integrals over the sides of
adjacent cubes cancel out. This proves the result for all z 2 K not on the
boundary of a cube Qk . Now if z 2 K and z belongs to the boundary of a cube,
then z does not belong to any of the segments k and so by continuity we have
that the formula Z
Xn
1 f( )
f (z) = d
2 i k z
k=1

holds for all z 2 K.

Lemma 206 Let be a Lipschitz continuous oriented curve in C parametrized

by ' : [a; b] ! C, let f : '([a; b]) ! C be a continuous function, and let K C

105
be a compact set with K S \ '([a; b]) = ;. Then for every " > 0 there exists a
n
rational function R : C n j=1 fzj g ! C, where zj 2 '([a; b]) such that
Z
f( )
d R(z) " for all z 2 K:
z
Proof. The function
f ('(t))
g(z; t) := ; (t; z) 2 [a; b] K
'(t) z
is uniformly continuous, therefore we can …nd a partition a = t0 < t1 < <
tn = b such that
f ('(t)) f ('(tj )) "
for all (t; z) 2 [tj 1 ; tj ] K;
'(t) z '(tj ) z M (b a)
where k'0 k1 M . De…ne
Xn
f ('(tj ))
R(z) := ('(tj ) '(tj 1 )):
j=1
'(tj) z

Then
Z n Z
X tj Xn Z
f( ) f ('(t)) 0 f ('(tj )) tj 0
d R(z) = ' (t) dt ' (t) dt
z j=1 tj 1
'(t) z j=1
'(tj ) z tj 1
Xn Z tj
f ('(t)) f ('(tj ))
= '0 (t) dt
j=1 tj 1
'(t) z '(tj ) z

and so
Z n Z
X tj
f( ) f ('(t)) f ('(tj ))
d R(z) j'0 (t)j dt
z j=1 tj 1
'(t) z '(tj ) z
Xn Z tj
"
M dt = ":
j=1 tj 1
M (b a)

This concludes the proof.

Wednesday, April 8, 2020
Lemma 207 Let G C be a set and let F(G) be the family of functions f :
G ! C for which there exists a sequence of polynomials pn such that pn ! f
uniformly in G as n ! 1. If fk 2 F(G) and fk ! f uniformly in G, then
f 2 F(G).
Proof. The proof uses a diagonal argument. Since fk 2 F(G) there exists
a sequence of polynomials pn;k such that pn;k ! fk uniformly in G as n ! 1.
Hence we can …nd nk k such that
1
sup jpn;k (z) fk (k)j
z2K k

106
for all n nk . De…ne
qk (z) := pnk ;k (z):
Then
jf (z) qk (z)j = jf (z) pnk ;k (z)j jf (z) fk (z)j + jfk (z) pnk ;k (z)j
1
jf (z) fk (z)j + :
k
Taking the supremum over all z 2 G, we have that the right-hand side converges
uniformly to zero in G as k ! 1.
Lemma 208 Let K C be a compact set such that C n K is connected. Given
z0 2 C n K, let gz0 (z) := z 1z0 . Then there exists a sequence of polynomials
which converges to gz0 uniformly in K.
Proof. Let F(K) be the space of all functions f : K ! C such that there
exists a sequence of polynomials pn : C ! C such that pn ! f uniformly in K.
Note that if f; g 2 F(K), then f g and f + g 2 F(K). Moreover, if fk 2 F(K)
and fk ! f uniformly in K, then by the previous lemma, f 2 F(K).
Step 1: Let R > 0 be so large that K B(0; R), let z1 2 C n B(0; R),
and let gz1 (z) := z 1z1 . We claim that gz1 2 F(K). Find 0 < r < R such that
K B(0; r). For z 2 K, write
1 1 1
= :
z z1 z1 1 zz1
Then
z r
=: < 1
z1 R
and so we can use geometric power series to write
1
1 X
k
1 1 1 z
= = 1 :
z z1 z1 1 zz1 z1 z1
k=0

Since this geometric series converges uniformly in K (since the number is inde-
P` k
pendent of z), we have that and the polynomials z11 k=0 1 zz1 converge
uniformly to gz1 in K.
Step 2: Let w1 2 C n K and assume that gw1 2 F(K). Let 0 < <
1
4 dist(w1 ; K). We claim that for every w2 2 C with jw1 w2 j < we have that
gw2 2 F(K) in K. To see this we proceed as in the previous step to write for
z 2 K,
1 1 1 1
gw2 (z) = = = w1 w2 :
z w2 z w1 (w1 w2 ) z w1 1 z w1

Then jz w1 j 4 and so
w1 w2 1
= <1
z w1 4 4

107
and so we can use geometric power series to write
1
X k
1 1 1 w1 w2
gw2 (z) = w1 w2 = ;
z w1 1 z w1
z w1 z w1
k=0

where this geometric series converges uniformly in K. Hence, the sequence of

functions
X̀ w1 w2 k
z w1
k=0
k
converges uniformly in K as ` ! 1. Since gw1 2 F(K) we have that gw 1
2
k k
F(K). In turn, (w1 w2 ) gw1 2 F(K) and so

X̀ k k
(w1 w2 ) gw 1
2 F(K):
k=0

P1 w1 w2
k
Hence, k=0 z w1 2 F(K) since the series converges uniformly in K. It
follows that gw2 2 F(K), since it is the product of gw1 and this series.
Step 3: Let R > 0 be so large that K B(0; R). Let z1 2 C n B(0; R).
Given z0 2 C n K, since C n K is connected, we can …nd a polygonal path that
joins z0 and z1 with range in C n K. Let 0 < < 41 dist( ; K). Without loss of
generality we can assume that the endpoints of the segments of have distance
less than . Hence, we can apply Step 2 starting from z1 until we reach z0 .
We turn to the proof of the theorem.
Proof of Runge’s theorem. By Lemma 205 there exist …nitely many
oriented segments 1 , . . . , n with range in U n K such that
n
X Z
1 f( )
f (z) = d
2 i k
z
k=1

for all z 2 K. By Lemma 205 for each " > 0 there exists a rational function Rk
such that Z
1 f( )
d Rk (z) "=n for all z 2 K:
2 i k z
Hence,
n
X
f (z) Rk (z) " for all z 2 K:
k=1

Now each Rk is a sum of rational functions whose denominator has the form
1
z z0 for some z0 2 U n K. We now apply Lemma 208.
Friday, April 10, 2020
We now present a more general version. Let S2 := @B((0; 0; 0); 1) be the
unit sphere in R3 . We can view the complex plane as the plane the plane
f(x; y; 0) : x; y 2 Rg inside R3 . Let N = (0; 0; 1) 2 S2 be the north pole. Given

108
a point z = x + iy there is a unique line passing through N and (x; y; 0) which
intersects S2 at a point S(z) 2 S2 n fN g. The map S gives a bijection between
C and S2 n fN g. Indeed, given (X; Y; Z) 2 S2 n fN g consider

X Y
x= ; y= :
1 Z 1 Z
Conversely, given z = x + iy 2 C we have that

2x 2y x2 + y 2 1
S(z) = ; 2 ; 2
x2 + y + 1 x + y + 1 x + y2 + 1
2 2

1
= (2 Re z; 2 Im z; jzj2 1):
1 + jzj2

If we set S(1) := N we have a bijection between C1 and S2 . Note that

S(z) ! N in R3 if and only if jzj ! 1 in C.
Hence, we can regard C1 as a subset of R3 . In turn, the metric in R3 induces
a metric on C1 . We leave as an exercise to show that this metric is given by

2jz wj 2
d(z; w) = p p ; d(z; 1) = p
1 + jzj2 1 + jwj2 1 + jzj2

for z; w 2 C and that this metric induces the same topology in C. Note that
since S2 is compact, so is C1 .

Theorem 209 (Runge) Let U C be an open set, let K U be a compact

set, let E C1 nU be such that E contains at least one point in each component
of C1 n K, and let f : U ! C be a holomorphic function. Then there exists a
sequence of rational functions rn : C n E ! C with poles in E such that rn ! f
uniformly in K.

We will need two more lemmas.

Lemma 210 Let V; W C be two open sets with V W and @V \ W = ;. If

H is any component of W and H \ V 6= ;, then H V .

Proof. Let H be as in the statement and let z0 2 H \ V . Then there exists

a connected component G of V such that z0 2 G. To conclude the proof, it is
enough to show that H = G.
We have that G H, since G is a connected subset of V (and so of W )
containing z0 and H is the union of all connected subsets of W containing z0 .
Write
H = G [ (H n G) = G [ ((H \ @G) [ (H n G)):
But H \ @G W \ @G W \ @V = ;. Hence, the connected set H is the union
of two disjoint open sets. Since G is nonempty, it follows that H n G = ;, which
shows that H = G.

109
Lemma 211 Let K C be a compact set, let z0 2 C n K, let g(z) := z 1z0 , and
let E C1 n K be such that E contains at least one point in each component
of C1 n K. Then there exists a sequence of rational functions Rn : C n E ! C
with poles in E such that Rn ! g uniformly in K.

Proof. Step 1: Let B(E) be the space of all functions f : K ! C such

that there exists a sequence of rational functions Rn : C n E ! C with poles in
E such that Rn ! f uniformly in K. Note that if f; g 2 B(E), then f g and
f + g 2 B(E). Moreover, if fk 2 B(E) and fk ! f uniformly in K, then by
Lemma 207 (which continues to hold if we replace polynomials with rational
functions), f 2 B(E).
Step 2: Assume that E CnK . Let W := CnK and let V be the set of all
w 2 W such that gw 2 B(E), where gw (z) = z 1w , z 2 K. We claim that V is an
open set. To see this, let w0 2 V and w 2 B(w0 ; r), where r := dist(w0 ; K) > 0.
For z 2 K, write
1 1 1 1
= = w w0 :
z w z w0 (w w0 ) z w0 1 z w0

Then jz w0 j r and so
w w0 jw w0 j
=: <1
z w0 r
and so we can use geometric power series to write
1
X k
1 w w0
w w0 = :
1 z w0
z w0
k=0

Since this geometric series converges uniformly in K (since the number is

P` k
independent of z), and k=0 wz ww00 belongs to B(E), because is it given by
1
products and sums of functions in B(E), by Step 1, w w0 2 B(E), and so
1 z w0
also gw 2 B(E). This shows that B(w0 ; r) V . Thus, V is open.
Next we claim that @V \ W = ;. Let w 2 @V and …nd wn 2 V such that
wn ! w. By what we just proved, if jwn wj < dist(wn ; K), then w 2 V . Since
w2 = V , it must be that

jwn wj dist(wn ; K) dist(w; K) jwn wj:

Letting n ! 1 gives dist(w; K) = 0, which implies that w 2 K, since K is

compact. Recalling that W := C n K, it follows that w 2 = W.
This proves that all the hypotheses of the previous lemma are satis…ed. Let
H be any component of W = C n K. By hypothesis there exists w 2 E \ H.
Moreover gw is a rational function itself with pole in E. Hence, w belongs to
V . By the previous lemma, it follows that H V . This shows that V = C n K,
that is, that for every w 2 C n K there exists a sequence of rational functions
Rn : C n E ! C with poles in E such that Rn ! gw uniformly in K.

110
Monday, April 13, 2020
Proof. Step 3: Assume that 1 2 E C1 n K. Since K is bounded, there
exists a unique unbounded connected component H of C n K. If w0 2 H and
jw0 j is very large, then the Taylor series of gw0 converges uniformly in K (see
Lemma 208). Thus, w0 2 B(S).
By applying Step 2 to (E [fw0 g)nf1g, we conclude that for every w 2 CnK
there exists a sequence of rational functions Rn : C n ((E [ fw0 g) n f1g) ! C
with poles in (E [ fw0 g) n f1g such that Rn ! gw uniformly in K. Write

Rn = Qn + Sn ;

where the poles of Qn are in E n f1g and Sn is either zero or has only a pole
in w0 . Since Sn can be approximated uniformly in K by polynomials, by a
diagonal argument, we can …nd a sequence of rational functions with poles in
E n f1g converging uniformly to gw0 in K. This concludes the proof.
We turn to the proof of Runge’s theorem.
Proof. We proceed as in the proof of Theorem 203 with the only di¤erence
that in place of Lemma 208 we apply the previous lemma.

17.1 Mittag-Le- er Theorem

This is the analog of Weierstrass representation theorem for meromorphic func-
tions. In the statement we will use the fact that if U C is an open set and
E U is a set with no accumulation points in U , then E is countable.

Theorem 212 Let U C be an open set, let E = fwn : n 2 Ig U be a set

with no accumulation points in U , where I N and let
an;1 an;`k
Sn (z) = + + :
z wn (z wn )`k

Then there exists a meromorphic function f : U n E ! C whose only poles are

at E and whose principal part at wn is Sn .

Proof. Step 1: Let K0 := ; and

Kj := B(0; j) \ fz 2 C : dist(z; C n U ) 1=jg:

S1
Then Kj Kj+1 and j=1 Kj = U .
Note that

C1 nKj = (C1 nB(0; j))[(B(0; j)nU )[fz 2 U \B(0; j) : dist(z; CnU ) < 1=jg:
(96)
We claim that each component of C1 n Kj contains a component of C1 n U .
Indeed, since C1 n U C1 n Kj , if we consider the component G of C1 n Kj
which contains 1, it must contain the component H of C1 n U which contains
1 (since H is connected, 1 2 H and H C1 n Kj ). On the other hand,
since Kj B(0; j), we have that G contains C1 n B(0; j), since the latter is

111
a connected set contained in C1 n Kj . It follows that if D is a component
of C1 n Kj which does not contain 1, then D B(0; j) and so by (96), D
contains a point z0 2 C with dist(z0 ; C n U ) < 1=j. It follows from the de…nition
of distance that there exists w0 2 C n U C1 n Kj with jz0 w0 j < 1=j. Hence,
z0 2 B(w0 ; 1=j). But B(w0 ; 1=j) C1 n Kj . Indeed, let w 2 B(w0 ; 1=j). If
w 2 (C1 n B(0; j)) [ (B(0; j) n U ) there is nothing to prove, so assume that
w 2 U and jwj j. Since w0 2 C n U ,

dist(w; C n U ) jw w0 j < 1=j;

and so by the de…nition of Kj , w 2 = Kj .

Thus, z0 2 B(w0 ; 1=j) C1 nKj . Since D and B(w0 ; 1=j) are connected and
contain z0 , D [ B(w0 ; 1=j) is connected. But D is maximal, so B(w0 ; 1=j) D.
Let D1 be the component of C n U which contains w0 . Then D D1 again
because D C n U C1 n Kj and w0 2 D. This proves the claim.
Step 2: Let
Ij := fn 2 I : wn 2 Kj n Kj 1 g:
The sets Ij are disjoint and each Ij has only …nitely many elements, since E
has no accumulation points in U . De…ne
X
Qj := Sn ;
n2Ij

if Ij is nonempty and Qj = 0 otherwise. Then Qj is a rational functions with

poles in Kj n Kj 1 . By Runge’s theorem with E = C n U , there exists a rational
functions Rj with poles in C n U such that

jQj (z) Rj (z)j 1=2j for all z 2 Kj 1:

We claim that the function

1
X
f (z) = Q1 (z) + (Qj (z) Rj (z))
j=2

is well-de…ned and has all the desired property of the theorem. To see this let
we beging by showing that f is holomorphic in U n E. Note that since each wn
is isolated and don’t accumulate at points of U , U n E is open. Let K U n E
be a compact set. Then there exists m such that K Km . If j m + 1, then
K Kj 1 and so

jQj (z) Rj (z)j 1=2j for all z 2 K:

P1
It follows that the series j=m+1 (Qj (z) Rj (z)) is uniformly convergent in K.
Pm
Since Q1 (z) + j=2 (Qj (z) Rj (z)) have poles in E or in C n U , we have that
f is holomorphic in K .SBy considering an increasing sequence of compact sets
Tl , with Tl Tl+1 and j Tl = U n E, we have that f is holomorphic in U n E.

112
Wednesday, April 15, 2020
Proof. It remains to show that f has poles at each wn and that its principal
part is Qn . Since wn is isolated, there exists r > 0 such that jwn wj j > r for
all j 6= n. For z 2 U \ B(wn ; r) n fwn g we can write

f (z) = Sn (z) + f (z) Sn (z);

and the function f Sn is holomorphic in U \ B(wn ; r) since the poles of Rj

are in C n U for all j and Qj has poles in wj 2
= B(wn ; r) for all j 6= n. Thus, Sn
is the principal part of f at wn .

18 Simply Connected Domains

Using Runge’s theorem we can give another characterization of simply connected
sets. Given z 2 C and a Lipschitz continuous closed oriented curve with range
not containing z the winding number of around z is de…ned as
Z
1 d
ind (z) := : (97)
2 i z

It is also called the index of z with respect to .

Theorem 213 Let be a recti…able closed oriented curve in C with range .

Then

(i) for every z 2 C n , ind (z) is an integer,

(ii) if z; w belong to the same connected component of C n , then ind (z) =
ind (w),
(iii) ind (z) = 0 for all z in the unbounded connected component of C n .

Proof. (i) Fiz z 2 C n . Assume that is a polygonal path. Let ' : [0; 1] !
C be a parametrization of and consider the function
Z t
'0 (r)
g(t) := dr:
0 '(r) z
'0 (t)
Then g is absolutely continuous and g 0 (t) = '(t) z for L1 -a.e. t 2 [0; 1]. De…ne

g(t)
h(t) = ('(t) z)e :

By the chain rule,

h0 (t) = '0 (t)e g(t)

('(t) z)e g(t) 0
g (t)
'0 (t)
= '0 (t)e g(t)
('(t) z)e g(t)
=0
'(t) z

113
for L1 -a.e. t 2 [0; 1] and since h is absolutely continuous, it follows that h is
constant,say h 1c . Since '(0) = '(1) we get

1 = eg(0) = c('(0) z) = c('(1) z) = eg(1)

and so R d
1=e z ;
R d
which implies that z is a multiple of 2 i. Hence, ind (z) is an integer.
1
On the other hand, if is only recti…able, by Lemma 64 for every 0 < " < 2
there exists a polygonal path " with the same endpoints of such that

jind (z) ind " (z)j ":

Since ind " (z) is an integer, letting " ! 0+ we conclude that ind (z) is also an
integer.
(ii) Since the function ind : C n ! Z is continuous and it is integer-valued,
it must be constant in any connected component of C n .
(iii) Let C > 0 be such that j' (t) j C for all t 2 [0; 1]. Hence, for
jzj > R > C, we have that

j'(t) zj jzj j'(t)j jzj C > 0;

and so
'0 (t) M M
<
'(t) z j'(t) zj jzj C
provided R is su¢ ciently large. It follows that for jzj > R,
1
jind (z)j ;
2
and since ind takes only integer values, ind (z) = 0. The result now follows
from part (ii).
Another important application of Theorem ?? is the following.

Theorem 214 Let U C be an open set and let 1 and 2 be two continuous,
closed, oriented curves that are homotopic in U . Then

ind 1 (z) = ind 2 (z)

for all z 2 C n U . In particular, if U is simply connected, then ind (z) = 0

for every continuous closed oriented curve with range contained in U and for
every z 2 C n U .

Proof. Fix z0 2 C n U and let 1 and 2 be as in the statement. Since the

the function f (z) = z 1z0 is holomorphic in U , it follows by Theorem 98, that
R d
R d
1 z0 = 2 z0 , and so ind 1 (z0 ) = ind 2 (z0 ). On the other hand, if U is
simply connected, then every continuous closed oriented curve g1 is homotopic

114
to a point.
R But for a curve 2 with constant parametric representation we have
that 2 d z0 = 0, and so by the …rst part of the theorem, ind 1 (z0 ) = 0.
Given n closed continuous oriented curves 1 , . . . , n , the family :=
f 1 ; : : : ; n g is called a cycle. The range of is given by the union of the
ranges of 1 , . . . , n . Given a point z 2 C not contained in the range of , we
de…ne the winding number of around z to be the integer
n
X
ind (z) := ind k
(z) :
k=1

Theorem 215 Let U C be an open set and let K U be a compact set.

Then there exists a cycle with range contained in U n K such that
1 if z 2 K;
ind (z) =
0 if z 2 C n U:

Proof. Let 0 < < 21 dist(K; @U ) and consider a grid of squares of diameter
less than . Since K is compact, only …nitely many closed squares Q1 ; : : : ; Qn
intersect K. If z 2 Qj for some j, then dist(z; K) < . Hence, Qj U . Also if
Qj and Qk have a side S in common, then if we consider the closed curves @Qj
and @Qk oriented counterclockwise, then S will be traversed in both directions
and so the integrals of any continuous function over S + and S will cancel out.
Let S1 , . . . , Sn be the segments which are the sides of only one the rectangles.
Note that if one of these segments Sj intersects K then necessarily there must
be two rectangles which intersect K, which contradicts the de…nition of Sj . It
follows that Sk U n K.
If z 2 K, then there exists j 2 f1; : : : ; mg such that z 2 Rj . If z 2 Rj , then
Z
1 d
ind@Rj (z) = = 1;
2 i @Rj z
Z
1 d
ind@Rk (z) = = 0; k 6= j:
2 i @Rk z
Hence, summing these two identities
n
X
ind (z) = ind@Rk (z) :
k=1

If z belongs to @Rj , then either z is a vertex, in which case it belongs to four

S4
rectangles, say Rj1 , Rj2 , Rj3 ; Rj4 . Then, setting R = l=1 Rjl ,
4
X 4
X Z Z
1 d 1 d
ind@Rjl (z) = = =1
2 i @Rj z 2 i @R z
l=1 l=1 l

since all the integral along common edges cancel out. On the other hand,
Z
1 d
ind@Rk (z) = = 0; k 2 = fj1 ; j2 ; j3 ; j4 g:
2 i @Rk z

115
Hence, as before ind (z) = 1. Finally if z belongs to @Rj but it is not a
vertex, in which case it belongs to two rectangles, say Rj1 , Rj2 Then, setting
S2
R = l=1 Rjl , as before
2
X 2
X Z Z
1 d 1 d
ind@Rjl (z) = = = 1:
2 i @Rj z 2 i @R z
l=1 l=1 l

Also, On the other hand,

Z
1 d
ind@Rk (z) = = 0; k2
= fj1 ; j2 g:
2 i @Rk z

This shows that ind (z) = 1.

If z 2 C n U , then z 2
= Rk for any k and since z belongs to the unbounded
component of C n @Rk , ind@Rk (z) = 0 for all k, which shows that ind (z) = 0.
This completes the proof.
Friday, April 17, 2020

Theorem 216 Let U C be an open connected set. Then the following are
equivalent:

(i) C1 n U is connected,
(ii) U is simply connected,
(iii) ind (z) = 0 for every continuous closed oriented curve with range con-
tained in U and for every z 2 C n U .

Proof. Step 1: We prove that (i) implies (ii). Assume that C1 n U is

connected. Fix an holomorphic function f : U ! C and a recti…able closed
oriented curve with range contained in U . Taking E = f1g in Runge’s
theorem there exists a sequence of rational functions rn : C ! C with poles
in 1 such that rn ! f uniformly in . But this implies that these rational
functions are polynomials. Since each polynomial has a primitive, by Remark
??, Z
rn dz = 0:
R
Letting n ! 1 and using uniform convergence in , it follows that f ds = 0.
Thus (ii) holds. In view of Theorem 196 it follows that U is simply connected.
Step 2: That (ii) implies (iii) follows from Theorem 214.
Step 3: Assume that (iii) holds but that C1 n U is not connected. Since
C1 n U is closed, its connected components are also closed. Moreover, since
C1 is compact, so is any closed subset of C1 . Hence, we can …nd two disjoint
nonempty compact sets C and K (with respect to the metric in C1 ) such that

C1 n U = C [ K:

116
Moreover, since U C we have that 1 2 C1 n U , so 1 2 C [ K. Assume that
1 2 C. Then 1 2 = K and so K must be bounded, since otherwise we could
…nd a sequence fzn gn in K such that jzn j ! 1. This would imply that 1 is an
accumulation point of K and so it would belong to K since K is closed. Thus
K is compact in C.
Let V := C n C. Then V is open and contains K. By Theorem 215 there
exists a cycle with range contained in V n K such that

1 if z 2 K;
ind (z) =
0 if z 2 C n V:

But V n K = (C n C) n K = C n (C [ K) = C n (C n U ) = U . Hence, the

range of is contained in U but ind (z) = 1 for all z 2 K C n U , which
contradicts hypothesis (iii), since the winding number of each closed curve in
the cycle should be zero.

Remark 217 Note that saying that C1 n U is connected is not equivalent to

saying that C n U is connected. Indeed, consider the set E = fz = x + iy : y 2
(0; 1)g. Then its complement is not connected in C n U but it is connected in
C1 n U .

Corollary 218 Let U C be an open bounded connected set. Then U is con-

nected if and only if C n U is connected.

Exercise 219 Let U C be an open set. Prove that C1 n U is connected if

and only if every component of C n U is unbounded.

19 Proof of Caratheodory’s Theorem

Given an open set U C an oriented continuous half-open curve in U is an
equivalence class of continuous equivalent functions ' : [a; b) ! U . We de…ne
the length of as
L( ) := lim Var ':
r!b [a;r]

We say that the curve ends at b if there exists

lim '(t) = b 2 U :
t!b

Exercise 220 Let be an oriented continuous half-curve with range in some

open set U C. Prove that if has …nite length, then it ends at some point
b 2 U.

We begin with a preliminary result.

Lemma 221 Let V C be an open set and assume that f : V ! f (V ) be a

conformal map with f (V ) B(0; R) for some R > 0. If z0 2 C and

C(r) := V \ @B(z0 ; r);

117
then
2 R
infp L (f (C(r))) p ; 0< < 1:
<r< log(1= )
In particular, there exists rn & 0+ such that L (f (C(rn ))) ! 0 as n ! 1.
Proof. Let Dr := ft 2 [0; 2 ] : z0 + reit 2 V g and de…ne '(t) = z0 + reit ,
t 2 Dr . The set Dr is the union of disjoint intervals, Let I be one of these
intervals and consider [a; b] I. Then f ' : [a; b] ! C is a curve of class C 1
and so Z b
L(f ('([a; b])) = jf 0 ('(t))jj'0 (t)j dt:
a
Letting [a; b] % I if needed, we get
Z
L(f ('(I)) = jf 0 ('(t))jj'0 (t)j dt:
I

Summing over all disjoint intervals in Dr we obtain

Z
g(r) := L (f (C(r))) = L(f ('(Dr )) = jf 0 ('(t))jj'0 (t)j dt:
Dr

In turn, by Hölder’s inequality

Z 2 Z Z
(g(r))2 = jf 0 ('(t))jj'0 (t)j1=2 j'0 (t)j1=2 dt j'0 (t)j dt jf 0 ('(t))j2 j'0 (t)j dt
D Dr Dr
Z Z
2 r jf 0 ('(t))j2 j'0 (t)j dt = 2 r jf 0 (z0 + reit )j2 r dt:
Dr Dr

It follows that
Z 1 Z 1 Z
dr
(g(r))2 2 jf 0 (z0 + reit )j2 r dtdr
0 r 0 Dr
Z
=2 jf 0 (x + iy)j2 dxdy
U

where we used polar coordinates. Recalling that

!
@u @u
0 2 @x (x; y) @y (x; y)
jf (x + iy)j = det @v @v
@x (x; y) @y (x; y)

(see (10)), using the theorem on change of variables for Lebesgue (or Riemann)
integration we get
Z 1 Z
dr
(g(r))2 2 jf 0 (x + iy)j2 dxdy = 2 L2 (f (V )):
0 r V

Since f (V ) B(0; R) we obtain

Z p
1 1 2 dr
log infp (g(r)) (g(r))2 2 2
R2 :
2 <r< r

118
Dividing by log 1 proves the …rst part of the theorem, while to prove the second
part of the statement it su¢ ce to observe that log1 1 ! 0 as ! 0+ .

Exercise 222 Let U; V C be open sets and let f : U ! V be continuous,

1
one-to-one, onto, with f : V ! U continuous.

(i) Let fzn gn be a sequence of points in U such that zn ! z0 2 @U . Assume

that there exists
lim f (zn ) = w0 2 C:
n!1

Prove that w0 2 @V .
(ii) Assume that U = B(0; 1) and that f can be extended continuously to
B(0; 1). Prove that f (@U ) = @V .

119
Monday, April 20, 2020
Given a closed set C X, where X is a metric space and x; y 2 XnC. We say
that x; y are separated by C if they belong to di¤erent connected components of
X n C. We say that are not separated by C if they belong to the same connected
component of X n C.

Lemma 223 ( Janiszweski) Let C1 ; C2 C1 be two closed sets such that

C1 \ C2 is connected. If the points a; b 2 C1 n (C1 [ C2 ) are not separated by
either C1 or C2 , then they are not separated by C1 [ C2 .

Proof. Assume that a = 0 and b = 1 (the other cases are similar). Since
12 = Ck , we have that Ck is bounded, since otherwise we could …nd a sequence
fzn gn in Ck such that jzn j ! 1. This would imply that 1 is an accumulation
point of Ck and so it would belong to Ck since Ck is closed. Hence, Ck is
compact. Note that 0 and 1 belong to the same connected component U of
C1 nCk which is open and connected. Since Ck is bounded, with Ck B(0; Rk )
we have that the connected set C n B(0; Rk ) is contained in U . Thus, U n f1g
is open and connected in C and so pathwise connected. Thus we can …nd a
simple in…nite polygonal path k joining 0 with 1 (we can take it to be the
union of a half line and a simple polygonal path of …nite length). Since the
range of k is connected and C n k is connected, by Theorem 216, C n k is
simply connected and does not contain 0 and 1. Hence, by Theorem 100 we
can de…ne a branch fk of the logarithm in C n k . The connected set C1 \ C2
lies in one connected component F of C n ( 1 [ 2 ). If C1 \ C2 is empty we
take F to be any connected component of C n ( 1 [ 2 ). In the …rst case, by
adding a constant we can assume that f1 = f2 in F . Since the compact sets
C1 n F and C2 n F are disjoint, we can …nd disjoint open sets V1 and V2 such
that Ck n F Vk C n k , k = 1; 2. De…ne

fk (z) z 2 Vk ; k = 1; 2;
f (z) :=
f1 (z) = f2 (z) z 2 F:

Then f is holomorphic in the open set V := V1 [ V2 [ F which contains C1 [ C2

and ef (z) = z for all z 2 V .
Assume by contradiction that C1 [ C2 separates 0 and 1. Then the con-
nected component G of C1 n (C1 [ C2 ) which contains 0 is bounded. Note that
@G @(C1 [ C2 ) and since V contains C1 [ C2 we have that @V \ @G = ;. Let
0 < < 12 dist(@V; @G) and consider a grid of closed squares with diameter less
than and such that 0 lies in the interior of one of these squares, say 0 2 Q1 .
Note that Q1 is contained in G. Let Q1 ; : : : ; Qn be the closed squares contained
in G. Since @G V and 0 < < 21 dist(@V; @G), we have that the sides of
Q1 ; : : : ; Qn which are not counted twice are contained in V . Since f 0 (z) = z1 for
z 2 V , we have that
n
X n
X Z Z
1 d
ind (0) = ind@Qk (0) = = f 0 ( ) d = 0:
2 i @Qk @
k=1 k=1

120
On the other hand,
Z
1 d
ind@Q1 (0) = = 1;
2 i @Q1
Z
1 d
ind@Qk (0) = = 0; k 2:
2 i @Qk

Hence, summing these two identities ind (0) = 1, which gives a contradiction.

We turn to the proof of Theorem 200.

Proof. (i) =) (ii). Assume that f can be extended continuously to
B(0; 1) and still denote by f the extension. Then by the previous exercise,
f (@B(0; 1)) = @U . It follows that we can parametrize @U as

'(t) = f (eit ); t 2 [0; 2 ];

and so @U is the range of an oriented closed curve.

(ii) =) (iii) This implication follows from that fact that the range of a
continuous curve is locally connected.
(iii) =) (iv) Assume that @U is locally connected. For every " > 0 let
0 < < " be such that if z; w 2 @U with 0 < jz wj < there exists a compact
connected set F @U such that z; w 2 F and diam F < ". Let z; w 2 C n U
with jz wj < . If the closed segment [z; w] does not intersect @U , then we
take F = [z; w]. If [z; w] \ @U 6= ;, let z 0 and w0 be the …rst and last points
of [z; w] where [z; w] intersects @U . Since jz 0 w0 j < and z 0 ; w0 2 @U , there
exists a compact connected set F @U such that z 0 ; w0 2 F and diam F < ".
0 0
But then [z; z ] [ F [ [w ; w] is a compact connected set in C n U with diameter
less than 3" which contains z; w. Hence, C n U is locally connected.
Wednesday, April 22, 2020
Proof. (iv) =) (i) Assume that C n U is locally connected. Without loss
of generality we may assume that f (0) = 0. Since U is bounded, there exist
R0 < R such that
B(0; R0 ) U B(0; R): (98)
We claim that f is uniformly continuous in B(0; 1) n B(0; 1=2). Fix 0 < " < R0 .
Since C n U is locally connected we can …nd 0 < < " such that if z1 ; z2 2 C n U
with 0 < jz1 z2 j < there exists a compact connected set F C n U such that
z1 ; z2 2 F and diam F < ". Let 0 < < 1=4 be such that 2 R(log(1= )) 1=2 <
.
Let z; w 2 B(0; 1) n B(0; 1=2) with jz wj < . We claim that

jf (z) f (w)j < 2": (99)

Assume by contradiction that

jf (z) f (w)j 2":

121
p
By applying Lemma 221 with V = B(0; 1) and z0 = z we can …nd r 2 ( ; )
such that
L (f (C(r))) < < "; (100)
where C(r) := B(0; 1) \ @B(z; r). There are two cases. If B(z; r) B(0; 1).
Then C(r) = @B(z; r) and f (@B(z; r)) is the boundary of the simply connected
open set f (B(z; r)) which contains f (z) and f (w). Since jz wj < < r, we
have that z; w 2 B(z; ) B(z; r), and so f (z) with f (w) belong to the interior
of the closed curve f (@B(z; r)). Consider the segment S joining f (z) with f (w)
and extend it on both sides until it meets f (@B(z; r)). The resulting segment has
length bigger than 2", which contradicts the fact that L (f (C(r))) < < " (the
length is the supremum of the length of all polygonal paths made of segments
with endpoints on f (C(r))).
Assume next that B(z; r) \ @B(0; 1) 6= ;. In view of (100) and Exercise 220,
the continuous recti…able curve f (C(r)) has endpoints a and b 2 @U C n U.
In view of (100), jb aj L (f (C(r))) < , and so, since C n U is locally
connected there exists a compact connected set F C n U such that a; b 2 F
and diam F < ". Then F [ f (C(r)) is a connected set and
F [ f (C(r)) B(a; "): (101)
On the other hand, by (98), the fact that a 2 @U and " < R0 , we have that
02 = B(a; "). Since jf (z) f (w)j 2", it follows that either f (z) or f (w) does
not belong to B(a; "). Denote this point by c, so c 2 = B(a; "). Using the fact
that 0 2= B(a; ") in view of (101) we have that c and 0 are not separated by the
connected set F [ f (C(r)). On the other hand c 2 U and f (0) = 0 2 U and so
c and 0 are also not separated by C n U . Note that (F [ f (C(r))) \ (C n U ) = F ,
which is connected. Hence, by Janiszweski’s theorem c and 0 are not separated
by F [f (C(r))[(CnU ). Since the Cn(F [f (C(r))[(CnU )) = U n(F [f (C(r))
is open, its connected components are open, and so pathwise connected. Hence,
there exists a polygonal path in U n (F [ f (C(r)) which joins c and 0. Let
= [']. Since f is a conformal map, f 1 ' is a curve joining f 1 (c) 2 fz; wg
and 0. Moreover, its range its contained in B(0; 1) n C(r) = B(0; 1) n @B(z; r).
Since z; w 2 B(0; 1) n B(0; 1=2) with jz wj < < r, we have that z; w 2
p
B(z; ) B(z; r), while dist(0; B(0; 1) n B(0; 1=2)) = 12 > > r. Hence,
0 2= B(z; r). In turn, any curve joining 0 and either z or w would intersect
@B(z; r), and so we have a contradiction.
Let E C be a connected set and let z 2 E. We say that z is a cut point of
E is E n fzg is no longer connected. If we have a continuous simple arc, then
every point except the endpoints is a cut point. If we have a closed simple curve
then no point is a cut point.
Theorem 224 Let U C be an open bounded simply connected set and let f
map conformally B(0; 1) onto U . Assume that @U is a closed oriented curve
and denote by f the continuous extension of f to B(0; 1) given by Theorem 200.
Then z 2 @U is a cut point of @U if and only if the set f 1 (fzg) has more
than one element and the components of @U n fzg are f (Ik ), where Ik are the
components of @B(0; 1) n f 1 (fzg).

122
Friday, April 24, 2020
Proof. Let m = card f 1 (fzg) 2 N [ f1g. Since f : B(0; 1) ! U is
continuous, the set f 1 (fzg) is closed and so @B(0; 1) n f 1 (fzg) is relatively
open and thus it can be written as a countable union of disjoint open maximal
arcs Ik . In turn, we may write
m
! m
[ [
@U n fzg = f (@B(0; 1) n f 1 (fzg)) = f Ik = f (Ik ):
k=1 k=1

Since f is continuous and the sets Ik are connected we have that the sets f (Ik )
are connected. Note that if f 1 (fzg) is a singleton, then @U n fzg = f (I1 ),
which is connected, and so z is not a cut point of @U .
Conversely, assume that m 2. Then the endpoints a and b of I1 are
!
distinct. Consider the oriented closed segment ab and let '(t) = tb + (1 t)a,
t 2 [0; 1]. Consider the continuous curve parametrized by f '. Since f is
injective in B(0; 1) and f (a) = f (b) = z, we have that is a continuous simple
closed curve with range in U [ fzg. Let = f ('([0; 1)) be its range. By the
Jordan’s curve theorem, C n has two connected components Vb and Vu , with
Vb bounded and Vu unbounded, and with @Vb = @Vu = = f ('([0; 1)).
!
Note that B(0; 1) n (ab [ f 1 (fzg)) has two connected components E1 and
!
E2 . Since f is continuous and f (B(0; 1) n ab) C n = Vb [ Vu , and since f
maps connected sets into connected sets, we must have that f (E1 ) and f (E2 )
are contained in Vb or in Vu . But since f : B(0; 1) ! U is open, if we take
!
z0 2 ab n fa; bg, we can …nd a small ball B(z0 ; r) such that f (B(z0; r)) is open
and so there exists B(f (z0 ); ) f (B(z0 ; r)). Since f (z0 ) 2 = @Vb = @Vu ,
there must be points of B(z0 ; r) which end up in Vb and points which end up in
Vu . Thus fS(E1 ) and f (E2 ) are contained one in Vb and the other in Vu . Thus
m
f (I1 ) and k=2 f (Ik ) are not connected. In turn, z is a cut point of @U .
There are examples in which f 1 (fzg) has countably many elements.
We are now ready to prove Carathéodory’s theorem.
Proof. Let U C be an open bounded simply connected set and let f
map conformally B(0; 1) onto U . If f has a continuous and injective extension
to B(0; 1) then @U is parametrized by f (eit ), t 2 [0; 2 ], which is an oriented
simple closed curve. Conversely assume that @U is the range of an oriented
simple closed curve. In particular, @U is locally connected and it has no cut
points. Then by Theorem 200, f can be extended continuously to B(0; 1). By
the previous theorem the set f 1 (fzg) is a singleton for every z 2 @U , which
implies that f is injective on @B(0; 1). This concludes the proof.

Remark 225 Note that we actually proved that f has a continuous and injec-
tive extension to B(0; 1) if and only if @U is locally connected and it has no cut
points.

123
20 Elliptic Functions
We are interested in meromorphic functions f : C ! C1 which have two peri-
ods, that is, there exist !1 ; !2 2 C n f0g such that

f (z + !1 ) = f (z); f (z + !2 ) = f (z)

for all z 2 C. A function with these properties is called doubly periodic.

Exercise 226 Let f : C ! C1 be a doubly periodic meromorphic function with

periods !1 ; !2 2 C n f0g. Assume that := !1 =!2 2 R. Prove that f is either
periodic with simple period or constant.,

In view of the previous exercise, we can assume that Im 6= 0. Since and

1
have imaginary parts of opposite sign, by interchanging !1 and !2 , in what
follows we can assume that Im > 0.
Consider the function

g(z) := f (!1 z); z 2 C:

Then

g(z + 1) = f (!1 z + !1 ) = f (!1 z) = g(z);

g(z + ) = f (!1 z + !1 ) = f (!1 z + !2 ) = f (!1 z) = g(z):

Moreover, g is meromorphic if and only if f is and it has the same number of

zeros and of poles. Any other property of f can be deduced by the analogous
property of g. Thus, in what follows we assume that f has periods 1 and ,
where Im > 0. By induction we have that

f (z + j + k ) = f (z) for all z 2 C and j; k 2 Z: (102)

Consider the lattice

:= fj + k : j; k 2 Zg: (103)
We will show that partitions C into pairwise disjoint parallelograms congruent
to
P0 := fz 2 C : z = x + y ; 0 x < 1; 0 y < 1g: (104)
To be precise, [
C= (j + k + P0 ):
j;k2Z

We say that 1 and generate the lattice and we call P0 the fundamental
parallelogram of f .
We say that z; w 2 C are congruent modulo if

z =w+j+k

for some j; k 2 Z and we write z w. Note that z w2 .

124
Remark 227 If f : C ! C1 is be a doubly periodic meromorphic function with
periods !1 ; !2 2 C n f0g such that !1 =!2 2
= R, then we de…ne

P0 = fz 2 C : z = x!1 + y!2 ; 0 x < 1; 0 y < 1g

the fundamental parallelogram of f .

Theorem 228 Let f : C ! C1 be a doubly periodic meromorphic function

with periods 1 and , where Im > 0. Then

(i) every point in C is congruent modulo to a unique point in the funda-

mental parallelogram P0 ,
(ii) given j; k 2 Z, every point in C is congruent modulo to a unique point
in the parallelogram j + k + P0 ,
(iii) we have [
C= (j + k + P0 );
j;k2Z

where the interiors of the parallelograms are parwise disjoint,

(iv) the function f is completely determined by its values in P0 .

Proof. (i) Since the vectors 1 and form a basis over the reals of the two-
dimensional vector space C, given z 2 C, we can write z = x + y, for some
x; y 2 R. Let j; k 2 Z be such that j x < j + 1 and k y < k + 1. Then

w := z j k = (x j) + (y k)

is congruent to z modulo . Moreover, 0 x j < 1 and 0 y k < 1, and

so w 2 P0 .
To prove uniqueness, let w1 ; w2 2 P0 be congruent modulo . Then wl =
xl + yj , where 0 xl < 1 and 0 yl < 1, l = 1; 2. Since w1 w2 we have that

x1 + y1 x2 y2 = w 1 w2 = j + k

for some j; k 2 Z. But since 0 x1 ; x2 < 1, we have that 1 < x1 x2 < 1

and so j = x1 x2 = 0. Similarly, k = y1 y2 2 ( 1; 1) and so k = 0. Thus
w1 = w2 .
(ii) Let P := j0 + k0 + P0 , where j0 ; k0 2 Z. Given z 2 C by item (i)
there exists a unique w 2 P0 with z w. In turn, j0 + k0 + w 2 P and
z j0 + k0 + w. By the uniqueness in part (i), it follows that j0 + k0 + w is
the unique point in P which is congruent to z modulo .
(iii) By part (i) each z 2 C is congruent to some w 2 P0 modulo , which
means that z = j + k + w for some w 2 P0 . Hence, z 2 j + k + P0 .
On the other hand, if P1 = j1 + k1 + P0 and P2 = j2 + k2 + P0 , and
z 2 P1 \ P2 , then
z = j1 + k1 + w1 = j2 + k2 + w2

125
with w1 ; w2 2 P0 . This means that z w1 and z w2 . Again by the uniqueness
in item (i), w1 = w2 . In turn, j1 + k1 = j2 + k2 , which implies that j1 = j2
and k1 = k2 .
(iv) In view of (102),

f (z) = f (w) if z w:

The result now follows from item (i).

Next we show why we are taking meromorphic functions instead of holomor-
phic functions.

Corollary 229 Let f : C ! C be holomorphic and doubly periodic with periods

1 and , where Im > 0. Then f is constant.

Proof. Let M := maxP0 jf j. By item (iv) of the previous theorem for every
z 2 C there exists w 2 P0 such that f (z) = f (w). Hence, jf (z)j = jf (w)j M .
It follows by Liouville’s theorem that f is constant.

De…nition 230 An elliptic function is a meromorphic function which is doubly

periodic with periods w1 ; w2 2 C n f0g such that w1 =w2 2
= R.

We begin by showing that an elliptic function must have more than one pole.

Theorem 231 Let f : C ! C1 be an elliptic function. Then f must have at

least two poles.

Proof. Without loss of generality we may assume that the periods are 1
and with Im > 0.
Step 1: Assume that f has no poles on @P0 . Then by the residue theorem
Z Xn
f dz = 2 i reszk f;
@P0 k=1

where z1 ; : : : ; zn are the poles of f inside P0 . Note that there must be at least
one in view of the previous two theorems. With a slight abuse of notation we
write Z Z 1 Z 1+ Z Z 0
f dz = f dz + f dz + f dz + f dz:
@P0 0 1 1+

Note that by (102),

Z 1 Z Z 1 Z 0
f dz + f dz = f dz + f ( + w) dw
0 1+ 0 1
Z 1 Z 0 Z 1 Z 1
= f dz + f (w) dw = f dz f (z) dz = 0;
0 1 0 0

and similarly, Z Z
1+ 0
f dz + f dz = 0:
1

126
Hence,
Z n
X
0= f dz = 2 i reszk f:
@P0 k=1

If n were 1, we would have 0 = resz1 f , which would impliy that f has a remov-
able singularity at z1 by Theorems and . This would contradict the previous
corollary. Hence, n 2.

127
Monday, April 27, 2020
Proof. Step 2: Since poles do no accumulate in the interior, it follows that
f has a …nite number of poles in P0 . Hence, if B(0; R) contains P0 then by
periodicity f has a …nite number of poles in B(0; R). In turn, for " > 0 the
function f" (z) := f (z + "(1 + )) has no poles on @P0 . By the previous step we
…nd that f" has at least two poles in P0 for every " small. Letting " ! 0 we
conclude that f has at least two poles.
The number of poles of an elliptic function in its fundamental parallelogram
counted with their multiplicity is called its order. Next we show that the number
of zeros of an elliptic function equals the number of poles.

Theorem 232 Let f : C ! C1 be an elliptic function of order `. Then f has

` zeros in its fundamental parallelogram counted with their multiplicity.

Proof. Without loss of generality we may assume that the periods are 1
and with Im > 0. Since zeros and poles do no accumulate in the interior, it
follows that f has a …nite number of poles and zeros in P0 .
Step 1: Assume that f has no poles and no zeros on @P0 . By the argument
principle,
Z
1 f0
dz = (number of zeros of f in P0 ) minus (number of poles of f in P0 )
2 i @P0 f
=: nz `:
f0
Since f is doubly periodic with periods 1 and , reasoning as in the previous
R 0
theorem, we can show that 21 i @P0 ff dz = 0. Hence, nz = `.
Step 2: Since poles and zeros do no accumulate in the interior, it follows
that f has a …nite number of poles and zeros in P0 . Hence, if B(0; R) contains
P0 then by periodicity f has a …nite number of poles and zeros in B(0; R). In
turn, for " > 0 the function f" (z) := f (z + "(1 + )) has no poles or zeros on
@P0 . By previous step we …nd that the number of zeros of f" in P0 is the same
as the number of poles of f" in P0 for every " small. Letting " ! 0 we conclude
that the number of zeros of f in P0 is the same as the number of poles of f in
P0 .
The next natural question is the existence of elliptic functions. We will
construct an elliptic function of order two. The idea is to consider the function
X 1
(z + !)2
!2

but the problem is that this double series does not converge absolutely. Indeed
we will see below that for a double series to converge we need the exponent to
be bigger than 2. To …x this problem, we follow the approach in your homework
for cot and we de…ne the function
1 X 1 1
}(z) := 2 + ;
z (z + !)2 !2
!2

128
where := n f0g. This function is called Weierstrass } function. Note that

1 1 ! 2 z 2 2z! ! 2 z 2 2z! 2z
= = 2
(z + !)2 ! 2 2
! (z + !)2 ! (z + !)2 !3

as j!j ! 1.

Theorem 233 The Weierstrass } function is an elliptic function of order two.

We begin with a preliminary result.

Lemma 234 The double series

X 1 X 1
;
(jjj + jkj)r jj + k jr
(j;k)2Z2 nf(0;0)g j+k 2

converge if and only if r > 2.

Proof. Step 1: Assume that r > 2. For every j 6= 0 we have

X 1 1 X 1 1 X 1
r
= r + r
= r +2
(jjj + jkj) jjj (jjj + jkj) jjj (jjj + k)r
k2Z k2Znf0g k2N
X 1 Z 1
1 1 dx 1 2 1
= r +2 + 2 = r + :
jjj nr jjjr jjj x
r jjj r 1 jjjr 1
n=jjj+1

Hence,
X 1 X 1 X X 1
= +
(jjj + jkj)r (0 + jkj)r (jjj + jkj)r
(j;k)2Z2 nf(0;0)g k2Znf0g j2Znf0g k2Z
X 1 X 1 2 1
2 + + <1
kr jjjr r 1 jjjr 1
k2N j2Znf0g

since r > 2.
To prove that the second series converges, it su¢ ces to show that there exists
a constant c > 0 such that

jj + k j c(jjj + jkj)

for all (j; k) 2 Z2 n f(0; 0)g. Write = x + iy, where x 2 R and y > 0. Then
p 1
jj + k j = (j + kx)2 + k 2 y 2 (jj + kxj + jkyj):
2
If x = 0, then
1 minf1; yg
(jjj + jkyj) (jjj + jkj):
2 2

129
Assume that x 6= 0. If jjj 2jkxj, then

1 jyj 1 1 jyj 1
jj + kxj + jkyj jkyj = jkxj + jkyj jjj + jkyj
2 jxj 2 4 jxj 2
jyj minf1=jxj; 1g
(jjj + jkj):
4
If jjj 2jkxj, then

1 minf1; yg
jj + kxj + jkyj jjj jkxj + jkyj jjj + jkyj (jjj + jkj):
2 2
This concludes the proof of the case r > 2.
1 1
Step 2: Assume that r 2. If 1 k j then j + k 2j and so j+k 2j .
Then

X X j
1 X X j
1 X X j1
1 1 1
= = 1:
(jjj + jkj)r j=1 k=1
(j + k)r j=1 k=1
(2j)r j=1
(2j)r
(j;k)2Z2 nf(0;0)g

To prove that the second series diverges, it su¢ ces to show that there exists a
constant c > 0 such that

jj + k j c(jjj + jkj)

for all (j; k) 2 Z2 n f(0; 0)g. We have

jj + k j jjj + jk j = jjj + jkjj j maxf1; j jg(jjj + jkj);

which concludes the proof.

We turn to the proof of Weierstrass theorem.
Proof. Let R > 0 and let jzj < R. Write

1 X 1 1 X 1 1
}(z) = + +
z2 (z + !)2 !2 (z + !)2 !2
j!j 2R j!j>2R

=: I + II + III:

To estimate III observe that for jzj < R and j!j > 2R,
1 1
jz + !j j!j jzj j!j + R jzj j!j;
2 2
and so
1 1 z 2 2z! R2 + 2Rj!j
= 2 2
(z + !)2 ! 2 ! (z + !)2 j!j4
4Rj!j 1
2 = 8R 3 :
j!j4 j!j

130
Hence X 1
jIIIj 8R
(jj + k j)3
j+k 2

which converges by the previous lemma.

The term II is a …nite sum and so it is a meromorphic function in B(0; R)
with double poles at those ! 2 inside B(0; R).
This shows that } is well-de…ned and meromorphic with double poles at each
point of the lattice . To prove that } is doubly periodic with periods 1 and
we compute the derivative of }. We have
2 X 2 X 2
}0 (z) = 3 3
= :
z (z + !) (z + !)3
!2 !2

Note that by the previous lemma the series converges absolutely whenever z 2
= .
Let’s prove that }0 has periods 1 and . Since ! +1 2 and ! + 2 whenever
! 2 , we have
X 2 X 2
}0 (z + 1) = 3
= = }0 (z);
(z + 1 + !) (z + )3
!2 2
X 2 X 2
}0 (z + ) = = = }0 (z):
(z + + !)3 (z + )3
!2 2

Hence, there exist a; b 2 C such that

}(z + 1) = }(z) + a; }(z + ) = }(z) + b (105)
for all z 2 C n .
Using the fact that ! 2 if and only if !2 we have that
1 X 1 1
}( z) = 2
+
( z) ( z + !)2 !2
!2
1 X 1 1
= + = }(z):
z2 (z !)2 ( !)2
!2

This shows that } is even. Taking z = 21 and z = 2 in (105) gives a = 0 and

b = 0. We have proved that } is doubly periodic with periods 1 and . Since
the only element of inside the fundamental parallelogram is 0, } has order 2.

Wednesday, April 29, 2020

Next we show some important properties of the function }.
Theorem 235 The function } satis…es the equality
(}0 (z))2 = 4(}(z) e1 )(}(z) e2 )(}(z) e3 );
where
e1 := }(1=2); e2 := }( =2); e3 := }((1 + )=2): (106)

131
Proof. Since } is even, }0 is odd, and so using also the fact that }0 is
periodic of period 1,
}0 (1=2) = }0 ( 1=2) = }0 ( 1=2 + 1) = }0 (1=2);
which implies that }0 (1=2) = 0. Similalrly,
}0 ( =2) = }0 ( =2) = }0 ( =2 + ) = }0 ( =2);
and so }0 ( =2) = 0. Finally,
}0 ((1 + )=2) = }0 ( (1 + )=2) = }0 ( (1 + )=2 + 1 + ) = }0 ((1 + )=2);
which implies that }0 ((1 + )=2). Since }0 is an elliptic function of order 3, it
follows from Theorem 232, that it has three zeros in the fundamental parallel-
ogram P0 (already counted with their multiplicity). Hence, 12 , 2 , and 1+2 are
simple zeros of }0 and they are the only ones in P0 .
Since the function } e1 is elliptic of order two, and it has a double zero
at 21 (since its derivative has a simple zero), it follows from Theorem 232 that
} e1 has no other zeros in P0 . Similarly, } e2 and } e3 have a double zero
at 2 and 1+2 , respectively, and no other zeros in P0 .
Consider the function
g(z) = (}(z) e1 )(}(z) e2 )(}(z) e3 ):
The only zeros of g in P0 are at 21 , 2 , and 1+2 and they have multiplicity 2.
Moreover, g is an elliptic function of with poles at . Since 0 is the only pole in
P0 , it has multiplicity 6 by Theorem 232. Thus, every pole in has multiplicity
6.
On the other hand, since }0 has poles of multiplicity 3 at , (}0 )2 has poles of
multiplicity 6 at . Also, by what we did before it only has zeros of multiplicity
2 at 21 , 2 , and 1+2 . Thus, if we consider the function (}0 )2 =g, we have that
it has removable singularities at each point of and at 21 , 2 , and 1+2 (and
0 2
their periodic translates). Hence, (} ) =g can be extended to an entire function.
Since it is doubly periodic with periods 1 and , by Corollary 229, (}0 )2 =g is
constant.
We have seen in the proof of Theorem 233 that if we htake R > 0i so
P 1 1
small that B(0; 2R) \ = f0g, then the function j!j>2R (z+!)2 !2 =
P h i
1 1
!2 (z+!)2 ! 2 is holomorphic in B(0; R). Hence,

lim z 2 }(z) = 1:
z!0

Similarly,
lim z 3 }0 (z) = 2:
z!0
It follows that
z 6 (}0 )2 4
c = lim = :
z!0 z 6 g(z) 1
This completes the proof.

132
Remark 236 The numbers 12 , 2 , and 1+2 are called half-periods. It follows
from the previous proof that }0 restricted to P0 has three simple zeros at the
half-periods and no other zeros. Hence, for every a 2 P0 , the function } }(a)
has a double zero at a if a is a half-period and otherwise a simple zero at a and
a since } is even.

We now demonstrate the importance of the function }.

Theorem 237 Every elliptic function f : C ! C1 with periods 1 and , where

Im > 0, is a rational function of } and }0 .

Proof. We want to construct a doubly periodic elliptic function g using }

which has the same zeros and poles of f .
Step 1: Assume that f is even. Then if f has a zero or a pole at some
a 2 P0 n f0g, then it also has a zero or a pole at a. Note that a is congruent
to a modulo if and only if a is a half-period. Indeed, if

a= a+j+k

for some j; k 2 Z, then a = 21 j+ 12 k 2 P0 , which can happen only if j; k 2 f0; 1g.

Substep 1: Assume that f has no zeros or poles at the origin and at the
half-periods. We recall that by Theorem 232, if f has order `, then it has `
zeros. Let a1 ; : : : ; a` 2 P0 n f0g be the zeros of f in P0 counted with their
multiplicity and let b1 ; : : : ; b` 2 P0 n f0g be the poles of f in P0 counted with
their multiplicity. We claim that

Ỳ }(z) }(an )
f (z) = f (0) :
n=1
}(z) }(bn )

Indeed, let g denote the function on the right-hand side. In view of the previous
1
remark } }(an ) has a simple zero at an while the function } }(b n)
has a
simple pole at bn . Thus, the function g has the same zeros and poles in P0 as
f . It follows that f =g has removable singularities at an and at bn , n = 1; : : : ; `.
Thus f =g can be extended to a doubly periodic entire function, and so it must
be constant in view of Corollary 229. Using the fact that

lim z 2 }(z) = 1;
z!0

we obtain that the constant must be f (0).

Substep 2: If f has a zero at at a half-period a, then the zero must have
even multiplicity. Indeed f (2n+1) is odd and we can reason as in the proof of
Theorem 235 to show that f (2n+1) vanishes at all the half-periods. Similarly,
if f has a pole at a half-period a, then f1 is still an even elliptic function with
the same periods and so the pole must have even multiplicity. Recalling that
} }(a) has a double zero if a is a half-period and a pole of multiplicity two
at the origin we can …nd integers k0 ; : : : ; k3 2 Z such that }k0 behaves like f
near z = 0 and (}(z) ej )kj , j = 1; 2; 3, behaves like f near 21 , 2 , and 1+2 ,

133
respectively. Let a1 ; : : : ; an 2 P0 n f0; 12 ; 2 ; 1+2 g be the other zeros of f in P0
counted with their multiplicity and let b1 ; : : : ; bm 2 P0 n f0; 21 ; 2 ; 1+2 g be the
other poles of f in P0 counted with their multiplicity. Consider the function
3
Y n
Y m
Y 1
g(z) := }k0 (z) (}(z) ej )kj (}(z) }(aj )) ;
j=1 j=1 j=1
}(z) }(bj )

where
2(k0 + k1 + k2 + k2 ) + n m=0
by Theorem 232. The function g has the same zeros and poles in P0 as f . It
follows that f =g has removable singularities at 0, 0; 12 ; 2 ; 1+2 , and at all the aj
and bk . Thus f =g can be extended to a doubly periodic entire function, and so
it must be constant in view of Corollary 229.
Step 2: If f is odd, then f =}0 is an even elliptic function and so by the
previous step it can be written as a rational function of }. Finally, in the general
case we can write f as the sum of an even function and an odd function, to be
precise,
1 1
f (z) = [f (z) + f ( z)] + [f (z) f ( z)]:
2 2
This concludes the proof.
Friday, May 1, 2020
No class.

134