If f(c)=0,f(c)=0, then f(z)=0f(z)=0 for all z∈Bϵ(c).z∈Bϵ(c). Hence, by the Identity Theorem
((Reference)), f(z)f(z) would equal 0 for all z∈S.z∈S.
so, we may as well assume that f(c)≠0.f(c)≠0.
Let rr be any positive number for which the closed disk B¯r(c)B¯r(c) is contained in Bϵ(c).Bϵ(c).
We claim first that there exists a point zz on the boundary CrCr
of the disk B¯r(c)B¯r(c) for which |f(z)|=|f(c)|.|f(z)|=|f(c)|.
Of course, |f(z|≤|f(c)||f(z|≤|f(c)| for all zz on this boundary by assumption.
By way of contradiction, suppose that
|f(ζ)|<|f(c)||f(ζ)|<|f(c)| for all ζζ on the boundary CrCr of the disk.
Write MM for the maximum value of the function |f||f| on the compact
set Cr.Cr. Then, by our assumption, M<|f(c)|.M<|f(c)|.
Now, we use the Cauchy Integral Formula:
|
f
(
c
)
|
=
|
1
2
π
i
∫
C
r
f
(
ζ
)
ζ
-
c
d
ζ
|
=
1
2
π
|
∫
0
2
π
f
(
c
+
r
e
i
t
)
r
e
i
t
i
r
e
i
t
d
t
|
≤
1
2
π
∫
0
2
π
|
f
(
c
+
r
e
i
t
)
|
d
t
≤
1
2
π
∫
0
2
π
M
d
t
=
M
<
|
f
(
c
)
|
,
|
f
(
c
)
|
=
|
1
2
π
i
∫
C
r
f
(
ζ
)
ζ
-
c
d
ζ
|
=
1
2
π
|
∫
0
2
π
f
(
c
+
r
e
i
t
)
r
e
i
t
i
r
e
i
t
d
t
|
≤
1
2
π
∫
0
2
π
|
f
(
c
+
r
e
i
t
)
|
d
t
≤
1
2
π
∫
0
2
π
M
d
t
=
M
<
|
f
(
c
)
|
,
(1)and this is a contradiction.
Now for each natural number nn for which 1/n<ϵ,1/n<ϵ, let znzn be a point
for which |zn-c|=1/n|zn-c|=1/n and |f(zn)|=|f(c)|.|f(zn)|=|f(c)|.
We claim that the derivative f'(zn)f'(zn) of ff at zn=0zn=0 for all n.n.
What we know is that the real-valued function
F(x,y)=|f(x+iy)|2=(u(x,y)2+(v(x,y))2F(x,y)=|f(x+iy)|2=(u(x,y)2+(v(x,y))2 attains a local maximum value at zn=(xn,yn).zn=(xn,yn).
Hence, by (Reference), both partial derivatives of FF must be 0 at (xn,yn).(xn,yn). That is
2
u
(
x
n
,
y
n
)
t
i
a
l
u
t
i
a
l
x
(
x
n
,
y
n
)
+
2
v
(
x
n
,
y
n
)
t
i
a
l
v
t
i
a
l
x
(
x
n
,
y
n
)
=
0
2
u
(
x
n
,
y
n
)
t
i
a
l
u
t
i
a
l
x
(
x
n
,
y
n
)
+
2
v
(
x
n
,
y
n
)
t
i
a
l
v
t
i
a
l
x
(
x
n
,
y
n
)
=
0
(2)and
2
u
(
x
n
,
y
n
)
t
i
a
l
u
t
i
a
l
y
(
x
n
,
y
n
)
+
2
v
(
x
n
,
y
n
)
t
i
a
l
v
t
i
a
l
y
(
x
n
,
y
n
)
=
0
.
2
u
(
x
n
,
y
n
)
t
i
a
l
u
t
i
a
l
y
(
x
n
,
y
n
)
+
2
v
(
x
n
,
y
n
)
t
i
a
l
v
t
i
a
l
y
(
x
n
,
y
n
)
=
0
.
(3)Hence the two vectors
V
→
1
=
(
t
i
a
l
u
t
i
a
l
x
(
x
n
,
y
n
)
,
t
i
a
l
v
t
i
a
l
x
(
x
n
,
y
n
)
)
V
→
1
=
(
t
i
a
l
u
t
i
a
l
x
(
x
n
,
y
n
)
,
t
i
a
l
v
t
i
a
l
x
(
x
n
,
y
n
)
)
(4)and
V
→
2
=
(
t
i
a
l
u
t
i
a
l
y
(
x
n
,
y
n
)
,
t
i
a
l
v
t
i
a
l
y
(
x
n
,
y
n
)
)
V
→
2
=
(
t
i
a
l
u
t
i
a
l
y
(
x
n
,
y
n
)
,
t
i
a
l
v
t
i
a
l
y
(
x
n
,
y
n
)
)
(5)are both perpendicular to the vector V→3=(u(xn,yn),v(xn,yn)).V→3=(u(xn,yn),v(xn,yn)).
But V→3≠0,V→3≠0, because ∥V→3∥=|f(zn)|=|f(c)|>0,∥V→3∥=|f(zn)|=|f(c)|>0,
and hence V→1V→1 and V→2V→2 are linearly dependent.
But this implies that f'(zn)=0,f'(zn)=0, according to (Reference).
Since c=limzn,c=limzn, and f'f' is
analytic on S0,S0,
it follows from the Identity Theorem that there exists an r>0r>0 such that f'(z)=0f'(z)=0 for all z∈Br(c).z∈Br(c).
But this implies that ff is a constant f(z)=f(c)f(z)=f(c) for all z∈Br(c).z∈Br(c).
And thenm, again using the Identity Theorem, this implies that f(z)=f(c)f(z)=f(c) for all z∈S,z∈S,
which completes the proof.