You are on page 1of 52

Figure 0.

1
COMPLEX MADE SIMPLE
DAVID C. ULLRICH
Solutions provided by Scott Larson.
Contents
Part 1. Complex Made Simple 3
0. Dierentiability and the Cauchy-Riemann Equations 3
1. Power Series 6
2. Preliminary Results on Holomorphic Functions 8
3. Elementary Results on Holomorphic Functions 13
4. Logarithms, Winding Numbers and Cauchys Theorem 19
5. Counting Zeroes and the Open Mapping Theorem 30
6. Eulers Formula for sin(z) 33
6.0. Motivation 33
6.1. Proof by the Residue Theorem 33
6.2. Estimating Sums Using Integrals 34
6.3. Proof Using Liouvilles Theorem 35
7. Inverses of Holomorphic Maps 36
8. Conformal Mappings 38
8.0. Meromorphic Functions and the Riemann Sphere 38
8.1. Linear-Fractional Transformations, Part I 38
8.2. Linear-Fractional Transformations, Part II 39
8.3. Linear-Fractional Transformations, Part III 39
8.4. Linear-Fractional Transformations, Part IV:
The Schwarz Lemma and Automorphisms of the Disk 39
8.5. More on the Schwarz Lemma 40
9. Normal Families and the Riemann Mapping Theorem 41
9.0. Introduction 41
9.1. Quasi-Metrics 41
9.2. Convergence and Compactness in C(D) 42
1
2 DAVID C. ULLRICH
9.3. Montels Theorem 43
9.4. The Riemann Mapping Theorem 44
9.5. Montels Theorem Again 44
10. Harmonic Functions 44
10.0. Introduction 44
10.1. Poisson Integrals and the Dirichlet Problem 45
10.2. Poisson Integrals and Aut(D) 46
10.3. Poisson Integrals and Cauchy Integrals 47
10.4. Series Representations for Harmonic Functions in the Disk 47
10.5. Greens Functions and Conformal Mappings 47
10.6. Intermission: Harmonic Functions and Brownian Motion 47
11.7. The Schwarz Reection Principle and Harnacks Theorem 47
12. Runges Theorem and the Mittag-Leer Theorem 47
13. The Weierstrass Factorization Theorem 50
16. Analytic Continuation 52
16.0. Introduction 52
16.1. Continuation Along Curves 52
3
Part 1. Complex Made Simple
0. Differentiability and the Cauchy-Riemann Equations
Proposition 0.0. Suppose that V is an open subset of C, that f : V C, and z V . Then f is
complex-dierentiable at z if and only if it is real-dierentiable at z and the real derivative Df(z) is
complex-linear.
Proposition 0.1. Suppose that V is an open subset of C, that f : V C, and z V . Then f = u+iv
is complex-dierentiable at z if and only if it is real-dierentiable at z and the real and imaginary parts
satisfy the Cauchy-Riemann equations
u
x
(z) = v
y
(z), u
y
(z) = v
x
(z).
Corollary 0.2. Suppose that V is an open subset of C, that f : V C, and z V ; write f = u +iv.
If f is complex-dierentiable at z V then u and v satisfy the Cauchy-Riemann equations at z.
Corollary 0.3. Suppose that V is an open subset of C, that f : V C, and z V ; write f = u+iv. If
the rst-order partial derivatives of u and v are continuous at z V and satisfy the Cauchy-Riemann
equations there then f is complex-dierentiable at z.
0.1. Suppose that f is dened in a neighborhood of the complex number z. Show that f is complex-
dierentiable at z if and only if there exists a complex number a such that f(z +h) = f(z) +ah+o(h)
as h 0.
Proof. Suppose that f is complex-dierentiable at z so
f

(z) = lim
h0
f(z +h) f(z)
h
.
Now let E(h) be a function such that f(z + h) = f(z) + f

(z)h + E(h) for all h in the neighborhood


of z for which f is dened. But E(h) satises E(0) = 0 and lim
h0
|E(h)|
|h|
= 0, so f

(z) = a.
Now suppose that there exists a complex number a such that f(z +h) = f(z) +ah+o(h) as h 0.
Then
a = lim
h0
f(z +h) f(z)
h

o(h)
h
= f

(z).

0.2. Suppose that f is complex-dierentiable at z. Show that f is continuous at z.


Proof. Since f is complex-dierentiable at z,
f

(z) = lim
h0
f(z +h) f(z)
h
.
Since this limit exists, lim
h0
f(z +h) = f(z) so f is continuous at z.
0.3. Suppose that T : C C is R-linear. Show that T is C-linear if and only if T(iz) = iTz for all
z C.
Proof. Suppose that T is C-linear so that for every c, z C we have
T(cz) = cTz.
Since i, z C we have
T(iz) = iTz.
Now suppose that T(iz) = iTz for all z C. Since T is R-linear, for every z, w C and c R we
have
T(z +w) = Tz +Tw, T(cz) = cTz.
4 DAVID C. ULLRICH
But if c = x +iy for x, y R, we have
T(cz) = T((x +iy)z) = T(xz) +T(iyz) = xTz +iT(yz)
= xTz +iyTz = (x +iy)Tz = cTz.
Thus T is C-linear.
0.4. Suppose that T : R
2
R
2
is the R-linear mapping dened by the matrix

a b
c d

(that is,
T(x, y) = (ax +by, cx +dy)). Show that T is C-linear if and only if a = d and b = c. Hint: Use the
fact that i(x, y) = (y, x).
Proof. Suppose that T is C-linear so that T(iz) = iTz for all z C. Let z = (x, y) so we have
(ay +bx, cy +dx) = T((y, x)) = T(iz) = iTz = i(ax +by, cx +dy)
= (cx dy, ax +by).
Thus b = c and a = d.
Now suppose that b = c and a = d. Let z = (x, y) for some z C so that
T(iz) = T((y, x)) = (ay +bx, cy +dx) = (cx dy, ax +by) = i(ax +by, cx +dy)
= iTz.
Therefore by Exercise 0.3, T is C-linear.
0.5. Suppose that f and g are complex-dierentiable at z.
(i) Show that f +g and fg are complex-dierentiable at z and that their derivatives are given by
the same formula as in elementary calculus.
(ii) Suppose in addition that g(z) = 0 and show that f/g is dierentiable at z with derivative
given by the same formula as in elementary calculus.
(iii) Suppose that n is a positive integer and dene f(z) = z
n
. Show that f is dierentiable and
f

(z) = nz
n1
.
(iv) Same as part (iii), but for negative integers n, with the restriction z = 0.
Proof (i).
(f +g)

(z) = lim
h0
(f +g)(z +h) (f +g)(z)
h
= lim
h0
f(z +h) f(z)
h
+ lim
h0
g(z +h) g(z)
h
= f

(z) +g

(z)
(fg)

(z) = lim
h0
(fg)(z +h) (fg)(z)
h
=
= f(z) lim
h0
g(z +h) g(z)
h
+g(z) lim
h0
f(z +h) f(z)
h
= (fg

)(z) + (f

g)(z)

Proof (ii).
(f/g)

(z) = lim
h0
(f/g)(z +h) (f/g)(z)
h
=
=
g(z)f

(z) f(z)g

(z)
g
2
(z)
5

Proof (iii).
f

(z) = lim
h0
f(z +h) f(z)
h
= lim
h0
(z +h)
n
z
n
h
= lim
h0

n
i=0

n
i

z
ni
h
i
z
n
h
= lim
h0
n

i=1

n
i

z
ni
h
i1
= nz
n1
.

Proof (iv).
f

(z) = lim
h0
f(z +h) f(z)
h
= lim
h0
(z +h)
n
z
n
h
=

0.6. Suppose that f is dierentiable at z and that g is dierentiable at f(z). Show that the composition
g f is dierentiable at z, with
(g f)

(z) = g

(f(z))f

(z).
Proof.
0.7. Dene f : C C by
f(x +iy) =

0 (x = 0),
0 (y = 0),
1 (otherwise).
Show that f satises the Cauchy-Riemann equations at the origin although f is not complex dieren-
tiable at the origin.
Proof. We see that f(C) R so u((x, y)) = f((x, y)) and v((x, y)) = 0. Calculating the partial
derivatives of u and v at the origin gives us
u
x
(0) = lim
h0
u((h, 0)) u(0)
h
= 0,
u
y
(0) = lim
h0
u((0, h)) u(0)
h
= 0,
v
x
(0) = 0,
v
y
(0) = 0.
Thus the Cauchy-Riemann equations hold trivially. However, f is not continuous at the origin and
therefore must not be complex dierentiable at the origin either.
0.8. Dene f : C C by
f(z) =

|z|
2
sin(1/ |z|) (z = 0),
0 (z = 0).
Show that f is complex dierentiable at the origin although the partial derivative u
x
is not continuous
at the origin.
6 DAVID C. ULLRICH
Proof. We begin with the denition of complex dierentiable.
f

(0) = lim
h0
f(h) f(0)
h
= lim
h0
|h|
2
sin(1/ |h|)
h
= lim
h0
|h|
2
sin(1/ |h|)
|h| e
it
= lim
h0
e
it
|h| sin(1/ |h|)

e
it
hsin(1/ |h|)

|h|

= 0
Hence f is complex dierentiable at the origin.
Now we show that u
x
is not continuous at the origin. Since f(C) R, we have u((x, y)) = f(z). So
now we calculate u
x
((x, y)) for (x, y) = (0, 0), lim
x0
u
x
((x, 0)), and u
x
((0, 0)).
u
x
((x, y)) =

x

(x
2
+y
2
) sin

(x
2
+y
2
)
1/2

= x(x
2
+y
2
)
1/2
cos

(x
2
+y
2
)
1/2

+ 2xsin

(x
2
+y
2
)
1/2

lim
x0
u
x
((x, 0)) = lim
x0
cos(1/x) + 2xsin(1/x) (DNE)
u
x
((0, 0)) = lim
h0
h
2
sin(1/h)
h
= lim
h0
hsin(1/h) = 0 (h R)
Since u
x
((0, 0)) = lim
x0
u
x
((x, 0)), this shows that u
x
is not continuous at the origin.
1. Power Series
Lemma 1.1. Suppose (c
n
)

n=0
is a sequence of complex numbers, and dene R [0, ] by
R = sup {r 0 | (c
n
r
n
) is bounded} .
Then the power series

n=0
c
n
(z z
0
)
n
converges absolutely and uniformly on every compact subset
of the disk D(z
0
, R) and diverges at every point z with |z z
0
| > R.
Proposition 1.2. Suppose that the power series

n=0
c
n
(z z
0
)
n
has radius of convergence R > 0.
Then the function f(z) =

n=0
c
n
(z z
0
)
n
is dierentiable in the disk D(z
0
, R), with derivative
f

(z) =

n=1
nc
n
(z z
0
)
n1
=

n=0
(n + 1)c
n+1
(z z
0
)
n
.
Corollary 1.3. If the power series

n=0
c
n
(z z
0
)
n
has positive radius of convergence R then the
function
f(z) =

n=0
c
n
(z z
0
)
n
is in fact innitely dierentiable in the disk D(z
0
, R); further, the doecients are given by
c
n
=
f
(n)
(z
0
)
n!
.
Corollary 1.4. If the power series

n=0
c
n
(z z
0
)
n
has positive radius of convergence R and there
exists r (0, R) such that

n=0
c
n
(z z
0
)
n
= 0 for all z D(z
0
, r) then c
n
= 0 for all n.
1.1. Suppose a [0, 1]. Show that there exist sequences x
n
> 0 and y
n
> 0 such that lim
n
x
n
= 0,
lim
n
y
n
= 0, and
lim
n
x
yn
n
= a.
7
Proof. First we show that if 0 < x < 1 then there exists y [0, 1] such that x
y
= . Let
0 < x < 1. Then we let y = log / log x and notice that the given inequality shows that this is
well dened and y [0, 1]. But
x
y
= x
log / log x
= x
log
x

= .
Next we let a = 0. If we let x
n
=
1
n
n
> 0 and y
n
=
1
n
> 0, we observe that lim
n
x
n
=
lim
n
y
n
= 0. But now we see
lim
n
x
yn
n
= lim
n

1
n
n

1/n
= lim
n
1
n
= 0.
Thus this is true for a = 0.
Now we let a (0, 1]. So let x
n
= 1/n > 0 so that lim
n
x
n
= 0. Now let > 0 be given. Observe
that there exists N N such that
0 <
1
N
< a 1.
Thus there exists
1
R such that 0 <
1
< and
0 <
1
N
< a
1
< a 1.
By the rst step of the proof, there exists y(N) [0, 1] such that

1
N

y(N)
= a
1
.
Notice it is clear that y(N) 0 as N and since a
1
< 1, y(N) (0, 1]. Now we dene our
sequence, y
n
, as follows:
y
n
=

y(N), n > N,
1, n N.
Thus we have constructed a sequence y
n
> 0 such that lim
n
y
n
= 0 with the string of inequalities
such that for any n > N,
|x
yn
n
a| =

1
n

yn
a

= |a
1
a|
=
1
< .
Therefore for any a [0, 1] there exist sequences x
n
> 0 and y
n
> 0 such that lim
n
x
n
= lim
n
=
0 and
lim
n
x
yn
n
= a.

1.2. Suppose the power series

n=0
c
n
z
n
has radius of convergence R > 0, and dene f : D(0, R) C
by
f(z) =

n=0
c
n
z
n
.
8 DAVID C. ULLRICH
We have shown that
f

(z) =

n=1
nc
n
z
n1
(z D(0, R));
explain why it would not be quite correct to write
f

(z) =

n=0
nc
n
z
n1
(z D(0, R))
instead, although it is perfectly correct to say f

n=0
f

n
, where f
n
(z) = c
n
z
n
.
Solution. Since 0 D(0, R), this formula is valid for f

(0). Expanding the sum we see that this formula


would be dividing by zero.
f

(0) = 0 c
0

1
0
+
However,
f

0
(0) = lim
h0
c
0
h
0
0 0
0
h
= 0.
So f

0
= 0 for all z D(0, R) so that it is perfectly correct to use the formula
f

n=0
f

n
,
where f
n
(z) = c
n
z
n
.
2. Preliminary Results on Holomorphic Functions
2.1. Dene f : C C by f(z) = z
2
sin(1/z) for z = 0, f(0) = 0. The last few paragraphs almost
seem to give a proof that f is dierentiable everywhere but f

is not continuous at the origin. This is


impossible where does the proof fail?
Proof. To see where the proof for f C(R) fails, let us look at the derivative of f at the origin and
notice that h C.
f

(0) = lim
h0
f(h)
h
= lim
h0
h
2
sin(1/h)
h
= lim
h0
hsin(1/h)
The proof for f C(R) says that |hsin(1/h)| |h| because |sin(x)| 1 for x R. But this is not true
for |sin(z)| for z C. For example,
|sin(3i)| =

exp(3) exp(3)
2i

(1 + 3) 1/(1 + 3)
2i

= |(15/8)i|
= 15/8.

Lemma 2.0 (Cauchys Theorem for Derivatives). Suppose V is an open subset of the plane, f : V C
is continuous, and there exists an F : V C such that f = F

in V . Then

f(z) dz = 0
9
for any smooth closed curve in V .
Lemma 2.1 (ML Inequality). Suppose that : [a, b] C is a smooth curve and f is continuous on

. If |f| M on

and L is the length of then

f(z) dz

ML.
2.2.
(i) Show that [b, a] =

[a, b].
(ii) Generalize this: Suppose that
1
is a smooth curve in the plane. How can you dene another
smooth curve
2
so that
2
=

1
?
Proof (i). By denition, [b, a] is the path
1
: [0, 1] C dened by

1
(t) = (1 t)b +ta,
and [a, b] is the path
2
: [0, 1] C dened by

2
(t) = (1 t)a +tb.
So calculating with the denition of
1
=

2
, we get

[b,a]
f(z) dz =

1
0
f

1
(t
1
)

1
(t
1
) dt
1
(Let t
2
= 1 t
1
)
=

0
1
f

1
(1 t
2
)

(a b) dt
2
=

1
0
f

2
(t
2
)

2
(t) dt
2
=

[a,b]
f(z) dz.

Proof (ii). Let


1
: [a, b] C be a smooth path in the plane. We can generalize (i) by dening a
smooth path
2
: [a, b] C such that
2
=

1
to be

2
(t) =
1

a +b t

, t [a, b].
We check that this is well dened by the following calculation.

1
f(z) dz =

b
a
f(
1
(t
1
))

1
(t
1
) dt
1
(Let t
2
= a +b t
1
)
=

a
b
f(
1
(a +b t
2
))

1
(a +b t
2
) dt
2
=

b
a
f(
2
(t
2
))

2
(t
2
) dt
2
(

2
(t) =

1
(t))
=

2
f(z) dz

2.3. Suppose that p [a, b] and show that [a, b] = [a, p]



+[p, b].
Proof. Let [a, b] be the path : [0, 1] C given by
(t) = (1 t)a +tb.
10 DAVID C. ULLRICH
Since p [a, b], there exists t
0
[0, 1] such that (t
0
) = p. Let [a, p] and [p, b] be the paths

1
,
2
: [0, 1] C given by

1
(t) = (1 t)a +tp,
2
(t) = (1 t)p +tb.
Let t
1
= t/t
0
and t
2
= (t t
0
)/(1t
0
) and notice dt
1
= 1/t
0
dt and dt
2
= 1/(1t
0
) dt for the following
calculations.

[a,b]
f(z) dz =

1
0
f((t))

(t) dt
=

t0
0
f((t))(b a) dt +

1
t0
f((t))(b a) dt
=

1
0
f((t
0
t
1
)) t
0
(b a) dt
1
+

1
0
f((t
2
(1 t
0
) +t
0
))

b a
1 t
0

dt
2
=

1
0
f((1 t
1
)a +t
1
p)(p a) dt
1
+

1
0
f((1 t
2
)a +t
2
p)(b p) dt
2
=

1
0
f(
1
(t
1
))

1
(t
1
) dt
1
+

1
0
f(
2
(t
2
))

2
(t
2
) dt
2
=

[a,p]
f(z) dz +

[p,b]
f(z) dz
Therefore [a, b] = [a, p]

+[p, b].
a
b
c
p
q
r
T
1
T
2
T
3
T
4
Figure 2.1. Triangle T = [a, b, c], where T
1
= [a, p, r], T
2
= [b, q, p], T
3
= [c, r, q],
and T
4
= [p, q, r].
2.4. Show that T = T
1

+T
2

+T
3

+T
4
for any triangle T.
Proof. Let T be given by gure 2.1 so that T = [a, b]

+[b, c]

+[c, a]. But T
1
= [a, p]

+[p, r]

+[r, a],
T
2
= [b, q]

+[q, p]

+[p, b], T
3
= [c, r]

+[r, q]

+[q, c], and T
4
= [p, q]

+[q, r]

+[r, p]. Thus calculating
with rules from previous exercises, we get
T = [a, b]

+[b, c]

+[c, a]
= [a, p]

+[p, b]

+[b, q]

+[q, c]

+[c, r]

+[r, a]
= [a, p]

+[p, r]

[p, r]

+[p, b]

+[b, q]

+[q, p]

[q, p]

+[q, c]

+[c, r]

+[r, q]

[r, q]

+[r, a]
= [a, p]

+[p, r]

+[r, a]

+[b, q]

+[q, p]

+[p, b]

+[c, r]

+[r, q]

+[q, c]

+[p, q]

+[q, r]

+[r, p]
= T
1

+T
2

+T
3

+T
4
.

11
Theorem 2.2 (Cauchy-Goursat: Cauchys Theorem for Triangles). Suppose that V C is open and
f : V C is dierentiable at every point of V . Then

T
f(z) dz = 0
for every triangle T V .
Proposition 2.3 (Moreras Theorem). Suppose that V is a convex open subset of the plane, and
suppose that f : V C is a continuous function with the property that

T
f(z) dz = 0 for every
triangle T V . Then there exists a function F : V C such that F

= f in V .
2.5. Let V = C\ {0} and dene f : V C by f(z) = 1/z. Then

T
f(z) dz = 0 for every triangle
T V but there does not exist F : V C such that F

= f in V .
Proof. First note that there is a convex open set V
T
such that T V
T
C\ {0}. We will show that
f|
V
T
H(V
T
). But if z V
T
, then z = 0 so we have
lim
h0
1/(z +h) 1/z
h
= lim
h0
z
z(z+h)

z+h
z(z+h)
h
=
1
z
2
.
Now by the Cauchy-Goursat theorem, we know

T
f(z) dz = 0.
But if F

= f for some F : V C, then : [0, 2] C dened by


(t) = e
it
is a smooth loop such that

V . Thus

f(z) dz = 0
by lemma 2.0. But if we simply calculate this integral we nd

f(z) dz =

2
0
ie
it
e
it
dt
= 2i.
Therefore there is no such F.

Theorem 2.4 (Cauchys Theorem in a Convex Set). Suppose that V is a convex open subset of the
plane and F : V C is dierentiable at every point of V . Then

f(z) dz = 0
for every smooth closed curve V .
Theorem 2.5 (Cauchys Integral Formula for Disks). Suppose V is an open subset of the plane and
f : V C is dierentiable at every point of V . Suppose that D(z
0
, r) V , and dene : [0, 2] C
by (t) = z
0
+re
it
(so that is the boundary of the disk D(z
0
, r), traversed counterclockwise). Then
f(z) =
1
2i

f(w) dw
w z
for all z in the open disk D(z
0
, r).
2.6. Assuming the hypotheses of theorem 2.5, show that
f(z
0
) =
1
2

2
0
f(z
0
+re
it
) dt.
12 DAVID C. ULLRICH
Proof. We dene a function
g(w) =
f(w)
w z
0
and notice that
g(z
0
+re
it
) =
f(z
0
+re
it
)
z
0
+re
it
z
0
=
f(z
0
+re
it
)
re
it
.
By Cauchys integral formula for disks we get the formula
f(z) =
1
2i

f(w) dw
w z
.
Thus we calculate f(z
0
) as follows:
f(z
0
) =
1
2i

g(w) dw
=
1
2i

2
0
g((t))

(t) dt
=
1
2i

2
0
g(z
0
+re
it
)ire
it
dt
=
i
2i

2
0
f(z
0
+re
it
) dt
=
1
2

2
0
f(z
0
+re
it
) dt.

Theorem 2.6. Suppose V is an open subset of the plane and f : V C is dierentiable at every point
of V . Suppose that D(z
0
, r) V . Then there exists a sequence of complex numbers (c
n
) such that
f(z) =

n=0
c
n
(z z
0
)
n
for every z D(z
0
, r) (and hence the series converges uniformly on every compact subset of D(z
0
, r)).
Further, the coecients are given by the formula
c
n
=
1
2i

f(w) dw
(w z
0
)
n+1
,
where : [0, 2] V is dened by (t) = z
0
+re
it
.
Corollary 2.7 (Cauchys Theorem in a Convex Set). Suppose that V C is a convex open set and
f H(V ). Then

f(z) dz = 0
for any smooth closed curve V .
Corollary 2.8 (Moreras Theorem). Suppose that V is an open subset of the plane and f : V C is
continuous. If

T
f(z) dz = 0 for every triangle T V then f H(V ).
Corollary 2.9. If f H(V ) then f

H(V ); hence by induction f is innitely dierentiable.


13
3. Elementary Results on Holomorphic Functions
Corollary 3.0. Suppose that f
n
H(V ) for n N and f
n
f in H(V ). Then f H(V ).
Corollary 3.1 (Cauchy Integral Formula for Derivatives). Suppose f H(V ) and D(z
0
, r) V .
Dene : [0, 2] V by (t) = z
0
+re
it
. Then
f
(n)
(z) =
n!
2i

f(w) dw
(w z)
n+1
(n N)
for all z D(z
0
, r).
Corollary 3.2 (Cauchys Estimates). Suppose that f is holomorphic in a neighborhood of the closed
disk D(z
0
, r) and |f| M in D(z
0
, r). Then

f
(n)
(z
0
)


Mn!
r
n
(n N).
Lemma 3.3. Suppose that V C is open and K V is compact. Then K is at positive distance
from the copmlement of V : There exists > 0 such that
|z w| >
for all z K and w C\V .
Lemma 3.4. Suppose that K C is compact, > 0, and let
K

zK
D(z, ).
Then K

is compact.
Proposition 3.5. Suppose that f
n
H(V ) for n N and f
n
f in H(V ). Then f

n
f

in H(V ).
3.1. Dene f : R R by f(x) = 1/(1 + x
2
). Show that for every x R there exists > 0 such that
f is represented by a power series convergent in (x , x +), although the power series at the origin
does not converge on the entire line.
Proof. Let V = C\ {i} and dene F : V C by
F(z) =
1
1 +z
2
.
Since F is dierentiable at every point of V , theorem 2.6 states there is a sequence of complex numbers
(c
n
) such that
F(z) =

n=0
c
n
(z z
0
)
n
for every z D(z
0
, |z i| /2). Since c
n
=
F
(n)
(z0)
n!
=
f
(n)
(z0)
n!
, the coecients are real for all z
0
R.
Now if x
0
R is given, we have
f(x) =

n=0
c
n
(x x
0
)
n
for all x (x
0
|z i| /2, x
0
+|z i| /2).
Suppose that

n=0
c
n
x
n
is a power series for f centered at the origin converging to f for all real
numbers, with c
n
=
f
(n)
(0)
n!
. Then (c
n
r
n
) is bounded for all r (0, ), so R = . But these are the
same coecients for the power series of f(z) =
1
1+z
2
centered at the origin, with R = 1. This is a
contradiction, so there is no such power series.
Corollary 3.6 (Liouvilles Theorem). A bounded entire function must be constant.
Corollary 3.7 (Fundamental Theorem of Algebra). A complex polynomial of positive degree has a
(complex root).
14 DAVID C. ULLRICH
Corollary 3.8. Suppose that f H(V ) and V is connected. If all the derivatives of f vanish at some
point of V then f is constant.
Lemma 3.9. Suppose that f H(V ) and f has a zero of order N at z V . Then there exists
g H(V ) with g(z) = 0 such that
f(w) = (w z)
N
g(w)
for all w V .
3.2. Suppose that A and B are open sets, f H(A), g H(B), and f = g in AB. Show that there
exists F H(A B) such that F = f in A and F = g in B.
Proof.
3.3. Give an example of two sets A, B C, a function f analytic on A and a function g analytic on
B such that A B = but there does not exist a function F analytic on A B with F = f on A and
F = g on B.
Example.
Corollary 3.10. Suppose that f H(V ) and V is connected. If Z(f) has a limit point in V then f
is constant.
Theorem 3.11 (Maximum Modulus Theorem). Suppose that f H(V ) and V is connected. Then
|f| cannot achieve a (local) maximum in V unless f is constant: If f is nonconstant then for every
a V and > 0 there exists z V with |f(z)| > |f(a)| and |z a| < .
Proposition 3.12 (Parseval Formula). Suppose that the power series
f(z) =

n=0
c
n
(z a)
n
converges for |z a| < r. Then
1
2

2
0

f(a +e
it

2
dt =

n=0
|c
n
|
2

2n
for every [0, r).
Lemma 3.13. If a holomorphic function is bounded near an isolated singularity then the singularity
is removable.
Proposition 3.14. Suppose that f H(D

(z, r)).
(i) f has a pole at z if and only if there exist a positive integer N and complex numbers (c
n
)
nN
such that c
N
= 0 and
f(w) =

n=N
c
n
(w z)
n
for w D

(z, r).
(ii) f has an essential singularity at z if and only if f(D

(z, )) is dense in C for all (0, r).


Proposition 3.15. Suppose that f H(D

(z
0
, r)) has the Laurent series expansion
f(z) =

n=
c
n
(z z
0
)
n
.
Then
(i) f has a removable singularity at z
0
if and only if c
n
= 0 for all n < 0.
(ii) f has a pole of order N > 0 at z
0
if and only if c
N
= 0 and c
n
= 0 for all n < N.
(iii) f has an essential singularity at z
0
if and only if there exist innitely many values of n < 0
such that c
n
= 0.
15
3.4. Suppose that f H(C) and |f(z)| e
Re z
for all z. Show that f(z) = ce
z
for some constant c.
Proof. We begin with the inequality

f(z)
e
z

=
|f(z)|
e
Re z
1.
This shows f(z)/e
z
is a bounded entire function so Liouvilles theorem states that f(z)/e
z
= c for
some constant c. Therefore f(z) = ce
z
.
3.5. Suppose that f H(C), n is a positive integer and |f(z)| (1 +|z|)
n
for all z. Show that f is a
polynomial.
Proof 1. Since f H(C) we can write f as
f(z) =

k=0
c
k
z
k
,
for all z C. With some calculations we see that
f(z) =

k=0
c
k
z
k
=
n

k=0
c
k
z
k
+

k=n+1
c
k
z
k
=
n

k=0
c
k
z
k
+z
n+1

k=0
c
k+n+1
z
k
.
Now let h H(C) be given by
h(z) =

k=0
c
k+n+1
z
k
.
For clarity note that we now have
(3.1) f(z) =
n

k=0
c
k
z
k
+z
n+1
h(z).
Since h H(C) there is some a C such that h(0) = a. If z = 0, then we get
h(z) =
f(z)

n
k=0
c
k
z
k
z
n+1
.
Thus we can calculate for z = 0,
|h(z)| =

f(z)

n
k=0
c
k
z
k
z
n+1

1
z
n+1

|f(z)| +

1
z
n+1

k=0
c
k
z
k

1
z
n+1

(1 +|z|)
n
+

1
z
n+1

k=0
|c
k
| |z|
k
=
1
|z|

1 +
1
|z|

n
+
n

k=0
|c
k
| |z|
kn1
.
16 DAVID C. ULLRICH
This shows lim
z
h(z) = 0 hence h is bounded and entire. By Liouvilles theorem we can let h(z) = c
for some c C. But notice that lim
z
h(z) = 0 so c = 0 and thus h(z) = 0 for all z C. Therefore
equation 3.1 shows that
f(z) =
n

k=0
c
k
z
k
,
so f is a polynomial.
Proof 2. Since f H(C) we can write f as

k=0
c
k
z
k
,
for all z C. Corollary 3.2 shows
|c
k
|
(1 +r)
n
r
k
, (r > 0).
If k > n, then
(1 +r)
m
r
k
0, (r ).
So c
k
= 0 for all k > m.
3.6. Suppose that f, g H(D(z, r)), 1 m n, f has a zero of order n at z and g has a zero of order
m at z. Show that f/g has a removable singularity at z.
Proof. Since f, g H(D(z, r)), there exist F, G H(D(z, r)) such that F(z) = 0 = G(z) and
f(w) = (w z)
n
F(w), g(w) = (w z)
m
G(w).
Notice that g(z) = 0 implies that f/g has a singularity at z. But for w D

(z, r) we have
f(w)
g(w)
=
(w z)
n
F(w)
(w z)
m
G(w)
= (w z)
nm
F(w)
G(w)
.
Since (wz)
nm
F(w)/G(w) H(D(z, r)), then f/g has a removable singularity at z by denition.
3.7. Suppose that f H(C) and f(n) = 0 for all n Z. Show that all the singularities of f(z)/ sin(z)
are removable.
Proof. First we show with the help of Appendix 3 that sin(z) = 0 if and only if z Z. So note that
if Re(z) = 0 then

e
iz

e
iz

, so that sin(z) = 0. So all the zeros of sin are real. It has been shown
in Appendix 3 that there are exactly two zeroes in the interval [0, 2), that sin(0) = sin() = 0, and
that sin has period 2. Thus sin(z) = 0 if and only if z Z.
Now we see that all singularities of f(z)/ sin(z) occur when z Z. But notice for n Z, sin

(n) =
cos(n) = . Thus sin(z) has a zero of order one at every n Z. By exercise 3.6 f(z)/ sin(z) has
a removable singularity at every n Z.
3.8. Suppose that f H(C), f(z +1) = f(z) for all z, f(0) = 0, and |f(z)| e
|Imz|
for all z. Show
that f(z) = c sin(z) for some constant c.
Proof. Note that f(0) = 0 and suppose f(m) = 0 for 0 m < n. Then f(n) = f((n 1) + 1) =
f(n) = 0. So f(n) = 0 for all n Z
+
. Similarly suppose f(m) = 0 for n < m 0. Then
f(n) = f((n + 1) 1) = f(n + 1) = 0 so f(n) = 0 for all n Z. Thus by exercise 3.7 all the
singularities of f(z)/ sin(z) are removable. Let g be the entire function that agrees with f(z)/ sin(z)
17
on C\Z and note that g(z + 1) = g(z) for all z C. Let z = x + iy so if y > 1 we have the following
inequalities:
|g(z)| =

f(z)
sin(z)

2ie
y
e
iz
e
iz

2e
y
|e
y
| |e
y
|

2e
y
e
y
e
y
=
2
1 e
2y

2
1 e
2
.
Similar arguments show g is bounded for y < 1 and thus g is bounded on {z C | |Im(z)| 1}.
Let K = {z C | |Re z| 1/2, |Imz| 1} and observe that g H(C) implies g(K) is bounded
because K is compact. Since g(z) = g(z + 1) we know g is bounded on {z C | Imz 1}. Thus g is
bounded on C.
Since g is a bounded entire function, corollary 3.6 states there is a c C such that g(z) = c for
all z C. If z C\Z, then f(z)/ sin(z) = c. If z Z, then f(z) = 0 = c 0 = c sin(z). Therefore
f(z) = c sin(z) for all z C.
3.11. Let V = C\ {0}. Show that there does not exist f H(V ) such that e
f(z)
= z for all z V .
Proof. Suppose f H(V ) such that e
f(z)
= z for z V . Taking the derivative we see that
e
f(z)
f

(z) = 1.
Thus
f

(z) =
1
e
f(z)
=
1
z
.
But if we let : [0, 2] C be dened by (t) = e
it
, then
0 =

(z) dz
=

1
z
dz
=

2
0
ie
it
e
it
dt
= i2.
This is a contradiction hence there is no such f.
3.16. Suppose k is a nonzero integer and dene f : R C by f(t) = e
ikt
. Let = /k and set
I =
1
2

2
0
f(t) dt.
Using the fact that f has period 2 show that
1
2

2+

f(t) dt = I.
Using the fact that f(s +t) = f(s)f(t) show that
1
2

2+

f(t) dt = I;
18 DAVID C. ULLRICH
conclude that I = 0.
Proof. Since f has period 2 we can calculate that
1
2

2+

f(t) dt =
1
2

f(t) dt +

2+
2
f(t) dt

=
1
2

f(t) dt +


0
f(t) dt

=
1
2

2
0
f(t) dt
= I.
Since f(s +t) = f(s)f(t) we calculate that
1
2

2+

f(t) dt =
1
2

2
0
f(t +) dt
=
1
2

2
0
f(t)f() dt
=
1
2

2
0
f(t) dt
= I.

3.18. Suppose that N is a positive integer, < N + 1, and f H(D

(z
0
, r)) satises
|f(z)| c |z z
0
|

.
Show that f has either a removable singularity or a pole of order no larger than N at z
0
.
Proof 1. Since |f(z)| c |z z
0
|

, we have

(z z
0
)
N+1
f(z)

c |z z
0
|
N+1
,
for all z D

(z
0
, r). But N+1 > 0 implies c |z z
0
|
N+1
0 as z z
0
. Thus (zz
0
)
N+1
f(z)
0 as z z
0
and hence is bounded near z
0
. So there exists (c
n
) such that
(z z
0
)
N+1
f(z) =

n=0
c
n
(z z
0
)
n
,
for all z D

(z
0
, r). Since c
0
= lim
zz0
(z z
0
)
N+1
f(z) = 0,

n=0
c
n
(z z
0
)
n
=

n=1
c
n
(z z
0
)
n
.
Thus for all z D

(z
0
, r) we have
f(z) = (z z
0
)
N1

n=1
c
n
(z z
0
)
n
=

n=1
c
n
(z z
0
)
nN1
=

n=N
c
n+N+1
(z z
0
)
n
.
So f has a removable singularity or a pole of order no larger than N at z
0
.
Proof 2. Proposition 3.15 states that
f(z) =

n=
c
n
(z z
0
)
n
19
for all z D

(z
0
, r), where
c
n
=
1
2i

D(z0,)
f(z)
(z z
0
)
n+1
dz
for any (0, r). This shows that
|c
n
|
1
2
c

n+1
2 = c
n
.
Now suppose that n < N is an integer. Then n (N + 1) since n is an integer, so n
+ (N + 1) > 0. So we have
|c
n
| lim
0
c
n
= 0
for all integers n < N, and so
f(z) =

n=N
c
n
(z z
0
)
n
.

3.19 (Not from book). Suppose that f is entire and |f(z)| |z|
3/2
for all z. Show that f = 0.
Proof. Since |f(z)| |z|
3/2
, letting g(z) = f(z)/z
2
for all z C\ {0} we have
|g(z)| =

f(z)
z
2

|z|
1/2
.
Thus |g(z)| 1 for all z C\D. But
|zg(z)| |z|
1/2
,
so |z|
1/2
0 as z 0 shows zg(z) 0 as z 0. Since zg(z) is bounded near 0, we let
zg(z) =

n=0
c
n
z
n
,
for all z D\ {0}. Since 0g(0) = 0, we have

n=0
c
n
z
n
=

n=1
c
n
z
n
. Thus
g(z) = z
1

n=1
c
n
z
n
=

n=1
c
n
z
n1
=

n=0
c
n+1
z
n
,
for all z D\ {0}. But now there exists M 0 such that |

n=0
c
n+1
z
n
| M for all z D, so
|

n=0
c
n+1
z
n
| max {1, M} for all z C. Thus corollary 3.6 states there exists c C such that

n=0
c
n+1
z
n
= c for all z C. So f(z)/z
2
= c for all z C\ {0} and |f(z)| =

cz
2

|z|
3/2
for all
z C\ {0}. So |c| |z|
1/2
for all z C\ {0} shows c = 0. Thus f(z) = 0 for all z C\ {0} and
|f(0)| |0|
3/2
= 0 shows f = 0.
4. Logarithms, Winding Numbers and Cauchys Theorem
Theorem 4.0. Suppose that V C is open and f H(V ). We have

f(z) dz = 0
for every smooth closed curve in V if and only if f = F

for some F H(V ).


Proposition 4.1. Suppose that V is open and f C(V ) satises e
f(z)
= z for all z V . Then
f H(V ) (so that f is a branch of the logarithm in V ).
Proposition 4.2. Suppose that V is a connected open subset of the plane, z
0
V and L H(V ).
The following are equivalent:
(i) L is a branch of the logarithm in V .
(ii) e
L(z0)
= z
0
and L

(z) = 1/z for all z V .


20 DAVID C. ULLRICH
Proposition 4.3. Suppose that V is an open subset of the plane. There exists a branch of the logarithm
in V if and only if there exists f H(V ) with f

(z) = 1/z for all z V .


Corollary 4.4. If V is a convex open subset of C\ {0} then there exists a branch of the logarithm in
V .
Proposition 4.5. Suppose that V is an open subset of C. There exists a branch of the logarithm in
V if and only if

dz
z
= 0
for all smooth closed curves in V .
Lemma 4.6. Suppose that : [0, 1] C is continuous and 0 C\

.
(i) If
0
R and e
i0
= (0)/ |(0)| then there exists a unique continuous function : [0, 1] R
such that (0) =
0
and
(t) = |(t)| e
i(t)
for all t [0, 1].
(ii) For every > 0 there exists > 0 such that if (t) = | (t)| e
i

(t)
and also

(t) (t)

< for
all t [0, 1].
(iii) If is a closed curve then ((1) (0)/2 is an integer; if is a smooth closed curve then
(1) (0)
2
=
1
2

dz
z
.
Proposition 4.7. Suppose that is a cycle in C. The function I(a) = Ind (, a) is an integer-
valued function on C\

which is constant on the components of C\

and vanishes on the unbounded


component.
Proposition 4.5 (Part 2). Suppose that V is an open subset of C. There exists a branch of the
logarithm in V if and only if 0 / V and
Ind (, 0) = 0
for all closed curves in V .
Lemma 4.8. Suppose that V is an open subset of the plane and f H(V ). Dene g : V V C by
g(z, w) =

f(z)f(w)
zw
(z = w),
f

(z) (z = w).
Then g is continuous in V V , and if we dene g
w
(z) = g(z, w) then g
w
H(V ) for every w V .
Theorem 4.9 (Cauchys Integral Formula; Homology Version). Suppose that V is an open subset of
the plane and is a cycle in V . If has the property that
Ind (, a) = 0
for all a C\V then
1
2i

f(w) dw
w z
= Ind (, z) f(z)
for all f H(V ) and all z V \

.
Theorem 4.10 (Cauchys Theorem; Homology Version). Suppose that V is an open subset of the
plane and is a cycle in V . If has the property that
Ind (, a) = 0
for all a C\V then

f(z) dz = 0
for all f H(V ).
21
Theorem 4.10 (Cauchys Theorem, Part 2). Suppose that V is an open subset of the plane and is
a cycle in V . If

f(z) dz = 0
for all f H(V ) of the form f(z) = 1/(z a) then

f(z) dz = 0
for all f H(V ).
4.1. Show directly from the denition that Ind (D(a, r), a) = 1 for any a C and r > 0.
Proof. Let D(a, r) = where : [0, 1] C is dened by
(t) = a +re
2it
.
So Ind (D(a, r), a) = Ind (, a). But Ind (, a) is given by
Ind (, a) =
1
2
((1) (0))
where : [0, 1] R is a continuous function such that
(t) a = |(t) a| e
i(t)
.
If we dene (t) = 2t, we note that (t) a = re
2it
= |(t) a| e
i(t)
. Calculating we nd that
Ind (, a) =
1
2
((1) (0)
=
1
2
(2)
= 1.

4.2. Suppose a C and r > 0. Show that


Ind (D(a, r), z) =

1, |z a| < r,
0, |z a| > r.
(What is Ind (D(a, r), z) when |z a| = r ?)
Proof. First note that D(a, r) separates C into two components, D(a, r) and C\D(a, r). Thus propo-
sition 4.7 states that there is n Z such that Ind (D(a, r), z) = n for all z D(a, r). But exercise 4.1
shows Ind (D(a, r), a) = 1 so Ind (D(a, r), z) = 1 for all z D(a, r). Since C\D(a, r) is unbounded,
proposition 4.7 also states that Ind (D(a, r), z) = 0 for all z C\D(a, r).
4.3. Show that theorem 4.9 follows from theorem 4.10, by an argument similar to the proof of theo-
rem 2.5 above.
Proof (Theorem 4.9). Let f H(V ),

V with
Ind (, a) = 0
for all a C\V . Since V is open, let D(z, r) V for z V \

and let : I C be given by


(t) = z +re
Ind(,z)2it
. Note that Ind


, z

= Ind (, z) Ind (, z) = 0. Since


f(w)
wz
H(V \ {z}),
(

)

V \ {z}, and Ind


, a

= 0 for all z C\(V \ {z}), theorem 4.10 states

f(w)
w z
dw

f(w)
w z
dw =

f(w)
w z
dw = 0.
22 DAVID C. ULLRICH
But
1
2

f(w)
wz
dw = Ind (, z) f(z) = Ind (, z) f(z). Therefore
1
2

f(w)
w z
dw = Ind (, z) f(z),
for all z V \

.
Theorem 4.11. Suppose that a C and 0 r < R . If f H(A(a, r, R)) then there exists a
sequence of complex numbers (c
n
)

n=
such that
f(z) =

n=
c
n
(z a)
n
(z A(a, r, R)).
The series converges uniformly on compact subsets of A(a, r, R) and the coecients are given by
c
n
=
1
2i

D(a,)
f(w) dw
(w a)
n1
for any (r, R).
Proposition 4.12. Suppose that
0
and
1
are closed curves in the open set V and
0

V

1
. Then
Ind (
0
, a) = Ind (
1
, a)
for all a C\V .
Theorem 4.13 (Cauchys Theorem; Homotopy Version). Suppose V is an open set in the plane,
1
and
2
are smooth closed curves in V , and
1

V

2
. Then
1
2i

1
f(z) dz =
1
2i

2
f(z) dz
for all f H(V ).
4.4. Suppose that V is simply connected, is a smooth closed curve in V and a C\V . Show that
Ind (, a) = 0.
Proof. If z V , then
c
(t) = z puts any a C\V in the unbounded component of

. Proposition 4.7
shows that Ind (
c
, a) = 0 and proposition 4.12 shows that Ind (, a) = Ind (
c
, a) = 0.
4.5. Show that any convex open set is simply connected.
Proof. Take a loop : [0, 1] V where V is our convex open set and let z V . Since V is convex we
can take the line segment
t
: [0, 1] V dened by

t
(t

) =
t
(1 t

) +t

,
where
t
= (t). Thus we construct a path homotopy : [0, 1]
2
V given by
(t

, t) =
t
(t

)
from
0
= to
1
(t

) = z for all t

[0, 1]. Therefore V is simply connected.


Theorem 4.14 (Cauchys Theorem for Simply Connected Sets). If V is simply connected and is a
cycle in V then

f(z) dz = 0
for all f H(V ).
Corollary 4.15. Any nonvanishing holomorphic function in a simply connected set has a holomorphic
logarithm. That is, if V is simply connected, f H(V ) and f has no zero in V then there exists
L H(V ) with
e
L
= f
in V .
23
4.6. Show that any nonvanishing holomorphic function in a simply connected set has a holomorphic
square root.
Proof. Let V be a simply connected set and let f H(V ) be our nonvanishing function. Corollary 4.15
states that there is L H(V ) with e
L
= f in V . Dene g = exp(
1
2
L) so that
g
2
= exp(
1
2
L) exp(
1
2
L) = exp(L) = f.

4.7.
(i) Suppose that f and g are holomorphic near z
0
and f has a simple zero at z
0
. Find an expression
for the residue of g/f at z
0
(and prove that it is correct).
(ii) Suppose that f has a simple pole at z
0
and g is holomorphic near z
0
. Show that
Res (fg, z
0
) = g(z
0
) Res (f, z
0
) .
(iii) Suppose that f is holomorphic in a neighborhood of z
0
, set g(z) = f(z)/(z z
0
)
n
, and show
that
Res (g, z
0
) = f
(n1)
(z
0
)/(n 1)!.
For simplicity we note that if f has a removable singularity at z
0
then
Res (f, z
0
) = 0 = lim
zz0
(z z
0
)f(z).
Thus Res (f, z
0
) = lim
zz0
(z z
0
)f(z) if f has a removable singularity or a simple pole.
Proof (i). Since g/f has a removable singularity or simple pole at z
0
, we calculate
Res (g/f, z
0
) = lim
zz0
(z z
0
)
g(z)
f(z)
= lim
zz0
(z z
0
)
f(z) f(z
0
)
g(z)
=
g(z
0
)
f

(z
0
)
.

Proof (ii). Since f has a simple pole at z


0
, then fg has either a removable singularity or a simple pole
at z
0
. Thus we calculate
Res (fg, z
0
) = lim
zz0
(z z
0
)f(z)g(z)
= Res (f, z
0
) g(z
0
).

Proof (iii). We have


f(z) =

k=0
c
k
(z z
0
)
k
near z
0
. But then
g(z) =

k=0
c
k
(z z
0
)
kn
=

k=n
c
k+n
(z z
0
)
k
.
24 DAVID C. ULLRICH
So Res (g, z
0
) = c
n1
by denition of residue. But
c
n1
=
f
(n1)
(z
0
)
(n 1)!
.

4.8. Suppose that f has an isolated singularity at z


0
. Fix r > 0 and n Z, and dene : [0, 2] C
by
(t) = z
0
+re
int
.
Show that Ind (, z
0
) = n and that
1
2i

f(z) dz = Ind (, z
0
) Res (f, z
0
)
if r > 0 is small enough (be explicit about how small r must be.)
Proof.
Lemma 4.16. Suppose that V C is open, S V is (relatively) closed in V , and every point of S is
isolated (that is, for every z S there exists r > 0 such that S D(z, r) = {z}). Suppose that is a
cycle in V \S such that Ind (, a) = 0 for all a C\V . Then
Ind (, z) = 0
for all but nitely many z S.
Theorem 4.17 (Residue Theorem). Suppose that V C is open, S V is (relatively) closed in V ,
and every point of S is isolated. Suppose that f H(V \S). (In other words, f is holomorphic in
V except for isolated singularities at the points of S.) If is a cycle in V such that

V \S and
Ind (, a) = 0 for all a C\V then
1
2i

f(z) dz =

zS
Ind (, z) Res (f, z) .
(Note that all but nitely many terms in the sum vanish, by Lemma 4.16.)
4.9. Show that the Cauchy Integral Formula (theorem 4.9) is an immediate consequence of the Residue
theorem.
Proof. Suppose that V is an open subset of the plane and is a cycle in V . Let have the property
that Ind (, a) = 0 for all a C\V . If f H(V ) and z V \

, then let g(w) = f(w)/(w z) so


g H(V \ {z}). Thus the residue theorem states that
1
2i

g(w) dw = Ind (, z) Res (g, z) = Ind (, z) f(z).

4.10. Show that

dx
1+x
2
= , using the Residue thoerem (4.17).
Proof. Let R > 1, =
1

+
2
where
1
= [R, R] and
2
: [0, ] C is given by
2
(t) = Re
it
.
Let f(z) =
1
1+z
2
and notice that f has isolated singularities at i. Observe theorem 4.17 and let
V = {z C | Im(z) > 1/2} and notice Ind (, a) = 0 for all a C\V . Since V is open and {i} is
25
relatively closed in V with f|
V
H(V \ {i}), then

f(z) dz =

f|
V
(z) dz
= 2i

z{i}
Ind (, z) Res (f|
V
, z)
= 2iInd (, z) Res (f|
V
, i)
= 2iRes (f|
V
, i) .
Since 1 + z
2
= (z i)(z + i), 1 + z
2
has a simple zero at i. Thus Res (f|
V
, i) =
1
2i
by exercise 4.7 (i).
Thus

f(z) dz = 2i
1
2i
= .
But

f(z) dz =

1
f(z) dz +

2
f(z) dz
=

[R,R]
1
1 +z
2
dz +


0
Rie
it
1 + (Re
it
)
2
dt
=

R
R
1
1 +t
2
dt +i


0
1
e
it
R
+Re
it
dt

1
1 +t
2
dt (R ).

4.11. Suppose that P and Q are polynomials with deg(Q) deg(P) + 2 and let f(z) = P(z)/Q(z).
Suppose that Q has no zero on the real axis and show that

f(x) dx is equal to 2i times the sum


of the residues of f in the upper half-plane.
Proof. Let
1
= [R, R] and
2
: [0, 1] C be given by

2
(t) = Re
it
,
and let =
1

+
2
. The residue theorem states that if S = {z C | Im(z) > 0, Q(z) = 0}, then

f(z) dz = 2i

zS
Ind (, z) Res (f, z) = 2i

zS
Res (f, z). But

2
f(z) dz =

2
P(z)
Q(z)
dz =

1
0
P(Re
it
)Rie
it
Q(Re
it
)
dt,
which goes to 0 as R because deg(Q(z)) deg(zP(z)) + 1. But

f(x) dx = lim
R

1
f(z) dz
= lim
R

f(z) dz
= 2i

zS
Res (f, z) .

The next exercise gives a version of Parsevals Formula for Laurent series:
4.12. Suppose that f H(A(z
0
, r, R)) has Laurent series f(z) =

n=
c
n
(z z
0
)
n
. Show that
1
2

2
0

f(z
0
+e
it
)

2
dt =

n=

2n
|c
n
|
2
26 DAVID C. ULLRICH
for r < < R.
Proof. Let s
N
=

N
n=N
c
n
(z z
0
)
n
, so
1
2

2
0

s
N
(z
0
+e
it
)

2
dt =
1
2

2
0
N

n=N
c
n

n
e
int
N

m=N
c
m

m
e
imt
dt
=
N

m,n=N
c
n
c
m

m+n
1
2

2
0
e
i(nm)t
dt
=
N

n=N
|c
n
|
2

2n
.
Now if < r then s
N
f uniformly on D(a, ), and hence
1
2

2
0

f(a +e
it
)

2
dt = lim
N
1
2

2
0

s
N
(a +e
it
)

2
dt
= lim
N
N

n=N
|c
n
|
2

2n
=

n=
|c
n
|
2

2n
.

4.13. Prove the homology version of the Cauchy Integral Formula for derivatives: Under the same
hypotheses as theorem 4.9 we have
n!
2i

f(w) dw
(w z)
n+1
= Ind (, z) f
(n)
(z)
for n N.
Proof. Let g(w) = f(w)/(w z)
n+1
. Exercise 4.7 shows
1
2i

g(w) dw =

wS
Ind (, w) Res (g, w)
= Ind (, z) Res (g, z)
= Ind (, z)
f
(n)
(z)
n!
.

4.14. Show that exercise 4.4 follows immediately from proposition 4.12.
Proof. Let V be simply connected and a smooth loop in V with a C\V . Then Ind (, a) =
Ind (
a
, a) = 0, where
a
(t) = a for all t [0, 1].
The next few exercises give an example of using the residue theorem (actually just Cauchys theorem
here) to evaluate an improper integral that is a little more elaborate than exercise 4.10 above.
4.15. Suppose that f has a simple pole at z
0
and a < b. For r > 0 dene
r
: [a, b] C by

r
(t) = z
0
+re
it
, so that
r
is an arc on the circle with center z
0
and radius r. Show that
lim
r0
+

r
f(z) dz = (b a)iRes (f, z
0
) .
Proof.
27
4.16. Suppose that f
n
: [a, b] C is continuous for n N, |f
n
(t)| M for all n and all t, and suppose
that f
n
0 uniformly on [a + , b ] for every (0, (b a)/2). Show that

b
a
f
n
(t) dt 0 as
n .
Proof. Let > 0 be given. Then if < (b a)/2, we have

a+
a
f
n
(t) dt

a+
a
|f
n
(t)| dt

a+
a
=
M. Similarly,

b
b
f
n
(t) dt

M. Let = min(/4M, (b a)/2). Now by uniform convergence,


there is N N such that |f
n
(t)| < /(2(b a 2)) for n > N, t [a +, b ]. Therefore if n > N,

b
a
f
n
(t) dt

< M +

2(b a 2)
(b a 2) +M = .

4.17. Find
lim
0
+
,R


R
+

e
it
t
dt.
Let f(z) = e
iz
/z. Find

f(z) dz and apply the previous two exercises to gure out what happens
to two pieces of the integral as 0 and R .
Proof. For 0 < < R < let =
1

+
2

3

+
4
, where
1
= [R, ],
2
: [0, ] C is dened by

2
(t) = e
i(t)
,
3
= [, R], and
4
: [0, ] C is dened by (t) = Re
it
. Exercise 4.7 and exercise 4.15
show that
lim
0
+

2
e
iz
z
dz = iRes

e
iz
z
, 0

= ie
i0
= i.
lim
R

4
e
iz
z
dz = lim
R


0
e
iRe
it
iRe
it
Re
it
dt
= lim
R
i


0
e
iRe
it
dt
=
Therefore the answer is i.
4.18. Consider the integral
lim
R

R
R
sin(t)
t
dt.
(i) Explain why you cannot evaluate this integral using the method we used in the previous
exercise.
(ii) Use the result of the previous exercise to evaluate the integral.
(i). The method in the last exercise will not work here because lim
R
sin(Re
it
) is not bounded.
28 DAVID C. ULLRICH
(ii).
i =

R
R
e
it
t
dt
=

R
R
cos(t) +i sin(t)
t
dt
=

R
R
cos(t)
t
+i

R
R
sin(t)
t
dt.
Thus

R
R
sin(t)
t
dt = .

The next two exercises allow one to give a rigorous proof that various winding numbers are what
they are without doing any work at all, in most of the cases where the value of the winding number is
supposed to be obvious:
4.19. Suppose that
1
: [a, b] C\ {z
0
} is a smooth closed curve and let
2
(t) =
1
(t) z
0
. Show that
Ind (
1
, z
0
) = Ind (
2
, 0).
Proof.
Ind (
1
, z
0
) =
1
2i

1
dz
z z
0
=
1
2i

b
a

1
(t)

1
(t) z
0
dt
=
1
2i

b
a

2
(t)

2
(t)
dt
=
1
2i

2
dz
z
= Ind (
2
, 0) .

4.20. Suppose that : [a, b] C\ {z


0
} is a smooth closed curve. Suppose that a < c < b. Show
that if Re((t)) Re(z
0
) for all t [a, c], Re((t)) Re(z
0
) for all t [c, b], Im((a)) < Im(z
0
) and
Im((c)) > Im(z
0
) then Ind (, z
0
) = 1. (The previous exercise shows that you can assume that z
0
= 0.
Now use the existence of a branch of the logarithm in the right half-plane and a dierent branch of
the logarithm in the left half-plane.)
Proof. Let =

8
i=1

i
where
i
: [0, 1] C are given by
1
= |
[a,c]
,
2
(t) =
c
(1 t) + it,
3
(t) =
e
i(/2t)
,
4
(t) = i(1t)+(a)t,
5
(t) = (a)(1t)it,
6
(t) = e
i(3/2t)
,
7
(t) = i(1t)+(c)t,

8
(t) = |
[c,b]
. If
1
=

4
i=1

i
and
2
=

8
i=5

i
, then 0 = Ind (
1
, 0) + Ind (
2
, 0) because they
are contained in simply connected domains. But this is Ind (, 0) = Ind (, 0) + Ind

3

+
6
, 0

=
Ind (, 0) 1. Thus Ind (, 0) = 1.
4.21. Suppose that P(z) is a polynomial of degree n 2 with n distinct zeroes z
1
, . . . , z
n
. Explain
why it follows that every zero of P is simple, and show that
n

j=1
1
P

(z
j
)
= 0.
(Exercise 4.7 (i) shows that 1/P

(z
j
)) is the residue at z
j
of ...)
Proof. Since P(z) is a polynomial of degree n with n distinct zeroes, the fundamental theorem of
algebra states that we can write P(z) =

n
i=1
(z z
i
). This makes it clear that every zero of P
29
is simple. Thus Res (1/P(z), z
j
) = 1/P

(z
j
). Let : [0, 1] C be given by (t) = Re
2it
so that
z
j
D(0, R) for all 1 j n. Then
n

j=1
1
P

(z
j
)
=
n

j=1
Res

1
P(z)
, z
j

=
1
2i

1
P(z)
dz
=
1
2i

1
0
2iRe
2it

n
j=1
(Re
2it
z
i
)
dt

1
0
Re
2it

n
j=1
(Re
2it
z
j
)
dt

R
||

1
0
dt

n
j=1
|Re
2it
z
j
|
=
1
R
n1
||

1
0
dt

n
j=1

1
zj
Re
2it

1
R
n1
||
0 (R ).
So

n
j=1
1/P

(z
j
) = 0.
In the next exercise we give a sketch of an alternate proof of lemma 4.6 (i). You should study both
proofs and convince yourself that they are really the same proof, or at least based on the same idea,
but expressed very dierently. The notation in the proof in the exercise is simpler than that in the
proof in the text; on the other hand the proof in the exercise is somewhat more abstract, so it is not
clear to me which proof you will nd easier to understand. (I suspect this may vary from student to
student!)
So you should learn both proofs, and then the next time you see an argument that reminds you of
the version of the proof you nd harder to understand, you should consider rephrasing it in the style
of the proof you found more transparent:
4.22. Prove part (i) of lemma 4.6 as follows: Let A be the set of all x [0, 1] with the property that
there exists a continuous function : [0, x] R such that (0) =
0
and
(t) = |(t)| e
i(t)
(t [0, x]).
Show that A is nonempty, closed, and relatively open in [0, 1], and conclude that A = [0, 1] since [0, 1]
is connected.
Proof. Let Arg (0) =
0
so A is nonempty. Now let lim
n
x
n
= x for x
n
a nondecreasing sequence
in [0, 1] such that there exists continuous
n
: [0, 1] R with
n
(0) =
0
and (t)/ |(t)| = e
in(t)
for
t [0, x
n
]. But lim
n
e
in(xn)
= lim
n
(x
n
)/ |(x
n
)| = (x)/ |(x)|, so A is closed.
To show that A is relatively open in [0, 1], let a A and we will show some neighborhood of a is
contained in A. If a = 1 then we are done so suppose a = 1. Since a A there exists a continuous
function : [0, a] R such that (t) = |(t)| e
i(t)
for t [0, a]. Note since 0 /

, there exists > 0


such that 0 / D((a), ). Let > 0 be a number such that ([a, a+)) D((a), ). Thus there exists
L H(D((a), )) such that e
iL(z)
= z, specically e
iL((t))
= (t) for all t [a, a +). So e
iL((a))
=
(a) = |(a)| e
i(a)
, thus (a) = Re(L((a))) + 2k for some k Z. So dene f : [0, a +) R by
f(t) =

(t), t [0, a],


L((t))) + 2k, t [a, a +).
By the pasting lemma, f is continuous. Thus (a , a + ) [0, 1] A and therefore A is relatively
open in [0, 1].
30 DAVID C. ULLRICH
4.23. Derive lemmea 3.13 from theorem 4.9 as follows: Suppose that f H(A(z
0
, 0, R)) is bounded.
Fix r (0, R), and show that
f(z) =
1
2i

D(z0,r)
f(w) dw
w z

1
2i

D(z0,)
f(w) dw
w z
whenever 0 < < |z z
0
| < r. Now x z and consider what happens when tends to 0.
Proof. First note if z C\D(z
0
, R) then Ind (D(z
0
, r), z) = Ind (D(z
0
, ), z) = 0 and
Ind

D(z
0
, r)

D(z
0
, ), z
0

= 0
. So by theorem 4.9 we have
f(z) =
1
2i

D(z0,r)
f(w) dw
w z

1
2i

D(z0,)
f(w) dw
w z
.
Now let z A(z
0
, , R) be given so that
lim
0
+
1
2i

D(z0,)
f(w) dw
w z
= lim
0
+
1
2i

1
0
f(z
0
+e
2it
) 2ie
2it
z
0
+e
2it
z
dt

1
0
f(z
0
+e
2it
)e
2it
(z z
0
) e
2it
dt

1
0

f(z
0
+e
2it
)e
2it
(z z
0
) e
2it

dt

1
0

f(z
0
+e
2it
)

|z z
0
|
dt
M/ |z z
0
| .
So lim
0
+

D(z0,)
f(w) dw
wz
= 0. Thus f(z) =
1
2i

D(z0,r)
f(w) dw
wz
for all z A(z
0
, 0, R). Dene
g : D(z
0
, R) C by
g(z) =

f(z), z = z
0
,
1
2i

D(z0,r)
f(w) dw
wz0
, z = z
0
.
But calculating we nd
lim
h0
g(z
0
+h) g(z
0
)
h
= lim
h0
1
h
1
2i

D(z0,r)
f(w)
w z
0
+h

f(w)
w z
0
dw
= lim
h0
1
2i

D(z0,r)
f(w)
(w z
0
)
2
+h(w z
0
)
dw
=
1
2i

D(z0,r)
f(w)
(w z)
2
dw
Thus f has a removable singularity at z
0
.
5. Counting Zeroes and the Open Mapping Theorem
Theorem 5.0. Suppose that f H(V ), and f is not constant on any component of V . Suppose that
is a smooth curve in V such that Ind (, z) is either 0 or 1 for all z C\

, and equals 0 for all


z C\V . Suppose that f has no zero on

, and let = {z V | Ind (, z) = 1}. Then the number


of zeroes of f in is given by
1
2i

(z)
f(z)
dz.
Theorem 5.1 (The Argument Principle). Suppose that f, , etc. are as in Theorem 5.0. Dene a
curve (t) by (t) = f((t)). Then the number of zeroes of f in is equal to
Ind ( , 0) .
Lemma 5.2. If z, w C and |z w| < |z| +|w| then 0 does not lie on the line segment joining z and
w.
31
Proposition 5.3. Suppose that
0
and
1
are two closed curves in C\ {0} with the same parameter
interval. If
|
1
(t)
0
(t)| < |
1
(t)| +|
0
(t)|
for all t then
0
and
1
are homotopic in C\ {0}, and in particular we have
Ind (
0
, 0) = Ind (
1
, 0) .
Theorem 5.4 (Rouches Theorem). Suppose that is a smooth closed curve in the open set V
such that Ind (, z) is either 0 or 1 for all z C\

and equals 0 for all z C\V , and let =


{z V | Ind (, z) = 1}. If f, g H(V ) and
|f(z) g(z)| < |f(z)| +|g(z)|
for all z

then f and g have the same number of zeroes in .


5.1. Suppose that n is a positive integer and show that the polynomial z
n
(z 2) 1 has exactly n
roots in the open unit disk D(0, 1).
Proof. Let g(z) = z
n
(z 2) so that if z D(0, 1)\ {1}, then
|f(z) g(z)| = 1
< |z
n
(z 2) 1| + 1
< |z
n
(z 2) 1| +|z
n
(z 2)|
= |f(z)| +|g(z)| .
If z = 1, then |f(1) g(1)| = 1 < 3 = |f(1)| + |g(1)|. So |f(z) g(z)| < |f(z)| + |g(z)| for all
z D(0, 1), so Rouches Theorem states that f and g have the same nubmer of zeroes in D(0, 1).
Since g clearly has n zeroes in D(0, 1), then the polynomial z
n
(z2)1 also has n roots in D(0, 1).
Theorem 5.5 (The Open Mapping Theorem). If V is a connected open subset of the plane and
f H(V ) is nonconstant then f is an open mapping.
Theorem 5.6 (The Open Mapping Theorem, Explicit Version). Suppose that f is holomorphic in a
neighborhood of D(z
0
, r) and

f(z
0
+re
it
) f(z
0
)

> 0
for all t R. Then
D(f(z
0
), ) f(D(z
0
, r)).
Theorem 5.7 (The Open Mapping Theorem, with Bounds). Suppose that f H(V ) and |f

(z)| B
for all z V . Suppose that z
0
V and f

(z
0
) = 0. If r > 0 is small enough that D(z
0
, r) V and
also r <
2|f

(z0)|
B
then
D(f(z
0
), ) f(D(z
0
, r))
for
= r |f

(z
0
)|
Br
2
2
.
Recall that D = D(0, 1), the open unit disk.
5.2. Suppose that f H(D) and let u = Ref, v = Imf.
(i) Show that if |u| +|v| = 1 at every point of D then f is constant.
(ii) Show that if |u|+|v| 1 at every point of D then either f is constant or |u|+|v| < 1 everywhere.
Proof (i). Since D is a connected open set, the open mapping theorem states that f(D) is open in C
or f is constant. Suppose f is not constant so there exists r > 0 such that D(f(0), r) f(D). But if
|u(0)| + |v(0)| = 1, then |u(0) r/2| + |v(0)| = 1 which contradicts this belonging to the image of D,
hence f is constant.
32 DAVID C. ULLRICH
Proof (ii). Since proof to part (i) did not depend on which z D we chose, if there exists z D such
that |u| +|v| = 1 then f is constant. Thus if f is nonconstant, then |u| +|v| < 1 everywhere.
5.3. Determine the number of zeroes of f(z) = 1 + 6z
3
+ 3z
10
+z
1
1 in the annulus A(0, 1, 2).
Proof. First we determine the number of zeroes in D(0, 2). Let g(z) = z
10
(z + 3) so that
|f(z) g(z)| =

6z
3
+ 1

= 49
<

1 + 6z
3
+ 3z
10
+z
11

+ 2
10
=

1 + 6z
3
+ 3z
10
+z
11

z
10
(z + 3)

= |f(z)| +|g(z)| .
So f has ten zeroes in D(0, 2).
Now we determine the number of zeroes in D(0, 1). Let g(z) = 6z
3
+ 3z
10
+z
11
so that
|f(z) g(z)| = 1 < 2

1 + 6z
3
+ 3z
10
+z
11

6z
3
+ 3z
10
+z
11

= |f(z)| +|g(z)| ,
where the second inequality comes from the second absolute value combined with the triangle inequal-
ity.
The number of zeroes in D(0, 1) is shown to be zero by
|f(z)| =

1 + 6z
3
+ 3z
10
+z
11

6 3 1 1 = 1.
Therefore the total number of zeroes in A(0, 1, 2) is given by 10 3 = 7.
5.4. Give another proof of Rouches Theorem (Theorem 5.4) as follows: Let W = C\(, 0] be the
plane with the non-negative real axis removed. Show that there exists a branch of the logarithm
L H(W). (You should write down an explicit formula for L, and also note that the existence of a
branch of the logarithm in W follows from one of the general results in Chapter 4.) Now show that there
exists an open set O with

O such that f(z)/g(z) W for all z O; hence F = L(f/g) H(O).


Theorem 4.0 shows that

(z) dz = 0; explain how the result follows. Give an analogous proof for
Proposition 5.3.
Proof Theorem 5.4. Suppose that is a smooth loop in W such that Ind (, z) is either 0 or 1 for all
z C\

and equals 0 for all z C\W, and let = {z W | Ind (, z) = 1}. Let f, g H(W) and
|f(z) g(z)| < |f(z)| + |g(z)| for all z

. Since W is open, 0 / W and W is simply-connected,


then there exists L H(W). If z = re
it
for t (, ), then we let L(z) = log(r) +it. Since f, g are
not zero on

, each point of

has a neighborhood such that f, g are non-zero values. Let O be the


union of all of these neighborhoods, so

O with f(z)/g(z) W for all z O. Theorem 5.0 shows


that f/g has no zeroes on . Therefore f and g have the same number of zeroes in .
Proof Theorem 5.3. Suppose that
0
and
1
are two closed curves in C\ {0} with the same parameter
interval. Let
|
1
(t)
0
(t)| < |
1
(t)| +|
0
(t)|
for all t. Then
0
/
1
is a closed curve in W, a simply-connected set, so
0
/
1
is homotopic to a point.
Multiplying the homotopy by
1
gives rise to a homotopy between
0
and
1
.
5.5. Suppose that f : D D is continuous and f is holomorphic in D. Show that f has a xed point
in D.
33
Proof. Suppose that f has no xed points in D with |f(z)| = r < 1 so f(z) z = 0 and z = 0 on
D. So |f(z) z +z| = |f(z)| < 1 |f(z) z| + |z|. Thus f(z) z has one zero in D. Now let
f
r
(z) = rf(z) so f
r
has a xed point for every r [0, 1). But then f(z
r
) = z
r
/r for z
r
the xed point
of f
r
. So as r 1, z
1
D by closure so f(z
1
) = z
1
is a xed point of f in D.
5.6. Prove Hurwitzs Theorem: Suppose that D is an open set, f
n
H(D), f
n
f uniformly on
compact subsets of D, D(z, r) D, and f has no zero on D(z, r). Then there exists N such that
f
n
and f have the same number of zeroes in D(z, r) for all n > N. In particular, if D is connected,
f
n
H(D), f
n
f uniformly on compact sets, and f
n
has no zero then either f has no zero or f
vanishes identically.
Proof.
5.7. Suppose that D is a connected open set, f
n
H(D), and f
n
f uniformly on compact subsets of
D. If f is nonconstant and z D then there exists N and a sequence z
n
z such that f
n
(z
n
) = f(z)
for all n > N.
Proof. We can assume that f(z) = 0. Since f is nonconstant then this zero is isolated so take
D(z, ) such that f(w) = 0 for all w D(z, )\ {z}. Let > 0 be given and let D(z, ) D with
= min {, , }. By exercise 5.6 there exists N such that both f
n
and f have one zero in D(z, ) for
all n > N. So let z
n
D(z, ) be the point such that f
n
(z
n
) = 0. But |z
n
z| < for all n > N
and since was arbitrary, z
n
z as n .
6. Eulers Formula for sin(z)
6.0. Motivation.
6.1. Proof by the Residue Theorem.
Theorem 6.1.0 (Cauchy Integral Formula for Functions with Simple Poles). Suppose that V C is
open and is a cycle in v such that Ind (, a) = 0 for all z C\V . Suppose that S is a (relatively)
closed subset of V , S

= , and every point of S is isolated. Suppose that f H(V \S) and f has
a simple pole or removable sinularity at every point of S. Then for z V \(S

) we have
Ind (, z) f(z) =
1
2i

f(w)
w z
dw +

pS
Ind (, p) Res (f, p)
z p
.
Lemma 6.1.1. There exists a constant M such that
|cot(z)| M
whenever |Im(z)| 1 or Re(z) = n + 1/2 (n Z).
Theorem 6.1.2. If z C\Z then
cot(z) =
1
z
+

n=1

1
z n
+
1
z +n

.
6.1. Show that the series

n=0
1
z n
diverges for all z C\Z.
Proof.
6.2. Show that the series

n=1

1
z n
+
1
z +n

converges absolutely for all z C\Z.


34 DAVID C. ULLRICH
Proof.
Lemma 6.1.3. If z
1
, . . . , z
n
C and

n1
j=1
|1 z
j
| < 1/2 then

1
n

j=1
z
j

2
n

j=1
|1 z
j
| .
Lemma 6.1.4.
(i) If z
1
, . . . C and

j=1
|1 z
j
| < then

j=1
z
j
= lim
n
n

j=1
z
j
exists; furthermore,

n
j=1
z
j
= 0 unless z
j
= 0 for some j.
(ii) If (f
j
) is a sequence of complex-valued funtions on some set S and the sum

j=1
|1 f
j
|
converges uniformly on S then P
n
=

n
j=1
f
j
tends to P =

j=1
f
j
uniformly on S; if z S
then P(z) = 0 unless f
j
(z) = 0 for some j.
Proposition 6.1.5. Suppose that V is a connected open set, f
1
, f
2
, . . . H(V ) and f
n
f uniformly
on compact subsets of V . Suppose that f is not identically zero. Then
L(f
n
) L(f)
uniformly on K, if K is any compact subset of V on which f has no zero.
Lemma 6.1.6. Suppose that V is a connected open subset of C, f and g are holomorphic functions
in V neither of which vanishes identically, and
L(f) = L(g)
on the set where neither f nor g vanishes. Then f = cg for some constant c.
6.2. Estimating Sums Using Integrals.
6.3. Suppose that

is continuous on [1/2, 1/2].


(i) Show that

1/2
1/2
(t) dt (0) =
1
2

1/2
1/2

|t|
1
2

(t) dt.
(ii) Deduce that

1/2
1/2
(t) dt (0)

1
8

1/2
1/2
|

(t)| dt.
Proof (i).
1
2

1/2
1/2

|t|
1
2

(t) dt =
1
2

1/2
0

t
1
2

2
(

(t) +

(t)) dt
=
1
2

t
1
2

2
(

(t)

(t))

1/2
0

1/2
0
(

(t)

(t))(2t 1) dt

=
1
2

(2t 1)((t) +(t))|


1/2
0

1/2
0
2((t) +(t)) dt

1/2
0
((t) +(t)) dt (0)
=

1/2
1/2
(t) dt (0).
35

Proof (ii).

1/2
1/2
(t) dt (0)

1
2

1/2
1/2

|t|
1
2

(t) dt

1
2

1/2
1/2

|t|
1
2

(t)

dt

1
8

1/2
1/2
|

(t)| dt.

Theorem 6.2.0. Suppose that : (0, ) (0, ) is continuous.


(i) If is nonincreasing then


1
(t) dt

n=1
(n)


0
(t) dt.
(ii) If

is continuous then


1/2
(t) dt

n=1
(n)

1
8


1/2
|

(t)| dt.
(iii) If is convex then

n=1
(n)


1/2
(t) dt.
6.3. Proof Using Liouvilles Theorem.
6.4. Suppose that P : C C is continuous, P(z + 2) = P(z), and |P(z)| e
|z|
. Show that there
exists c such that |P(z)| ce
|Im(z)|
for all z.
Proof. First notice that |z| |Re(z)| +|Im(z)|, so e
|z|
e
|Re(z)|
e
|Im(z)|
. Thus
|P(z)|
e
|Im(z)|

e
|z|
e
|Im(z)|
e

for |Re(z)| 1. But if z C, then z = z


0
+ 2k for some |Re(z
0
)| 1 and k Z. Therefore
|P(z)| = |P(z
0
)| e

e
|Im(z)|
.

6.5. Prove Wallis formula:

2
=

n=1

4n
2
(2n 1)(2n + 1)

2
1

2
3

4
3

4
5

.
Proof. Using Eulers formula for sin(z) gives us
1 = sin(/2) =

2

n=1

1
1
4n
2

=

2

n=1

(2n 1)(2n + 1)
4n
2

.
36 DAVID C. ULLRICH
This shows the rst equality and now we show the second equality. Let a
n
= ln
n+1
n
and notice this is
a strictly decreasing positive sequence approaching zero, so

n=1
(1)
n
a
n
= L for some nite L. Now
we calculate
e
L
= e

n=1
(1)
n
an
=

n=1
e
(1)
n
ln
n+1
n
=

2
1

2
3

4
3

4
5

.
Since this converges, the second equality must also hold.

6.6. Find

n=1
1
n
2
, using the innite series for cot(z).
Solution. The innite series for cot(z) is given by
cot(z) =
1
z
+

n=1

1
z n
+
1
z +n

for z C\Z. So we calculate

n=1

1
n
2
z
2

=
1
2z
2

1
2z
2
z
sin(z)
cos(z)
=
z
sin(z)

sin(z)
z
2z
2

cos(z)
2z
2

=
z
sin(z)

1
2z
2


2
z
2
12z
2
+
1
2z
2
+

2
z
2
4z
2

=
z
sin(z)

2
z
2
6z
2
+

.
Now taking the limit as z 0 shows us that

n=1
1
n
2
=

2
6
.

7. Inverses of Holomorphic Maps


Theorem 7.0. Suppose that f H(V ) and f is one-to-one in V . Then f

has no zero in V .
Lemma 7.1. Suppose that V is convex and f H(V ). If K is a closed convex subset of the plane
and f

(z) K for all z V then


f(z) f(w)
z w
K
for all z, w V with z = w.
Proposition 7.2. Suppose that V is convex and f H(V ) is nonconstant; let W = f(V ) (note that
W iso pen by Theorem 5.5). Suppose that K isa closed convex subset of the plane and
0 / K.
If f

(z) K for all z K then it follows that f is one-to-one in V and the inverse f
1
: W V is
holomorphic.
Proposition 7.3. Suppose V is convex and f H(V ) is nonconstant. Suppose K is a closed convex
set, 0 / K

, and f

(z) K for all z V . Then f is one-to-one in V .


37
Proposition 7.4. Suppose that f H(V ), z
0
V , and f
p
rime(z
0
) = 0. Then there is an open set
containing z
0
in which f is one-to-one; furthermore, (f|

)
1
is holomorphic in f().
Theorem 7.5. Suppose that f H(V ) and f is one-to-one in V ; let W = f(V ). Then f
1
H(W)
and
(f
1
)

(w) =
1
f

(f
1
(w))
.
Theorem 7.6. Suppose that f H(V ) and f is one-to-one in V . Suppose that D(a, r) V , and let
W = f(D(a, r)). Then
f
1
() =
1
2i

D(a,r)
f

(w) w
f(w)
dw
for all W.
7.1. Show that theorem 7.5 follows immediately from theorem 7.6.
Proof. Suppose f H(V ), f is one-to-one, and W = f(V ). Take some w W and let f
1
(w) = v.
Now take D(v, r) V so f(D(v, r)) W and notice this shows W is open by the open mapping
theorem so H(W) is well-dened. Now let f() f(D(v, r)) so D(v, r) hence by theorem 7.6 and
dierentiability of f,
(f
1
)

(w) = lim
f()w
f
1
(f()) f
1
(w)
f() w
=
1
lim
f
1
(w)
f() f(f
1
(w))
f
1
(w)
=
1
f

(f
1
(w))
.
Therefore since w was arbitrary, f
1
H(W) with the given formula.
7.2. Fill in the details in the following alternate (perhaps more intuitive) proof of theorem 7.6: Let
= f(V ). Note that f
1
H(). Dene a curve : [0, 2] by (t) = f(a + re
it
). If you apply
the denition of the contour integral, you see that
1
2i

D(a,r)
f

(z) z
f(z)
dz =
1
2i

f
1
(w)
w
dw,
and the Cauchy Integral Formula shows that
1
2i

f
1
(w)
w
dw = f
1
().
Proof.
1
2i

D(a,r)
f

(z) z
f(z)
dz =
1
2i

2
0
f

(a +re
it
) (a +re
it
) ire
it
f(a +re
it
)
dt
=
1
2i

2
0
f
1
(f(a +re
it
))f

(a +re
it
)ire
it
f(a +re
it
)
=
1
2i

f
1
(w)
w
dw
= f
1
().
38 DAVID C. ULLRICH
Note we can use the Cauchy Integral Formula for the last equality because for / ,
Ind (, ) =
1
2i

1
z
dz
=
1
2i

2
0
f

(a +re
it
) ire
it
f(a +re
it
)
dt
=
1
2i

D(a,r)
f

(z)
f(z)
dt
= 0
by Cauchys theorem.
Theorem 7.7. Suppose that f H(V ), z
0
V , and f f(z
0
) has a zero of nite order m at z
0
.
Then there is an open set with z
0
and a one-to-one function g H() such that g(z
0
) = 0,
g() = D(0, r) for some r > 0, and such that
f(z) = f(z
0
) + (g(z))
m
for all z . (In particular f is exactly m-to-one in \ {z
0
}.)
Corollary 7.8. Suppose that f H(V ), z
0
V , and f f(z
0
) has a zero of nite order m at z
0
.
Then there exists an open set with z
0
, a number r > 0 and a function H(D(0, r)) such that
maps D(0, r) bijectively onto , and such that
f (z) = f(z
0
) +z
m
for all z D(0, r).
7.3. Suppose that f H(V ), W is an open set in the plane, and h: W V satises
f(h(z)) = z
for all z W.
(i) Give an example showing that it does not follow that h H(W).
(ii) Suppose in addition that h is continuous and show that h H(W).
Solution (i).
Proof (ii).
8. Conformal Mappings
8.0. Meromorphic Functions and the Riemann Sphere.
Theorem 8.0.0. If f : C

is holomorphic then there exist polynomials P and Q such that


f(z) =
P(z)
Q(z)
for all z C

.
8.1. Linear-Fractional Transformations, Part I.
Proposition 8.1.0. Suppose that f H(V ), z
0
V , and f

(z
0
) = 0. Suppose that
0
,
1
: [0, 1] V
are curves with nonvanishing (right) derivatives at 0, with
j
(0) = z
0
. Let
j
(t) = f(
j
(t)). Then
arg

1
(0)

0
(0)

= arg

1
(0)

0
(0)

(by which we mean that every argument of

1
(0)/

0
(0) is an argument of

1
(0)/

0
(0) and conversely).
Theorem 8.1.1. Aut(C

) is the set of all linear-fractional transformations.


Theorem 8.1.2. Aut(C) is equal to the set of nonconstant ane maps.
39
8.1. Show directly (without using any results in Chapters 7 or 8) that an entire function with a pole
at must be a polynomial.
Proof.
8.2. Give a direct proof of Theorem 8.1.2, without using Theorem 8.1.1.
Proof.
8.3. Show that Theorem 8.1.1 follows from Theorem 8.1.2.
Proof.
8.2. Linear-Fractional Transformations, Part II.
8.4. Give te details of both proofs of the identity
A

B
=
AB
sketched above: Write out a direct
proof and also verify carefully that M
A
=
A
.
Proof.
Lemma 8.2.0. The group M is generated by the subgroups {

| C}, {
t
| t > 0}, {

| R},
and {j, I} (here I is the identity: I(z) = z).
Theorem 8.2.1. If C C and M then (C) C.
Theorem 8.2.2. If z
1
, z
2
, z
3
are three distinct elements of C

and w
1
, w
2
, w
3
are three distinct ele-
ments of C

then there xists a unique M with (z


j
) = w
j
(j {1, 2, 3}).
8.5. The proof of Theorem 8.2.1 contains a proof that if Mxes 0,1, and then is the identity.
Show how it follows immediately that if M has three xed points it must be the identity. (The
proof should exploit the fact that M is a group, as in the proof of Theorem 8.2.1, instead of doing
calculations with the coecients of linear-fractional transformations.)
Proof.
8.6. Suppose that C is a circle. Show that the only subset of C homeomorphic to a circle is C itself.
Proof.
8.3. Linear-Fractional Transformations, Part III.
8.4. Linear-Fractional Transformations, Part IV:
The Schwarz Lemma and Automorphisms of the Disk.
Theorem 8.4.1 (Schwarz Lemma). Suppose that f : D D is holomorphic and f(0) = 0. Then
|f(z)| |z| for all z D and |f

(0)| 1. Furthermore, if |f

(0)| = 1 or |f(z)| = |z| for some nonzero


z then f is a rotation: f(z) = z for some constant with || = 1.
Theorem 8.4.2. Suppose that Aut(D) and (0) = 0. Then is a rotation:
(z) = z
for some C with || = 1.
Lemma 8.4.3.
(i)
a
(a) = 0 and
a
(0) = a.
(ii)
a
is its own inverse. (That is,
a
is an involution.)
(iii)
a
Aut(D) for any a D.
(iv) For any a, z D we have
1 |
a
(z)|
2
=
(1 |a|
2
)(1 |z|
2
)
|1 az|
2
.
(v) |

a
(0)| = 1 |a|
2
and |

a
(a)| = 1/(1 |a|
2
).
40 DAVID C. ULLRICH
Theorem 8.4.4. For any Aut(D) there exists a unique a D and C with || = 1 such that
(z) =
a
(z)
for all z D.
8.7. Suppose that Aut(
+
). Show that there exist a, b, c, d R, with ad bc = 1, such that
(z) =
az +b
cz +d
for all z
+
. Show that a, b, c, d are not unique but are almost unique.
Proof.
8.8. Explain why it is obvious that for any a
+
there exists Aut(
+
) with (a) = i. (This
follows from the corresponding fact in the disk. But an example is much more obvious in the upper
half-plane; pretend you dont know the result for the disk and give the obvious argument for Pi
+
.)
Proof.
8.5. More on the Schwarz Lemma.
Theorem 8.5.0 (Invariant Schwarz Lemma). Suppose that f : D D is holomorhpic. If f / Aut(D)
then
d(f(z), f(w)) < d(z, w)
for all z, w D with z = w and
|f

(z)|
1 |f(z)|
2
<
1
1 |z|
2
for all z D, while if f Aut(D) then
d(f(z), f(w)) = d(z, w)
and
|f

(z)|
1 |f(z)|
2
=
1
1 |z|
2
for all z, w D. (Here d(z, w) = |
z
(w)|, as above.)
8.9. Suppose that f : D D is holomorphic and dene : D D C by
(z, w) =

d(f(z),f(w))
d(z,w)
(z = w),
|f

(z)|(1|z|
2
)
1|f(z)|
2
(z = w).
Show that is continuous.
Proof.
8.10. Suppose that f : D D is holomorphic and is a smooth curve in D. Show that
L(f ) L(),
with equality if and only if f Aut(D) or L() = 0.
Proof.
8.11. Suppose z D and K D is compact. Show that there exists c > 0 such that
1 |f(w)| c(1 |f(z)|) (w K)
for every holomorphic f : D D.
Proof.
41
9. Normal Families and the Riemann Mapping Theorem
9.0. Introduction.
9.1. Quasi-Metrics.
Lemma 9.1.0. Suppose that : I R is concave. Suppose that a, b, a

, b

I, a < b, a

< b

, a

a,
and b

b. Then
(b

) (a

)
b


(b) (a)
b a
.
Lemma 9.1.1. Suppose that : [0, ) R is concave and
(0) = 0.
Then
(x +y) (x) +(y)
for all x, y 0.
Lemma 9.1.2. Suppose that : [0, ) R is concave and
(0) = 0.
Suppose further that is nondecreasing, (t) > 0 for all t > 0, and is continuous at 0.
If d is a quasi-metric on X then

d = d is also a quasi-metric on X such that
(i)

d(x, y) = 0 if and only if d(x, y) = 0,
(ii)

d(x
n
, y
n
) 0 if and only if d(x
n
, y
n
) 0.
Lemma 9.1.3. Suppose that d
j
is a quasi-metric on X for j N. Dene
d(x, y) =

j=1
2
j
d
j
(x, y)
1 +d
j
(x, y)
(x, y X).
Then d is a quasi-metric on X with the property that d(x
n
, y
n
) 0 as n if and only if
d
j
(x
n
, y
n
) 0 for every j.
Furthermore, d is a metric on X if and only if for every x, y X with x = y there exists a positive
integer j such that d
j
(x, y) > 0 (that is, if and only if the family (d
j
) separates points of X).
Lemma 9.1.4. Suppose that c
j
0 for j N and

j=1
c
j
< .
Suppose that (a
j
, n) is a double sequence such that
|a
j,n
| < c
j
for all j and n, and such that
lim
n
a
j,n
= a
j
for all j. Then
lim
n

j=1
a
j,n
=

j=1
a
j
.
Lemma 9.1.5. Suppose that (d
j
) is a sequence of quasi-metrics on X which separates points of X, so
that
d(x, y) =

j=1
2
j
d
j
(x, y)
1 +d
j
(x, y)
(x, y X)
denes a metric on X, as in Lemma 9.1.3. For x X and r > 0 denes
B
d
(x, r) = {y X | d(x, y) < r} ,
42 DAVID C. ULLRICH
as usual. For each positive integer N, x X and > 0 dene
B
N
(x, ) = {y X | d
j
(x, y) < , j {1, . . . N}} .
(i) Each B
N
(x, ) is open in the metric space (X, d).
(ii) For every r > 0 there exists a positive integer N and a number > 0 such that
B
N
(x, ) B
d
(x, r)
for all x X.
(iii) For every > 0 and positive integer N there exists a number r > 0 such that
B
d
(x, r) B
N
(x, )
for all x X.
Proposition 9.1.6. Suppose that (d
j
) is a sequence of quasi-metrics on X which separates points of
X (that is, for any x, y X with x = y there exists a j with d
j
(x, y) > 0). There exists a metric d on
X with the following two properties:
(i) For any sequences (x
n
) and (y
n
) in X we have d(x
n
, y
n
) 0 if and only if d
j
(x
n
, y
n
) 0 for
all j.
(ii) A set S X is totally bounded in the metric space (X, d) if and only if for every positive
integer N and ever > 0 there xist nitely many points x
1
, . . . , x
n
X such that
S
n

k=1
B
N
(x
k
, ).
9.1. Prove part (i) of Lemma 9.1.5
Proof.
9.2. Prove Proposition 9.1.6
Proof.
9.2. Convergence and Compactness in C(D).
Lemma 9.2.0. Suppose that D is an open subset of the plane. There exist compact sets K
j
(j N)
such that
K
j
K

j+1
and
D =

j=1
K
j
.
Theorem 9.2.1. Suppose that D is an open subset of the plane. Let (K
j
) be an exhaustion of D, and
dene
d(f, g) =

j=1
2
j
f g
Kj
1 +f g
Kj
.
Then d is a metric on C(D), and for a sequence (f
n
) C(D) and a function f C(D) we have
f
n
f in C(D) if and only if d(f
n
, f) 0.
Lemma 9.2.2. The metric space (C(D), d) is complete.
Lemma 9.2.3. The set S C(D) is totally bounded if and only if for every compact K D and
every > 0 there exist nitely many functions f
1
, . . . f
n
S such that
S
n

k=1
B
K
(f
k
, ).
Theorem 9.2.4 (Arzela-Ascoli). The set S C(D) is totally bounded if and only if it is bounded and
equicontinuous at every point of D.
43
Theorem 9.2.5 (Arzela-Ascoli). Suppose that D is an open set in the plane and S C(D). The
following are equivalent:
(i) S is pointwise bounded and pointwise equicontinuous.
(ii) Any sequence in S has a subsequence which converges uniformly on compact subsets of D.
9.3. Montels Theorem.
Lemma 9.3.0. If D C is open then H(D) is a closed subspace of C(D).
Theorem 9.3.1 (Montels Theorem). Suppose D C is open and S H(D) is closed. Then S is
compact if and only if it is uniformly bounded on compact sets (that is, if and only if for every compact
K D there exists a number M such that f
K
M for all f S).
Theorem 9.3.2 (Montels Theorem). If F H(D) is a normal family and (f
n
) is a sequence in F
then there exists a subsequence f
nj
such that f
nj
f H(D) uniformly on compact subsets of D.
9.3. Suppose that D is a connected open set in the plane. For F H(D) dene
F

= {f

| f F} .
(i) Show that if F is a normal family then so is F

.
(ii) Give an example showing that the converse of part (i) is false.
(iii) Show that the converse of (i) becomes true if you add one more very small hypothesis.
Proof (i). Let K D be compact and let f
K
M for all f F.
Solution (ii).
Proof (iii).
9.4. Suppose that D is a connected open set, F H(D) is a normal family, and (f
n
) is a sequence of
elements of F. Suppose that S D has a limit point in D and that f H(D) has the property that
f
n
(z) f(z) for every z S. Show that f
n
f in H(D).
Proof. Let X be a compact metric space, x X, (x
n
) X, and x
n
x. Since x
n
x, there exists
an innite subsequence (x
nj
) such that

x
nj
x

> . Since X is compact there exists a subsequence


of x
nj
that converges to x

X. Since

x
nj
x

> then x = x

.
Now suppose f
n
f in H(D). Since F is precompact, there exists a subsequence (f
nj
) H(D)
such that f
nj
g = f in H(D). Now consider the function h(z) = g(z) f(z). Since f
nj
(z) g(z)
and f
nj
(z) f(z) for all z S, f|
S
= g|
S
. Thus h|
S
= 0. Since Z(h) has a limit point, h vanishes on
D. Therefore f = g and hence f
n
f in H(D).
9.5. Suppose that D is an open set in the plane, M is a positive number, and let
F =

f H(D):

D
|f(z)|
2
dxdy M

.
Show that F is a normal family.
Proof.
9.6. Suppose that D is a connected open set in the plane, (f
n
)

n=1
H(D) is a normal family, and
each f
n
has no zero in D. If there exists z D with f
n
(z) 0 then f
n
0 uniformly on compact
subsets of D.
Proof. Suppose f
n
0 in H(D). The proof of exercise 9.4 shows there exists a subsequence f
nj

f = 0 in H(D). Let K be a compact subset of D. Since D is open, let D(z, r) D with f(w) >
for w D(z, r). Now let J N be a number such that if j > J, then

f
nj
(w) f(w)

< for all


w D(z, r). Since

f
nj
(w) f(w)

< <

f
nj
(w)

+|f(w)|
44 DAVID C. ULLRICH
for all w D(z, r), theorem 5.4 shows that if j > J, then f
nj
and f have the same number of zeroes
in D(z, r). This is a contradiction so f
n
0 in H(D).
9.7. Show that S C(D) is bounded (in the topological-vetor-space sense as above) if and only if for
every compact K D there exists M such that f
K
M for all f S.
Proof.
9.8. Let L be the space of all absolutely summable real sequences: x L if and only if x = (x
n
) where
each x
n
R and

n=1
|x
n
| < . Dene a metric d on L by
d(x, y) =

n=1
|x
n
y
n
| .
(i) Show that this formula actually denes a complete metric on L.
(ii) Show that a subset of L is bounded in the topological-vector-space sense if and only if it is
bounded as a subset of the metric space (L, d).
(iii) Show that the closed ball B(0, 1) is not compact.
9.4. The Riemann Mapping Theorem.
Theorem 9.4.0 (Riemann Mapping Theorem). Suppose that D C is a nonempty simply connected
open set and D = C. Then D is conformally equivalent to the unit disk D.
Lemma 9.4.1. Suppose that f : D D and : D D are holomorphic. If Aut(D) then
H
f
(z) = H
f
(z)
for all z D, while if / Aut(D) then
H
f
< H
f
(z)
for all z D with f

(z) = 0. In particular
H
f
2(z) < H
f
(z)
for all z D with f

(z) = 0.
Theorem 9.4.2 (Riemann Mapping Theorem, restated). Suppose that D C is open, connected,
nonempty, and has the property that any nonvanishing function holomorphic in D has a holomorphic
square root. If D = C then D is conformally equivalent to D.
Corollary 9.4.3. Suppose that D is a simply connected open set, D = C, and z
0
D. Then there
exists a unique conformal equivalence F : D D such that F(z
0
) = 0 and F

(z
0
) > 0.
9.5. Montels Theorem Again. See theorem 9.3.2.
10. Harmonic Functions
10.0. Introduction.
Lemma 10.0.0. Suppose that D R
2
is open and : D C is twice continuously dierentiable in
the real sense. Then is harmonic if and only if
2
/z z = 0.
Lemma 10.0.1.
(i) A holomorphic function is harmonic and hence the real and imaginary parts of a holomorphic
function are harmonic.
(ii) If is harmonic then f = /z is holomorphic.
(iii) If D is simply connected and is a real-valued harmonic function in D then there exists
f H(D) with = Re(f).
Theorem 10.0.2. Suppose that D is a connected open subset of the plane. Then D is simply connected
if and only if every real-valued harmonic function in D is the real part of some holomorphic function.
45
Theorem 10.0.2 (Part 2). A connected open subset of the plane is simply connected if and only if
every real-valued harmonic function has a harmonic conjugate.
10.1. Poisson Integrals and the Dirichlet Problem.
10.1. Explain why the following argument showing that there is no continuous function in the closed
unit disk which is holomorphic in the interior and equal to e
it
ate very boundary point e
it
is wrong: If
were such a function then v(z) = (z)1/z would dene a holomorphic function in D\ {0} vanishing
at every point of the boundary of the disk. But if the zero set of a holomorphic function has a limit
point the function must vanish identically. Thus (z) = 1/z for z = 0 and hence is not continuous
at the origin.
Solution.
Proposition 10.1.0. If is harmonic in a neighborhood of D(z, r) then
(z) =
1
2

2
0
(z +re
it
) dt.
Lemma 10.1.1. If 0 r < 1 and t R then
P
r
(t) =

n=
r
|n|
e
int
= Re

1 +re
it
1 re
it

=
1 r
2
1 2r cos(t) +r
2
=
1 r
2
(1 r cos(t))
2
+r
2
sin
2
(t)
.
Lemma 10.1.2.
(i) P
r
(t) > 0 for 0 r < 1, t R.
(ii)
1
2

2
0
P
r
(t) dt = 1 for 0 r < 1.
(iii) If 0 < then P
r
(t) 0 as r 1, uniformly for t [, ] [, ].
Theorem 10.1.3. Suppose that f C(D). Dene : D C by
(re
i
) =

f(e
it
), r = 1,
P[f](re
i
), 0 r < 1.
Then |
D
= f, C(D), and |
D
is harmonic in D.
10.2. Suppose that : D C is continuous in D and that (re
it
(e
it
) uniformly as r 1. Show
that is continuous in D.
Proof.
10.3. There is actually an application of the inequality

e
it
e
is

|t s|
for s, t R hidden in the proof above.
(i) Explain where this inequality was used.
(ii) Prove the inequality.
Solution (i).
Proof (ii).
46 DAVID C. ULLRICH
Lemma 10.1.4. Suppose that D is a bounded open set in the plane, and that C(D) satises a
weak mean value property in D: for every z D there exists = (z) > 0 such that D(z, (z)) D
and
(z) =
1
2

2
0
(z +re
it
) dt
for 0 < r < (z). If vanishes identiacally on D then vanishes identically in D.
Lemma 10.1.4 (Part 2). Suppose that D is a bounded open set in the plane, and that
1
,
2
C(D)
satisfy a weak mean value property in D: for every z D there exists = (z) > 0 such that
D(z, (z)) D and

j
(z) =
1
2

2
0

j
(z +re
it
) dt
for j {1, 2}, 0 < r < (z). If
1
=
2
at every point of D then
1
=
2
everywhere in D.
Theorem 10.1.5. Suppose that C(D) is harmonic in D. Then
|
D
= P[|
D
].
Theorem 10.1.6. Suppose that C(D) satises a weak mean value property in D: for every z D
there exists = (z) > 0 such that D(z, (z)) D and
(z) =
1
2

2
0
(z +re
it
) dt
for 0 < r < (z). Then is harmonic in D.
Proposition 10.1.7. Suppose that D
1
and D
2
are open subsets of the plane. If f : D
1
D
2
is
holomorphic and : D
2
C is harmonic then f is harmonic in D
1
.
10.4. Prove proposition 10.1.7.
Proof.
Theorem 10.1.6 (Part 2). Suppose that C(D(a, R)) satises a weak mean value property in
D(z, R): for every z D there exists = (z) > 0 such that D(z, (z)) D(a, R) and
(z) =
1
2

2
0
(z +re
it
) dt
for 0 < r < (z). Then is harmonic in D(a, R).
Theorem 10.1.8. Suppose that D is an open subset of the plane, C(D), and satises a weak
mean value property in D. Then is harmonic in D.
Theorem 10.1.6 (Part 3). Suppose that C(D) satises a very weak version of the mean value
property in D: for every z D there exists r (0, 1 |z|) such that
(z) =
1
2

2
0
(z +re
it
) dt.
Then is harmonic in D.
10.5. Suppose that D is an open subset of the plane and : D C.
(i) Show that if satises a small-radius mean value property in D it follows that satises a
small-radius mean value property in any open subset of D.
(ii) Explain why the same argument does not work for a one-radius mean value property.
Proof (i).
Solution (ii).
10.2. Poisson Integrals and Aut(D).
47
10.3. Poisson Integrals and Cauchy Integrals.
10.4. Series Representations for Harmonic Functions in the Disk.
10.5. Greens Functions and Conformal Mappings.
Lemma 10.5.5. Suppose that D is an open subset of the plane and K is a compact subset of D. There
exists a cycle in D such that Ind (, z) = 0 for all z C\D,

D\K, Ind (, z) = 0 or 1 for all


z D\

, and Ind (, z) = 1 for all z K.


10.6. Intermission: Harmonic Functions and Brownian Motion.
10.7. The Schwarz Reection Principle and Harnacks Theorem.
12. Runges Theorem and the Mittag-Leffler Theorem
12.1. Show that if P is a polynomial then there exists z such that |z| = 1 and

P(z)
1
z

1.
Proof. Suppose on the other hand that

P(z)
1
z

< 1 for all |z| = 1. Then

D
P(z)
1
z
dz

< 2
by the ML-inequality. But

D
P(z)
1
z
dz = 2i,
which is a contradiction. Therefore there exists z D such that

P(z)
1
z

1.
Lemma 12.0. Suppose that A C and R is a rational function such that all the nite poles of R are
contained in A. Then there exists a polynomial P and complex numbers (c
a,j
)
N
j=1
for each a A such
that
R(z) = P(z) +

aA
R
a
(z) = P(z) +

aA
N

j=1
c
a,j
(z a)
j
.
Note although A may be innite this is really a nite sum; we have c
a,j
= 0 unless a A actually is a
pole of R.
Lemma 12.1. Suppose that K C is compact and
D(, r) C\K.
Any function f R
D(,r)
can be uniformly approximated on K by functions in R
{}
.
12.2. Suppose that K C is compact. Suppose that (R
n
) and (S
n
) are sequences of elements of C(K)
which converge uniformly on K to f and g, respectively. Show that R
n
S
n
converges to fg uniformly
on K.
Proof. Let > 0 be given. Since f, g C(K), let |f(K)| M and |g(K)| N. Since R
n
, S
n
converge uniformly to f, g, then there exists N N such that n > N implies |R
n
(K)| M, |R
n
f| <
/(2N) and |S
n
g| < /(2M) on K. But now
|R
n
S
n
fg| = |R
n
S
n
R
n
g +R
n
g fg|
|R
n
S
n
R
n
g| +|R
n
g fg|
= |R
n
| |S
n
g| +|g| |R
n
f|
M


2M

+N


2N

= .

12.3. Give an example of (R


n
), (S
n
) C(R) such that R
n
f and S
n
g uniformly on R but R
n
S
n
does not converge to fg uniformly on R.
48 DAVID C. ULLRICH
Example. Let R
n
= x and S
n
= 1/n, so R
n
x and S
n
0 uniformly on R. But R
n
S
n
= x/n does
not converge to 0 uniformly on R.
Lemma 12.2. Suppose that K C is compact and
A C\K,
where
A = {z C | |z| > r}
for some r > 0. Then
(i) Any f R
A
can by uniformly approximated on K by polynomials.
(ii) If || > 2r then any polynomial can be uniformly approximated on K by functions in R
{}
.
Lemma 12.3 (Pole-Pushing Lemma). Suppose that K C is compact and V is a connected open
subset of C

\K. If V then any element of R


V
can by uniformly approximated on K by elements
of R
{}
.
Theorem 12.4. Suppose that K C is compact and , C

\K. Every function in R


{}
can be
uniformly approximated on K by elements of R

if and only if and lie in the same component of


C

\K.
Lemma 12.5. Suppose that D is an open subset of the plane. There exist compact sets K
j
(j N)
such that K
j
K

j+1
,
D =

j=1
K
j
,
and such that every component of C

\K
j
contains a component of C

\D.
Lemma 12.6. Suppose that is a smooth curve in the plane and f C(

). Dene F : C\

C
by
F(z) =

f(w)
w z
dw.
Suppose that K C\

is compact and > 0. Then there exist w


0
, . . . , w
n

such that

F(z)
n

j=1
f(w
j
)
w
j
z
(w
j
w
j1
)

< (z K).
Theorem 12.7 (Runges Thoerem for one compact set). Suppose that K C is compact and A
C

\K intersects every component of C

\K. If f is holomorphic in a neighborhood of K and > 0


then there exists R R
A
such that
|f(z) R(z)| < (z K).
Theorem 12.8 (Runges Theorem). Suppose that D C is open and A C

\D intersects every
component of C

\D. If f H(D) then there exists a sequence of rational functions (R


j
) R
A
such
that R
j
f uniformly on compact subsets of D.
Corollary 12.9. If D C is an open set such that C

\D is connected and f H(D) then there


exists a sequence of polynomials P
j
with P
j
f uniformly on compact subsets of D.
Corollary 12.10. If D C is a simply connected open set and f H(D) then there exists a sequence
of polynomials P
j
with P
j
f uniformly on compact subsets of D.
12.4. Show that there exists a sequence of polynomials (P
n
) such that P
n
(0) = 1 for all n, while
P
n
(z) 0 as n if z C and z = 0.
49
Proof. Let K
P
n
= D(0, n)\([(1/n, 0), (n, 1/n), (n, 1/n)]) and K
n
= K
P
n
{0} [1/n, n]. Now let
f
n
: K
n
C be given by
f
n
(z) =

0, z = 0,
1, z K
P
n
[1/n, n].
Then f
n
is holomorphic in a neighborhood of K
n
and C

\K
n
is connected, so theorem 12.7 shows
there exists a sequence of polynomials (P
n
) such that |P
n
f
n
| < 1/n in K
n
. So |P
n
| < 1/n on
K
P
n
[1/n, n] and |P
n
1| < 1/n on {0}. But if z C\ {0}, then there exists N N such that
z K
P
n
[1/n, n] for all n > N. Thus P
n
(z) 0 for all z C\ {0} and similarly P
n
(0) 1.
Since P
n
(0) 1, then there exists M N such that |P
n
(0) 1| < 1/2 for all n > M. Thus
P
n
(0) = 0 for all n > M, so we can dene

P
j
(z) = P
M+j
(z)/P
M+j
(0) for all j N. But now (

P
j
) is a
sequence of polynomials with the desired properties.
12.5. Show that there exists a sequence of polynomials (P
n
) such that P

n
(0) = 1 for all n, although
P

n
(z) 0 as n for all z C\ {0} and P
n
(z) 0 for all z C.
Proof. Let K
P
n
be the same as exercise 12.4, K
D
n
= D(0, d(0, K
P
n
)/2), and K
n
= K
P
n
K
D
n
[1/n, n].
Now let f
n
: K
n
C be given by
f
n
(z) =

z, z K
D
n
,
0, z K
P
n
[1/n, n].
Then theorem 12.7 states there exists P
n
such that |P
n
f
n
| < 1/n
2
for all n N. So if z C\ {0},
then there exists N N such that z K
P
n
[1/n, n] for all n > N. Thus |P
n
(z)| < 1/n
2
for all n > N.
Since z C\ {0}, there exists r > 0 such that D(z, r) C\ {0}. So Cauchys estimates show that
|P

n
(z)|
1
n
2
r
for all n > N.
Now if z = 0, then |P
n
(0) f
n
(0)| = |P
n
(0)| < 1/n
2
. Also, if z K
D
n
, then |P
n
(z) z| < 1/n
2
,
so Cauchys estimates state |P

n
(0) 1| < n/n
2
= 1/n. Thus P

n
(0) = 0 for all n > 2. Now dene

P
j
(z) = P
j+1
(z)/P

j+1
(0) for all j N. But now (

P
j
) is a sequence of polynomials with the desired
properties.
12.6. Fix a number M < . Show that there does not exist a sequence of polynomials (P
n
) such that
lim
n
P
n
(z) =

0 (z = 0),
1 (z = 0),
such that
|P
n
(z)| M (|z| 1, n N).
Proof. Let

P = P|
D
H(D) for P a polynomial and (P
n
) be a sequence of polynomials with

(

P
n
)


M. Since

(

P
n
)

M, exercise 9.4 shows that if (



P
n
) converges, then the limit function is in H(D).
Therefore there does not exist a sequence of polynomials with the given limit function.
Theorem 12.11 (Mittag-Leer Theorem). Suppose that D C is open and E D has no limit
point in D. Suppose that for each a E we are given a positive integer N
a
and a nite sequence of
complex numbers (c
a,j
)
Na
j=1
. Then there exists a function f meromorphic in D such that f has no poles
except at points of E and such that for each a E, f has a pole at a with principal part
Na

j=1
c
a,j
(z a)
j
.
12.7. Assume Runges theorem 12.8 and prove Cauchys theorem 4.10.
Proof. Suppose that V C is open, is a cycle in V , has the property that
Ind (, a) = 0
50 DAVID C. ULLRICH
for all a C\V , and f H(V ). Let A intersect every component of C

\V . Suppose that

R(z) dz =
0 for all rational functions R R
A
. Then by Runges theorem, we can take a sequence (R
j
) R
A
such that (R
j
) f uniformly on compact subsets of V . Thus

f(z) dz =

lim
j
R
j
(z) dz
= lim
j

R
j
(z) dz
= 0.
So we can assume that f = R R
A
. But now as in lemma 12.0,

R(z) dz =

P(z) +

aA
Na

j=1
c
a,j
(z a)
j

dz
=

P(z) dz +

aA
Na

j=1

c
a,j
(z a)
j
dz
=

aA

c
a,1
(z a)
1
dz
with the last equality provided by Cauchys theorem 2.0 for derivatives.
But now we have

R(z) dz =

aA
c
a,1

(z a)
1
dz
=

aA
c
a,1
Ind (, a)
= 0
because Ind (, a) = 0 for all a C\V . Therefore we have

f(z) dz = 0 for all f H(V ).


13. The Weierstrass Factorization Theorem
Lemma 13.0. For z D and n N we have
|1 E
n
(z)|

z
n+1

.
Theorem 13.1. Suppose that (z
j
)

j=1
is a sequence of nonzero complex numbers which tend to innity
as j ; suppose N is a non-negative integer. If (n
j
) is a sequence of non-negative integers such
that

j=1

r
|z
j
|

nj+1
<
for all r > 0 then the product
f(z) = z
N

j=1
E
nj

z
z
j

converges uniformly on compact subsets of the plane to an entire function with a zero of order N at
the origin, a zero at each z
k
and no other zeroes. (The order of the zero of f at z
k
is equal to the
number of times the value z
k
appears in the sequence (z
j
).)
Theorem 13.2 (Weierstrass Factorization Theorem for Entire Functions). Suppose that f H(C)
is nonconstant. Suppose that f has a zero of order N at the origin (0 N < ). Suppose that F
has M (0 M ) zeroes away from the origin and suppose that (z
j
)
M
j=1
is a sequence containing
51
the nonzero zeroes of f, each listed according to its multiplicity. If (n
j
)
M
j=1
is a sequence of positive
integers such that
M

j=1

r
|z
j
|

nj
<
for all r > 0 then there exists an entire function g such that
f(z) = e
g(z)
z
N
N

j=1
E
nj

z
z
j

for all z C; the product converges uniformly on compact subsets of the plane.
Theorem 13.3 (Weirstrass Factorization Theorem in an Open Set). Suppose that D C is open,
A D has no limit point in D, and m

is a positive integer for each A. There exists f H(D)


such that f has a zero of order m

at each A and no other zeroes.


Theorem 13.3 (Weierstrass Factorization Theorem in an Open Set, Part 2). Suppose that D C

is open, D = C

, A D has no limit point in D, and m

is a positive integer for each A. There


exists f H(D) such that f has a zero of order m

at each A and no other zeroes.


Theorem 13.4. If D is an open subset of the plane and f is meromorphic in D then there exist
functions g, h H(D) such that f = g/h.
Theorem 13.5 (Holomorphic Interpolation). Suppose that D C is open and A D has no limit
point in D. Suppose that for each A we are given a non-negative integer n

and a sequence of
complex numbers (c
,j
)
n
j=0
. There exists a function f H(D) such that
f
(j)
() = c
j,
( A, 0 j n

).
13.1. Suppose that D is a bounded and connected open subset of the plane. Show that there exists
f H(D) which cannot be extended to a function holomorphic in a strictly larger connected open set.
Proof. By lemma 12.5 there exist compact sets K
j
such that K
j
K

j+1
,
D =

j=1
K
j
.
Let C
j
=

D(z, 1/j) K
j+1
\K

j
| z K
j+1
\K

. Since K
j+1
\K

j
is compact, there exists a nite sub-
cover of C
j
, call it B
j
=

D
i
(z
i
, 1/j) K
j+1
\K

j
| z
i
K
j+1
\K

. Now let A
j
=

z
i
| D
i
(z
i
, 1/j) K
j+1
\K

j
B
j

and A =

j=1
A
j
. We show that D A while A consists of only isolated points in D.
Let b D, > 0, and consider D(b, ). But there exists N N such that if 2n > N, then
D(b, /2) K
2n
for some . So there exists m N such that K
m+1
\K

m
.
Since A has no limit points in D, let f be zero on A and nonzero everywhere else by the Weierstrass
factorization theorem. Now if

f H(E) with D E and

f|
D
= f, then there exists b D E.
Since

f is continuous,

f(b) = 0 so Z(

f) has a limit point. Therefore any holomorphic extension of f
is 0 on D contradicting the fact that f is nonzero on D\A.
13.2. Suppose that D is an open set in the plane and E and F are disjoint subsets of D, neither of
which has a limit point in D. Show that there exists a function f meromorhpic in D so that f has a
simple pole at every point of E, a simple zero at every point of F, and no other poles or zeroes.
Proof. Since E D has no limit point in D, let m
e
= 1 for all e E so theorem 13.3 states there
exists g H(D) such that g has a simple zero at each e E and no other zeroes. Similarly there
exists h H(D) with simple zeroes at each f F and no other zeroes. Since E and F are disjoint,
f = h/g is meromorphic in D with a simple pole at every point of E and a simple zero at every point
of F with no other poles or zeroes.
52 DAVID C. ULLRICH
13.3. Suppose that D is a simply connected open subset of the plane and f H(D) is not constant.
Show that there exists g H(D) such that f = g
2
if and only if every zero of f has even order.
Proof. Suppose there exists g H(D) such that f = g
2
and let z
0
D be a zero of f, so z
0
is a zero of
g of order M. Thus there exists h H(D) such that h(z
0
) = 0 and g(z) = (z z
0
)
M
h(z) for all z D.
But then f(z) = (g(z))
2
= ((z z
0
)
M
h(z))
2
= (z z
0
)
2M
(h(z))
2
. Since h
2
H(D) with h
2
(z
0
) = 0,
f has a zero of order 2M at z
0
.
Now suppose every zero of f has even order. Since f = 0, then there are no limit points of Z(f)
in D. For every Z(f), let 2m

be the order for the zero of f at . By Weirstrass factorization


theorem, there exists h H(D) such that h has a zero of order m

for each D and no other zeroes


in D. Thus h
2
has a zero of order 2m

for each Z(f) with no other zeroes. Let F H(D) be the


holomorphic function such that F|
D\Z(f)
= f/h
2
. Then let g = hF so that
f = h
2
F
2
= (hF)
2
= g
2
.

16. Analytic Continuation


16.0. Introduction.
16.1. Continuation Along Curves.
16.1. Suppose that g, h O
a
. Show that g = h if and only if
g
(k)
(a) = h
(k)
(a), k Z
+
.
Proof. First suppose that g = h and let (f, D) g. Then
g
(k)
(a) = f
(k)
(a) = h
(k)
(a).
Now suppose that g
(k)
(a) = h
(k)
(a) for all k Z
+
. If (g
1
, D) g and (h
1
, E) h, then g
(k)
1
(a) =
h
(k)
1
(a) for all k Z
+
. Thus there is some D(a, r) such that
g
1
(z) =

k=0
g
(k)
1
(a)
k!
(z a)
k
=

k=0
h
(k)
1
(a)
k!
(z a)
k
= h
1
(z)
for all z D(a, r). By denition g
1
(z)
a
h
1
(z) so g = h.

You might also like