Local Properties

1 Cauchy’s integral formula

We begin with a simple lemma:
1
Lemma 1. Let a ∈ C be fixed and let A be a set of finite measure. Then z−a is integrable
on A.

Proof. Let B := {z ∈ A : |z − a| > 1}. Then

Z Z
dm(z)
6 dm(z) 6 m(A) < ∞.
B |z − a| B

On the other hand,

Z Z Z 1 Z 2π
dm(z) dm(z) rdrdθ
6 = = 2π.
A\B |z − a| |z−a|61 |z − a| 0 0 r

We now prove the Cauchy–Green formula. An important consequence of the Cauchy–Green

formula is Cauchy’s integral theorem.

Theorem 2 (Cauchy–Green formula). Let U ⊂ C be a bounded domain with piece-wise reg-

ular boundary given positive orientation. Let f be a R-differentiable function on some
neighbourhood of U and ∂f be continuous on U. Then for each z ∈ U, we have:
Z Z
1 f(w) 1 ∂f(w)
f(z) = dw − dm(w)
2πi ∂U w−z π U w−z

Proof. THe proof is an application of Green’s theorem on the form ρ = f(w) w−z dw. However,
this form has a singularity at the point w = z. To deal with this issue, we delete from U a
small disk D(z, ε) where ε is chosen so small that the closed disk is fully-contained on U and
set Uε := U \ D(z, ε). Observe that Uε is also domain with piece-wise regular boundary and
if we orient C(z, ε) with the clockwise orientation, then ∂Uε would have partial orientation.
Write ρ as f(w) f(w)
w−z dx + i w−z dy, then rho is of the form Pdx + Qdy where P = w−z and
f(w)

f(w)
Q= w−z and thus
   
∂ f(w) ∂ f(w) ∂f 1 ∂f 1
Qx − Py = i − = 2i = .
∂x w−z ∂y w − z ∂w 2i ∂w w − z

1
 2 Holomorphic functions are analytic

The RHS of the above is continuous function on Uε . Thus we can apply Green’s theorem
to conclude that Z Z
∂f 1
ρ = 2i dm(w).
∂Uε Uε ∂w w−z
Now,
Z Z Z
rho = + ρ
∂Uε ∂U C(z,ε)
Z Z 2π
f(w)
= dw − i f(z + εiθ )dθ,
∂U w−z 0
R
and this converges to ∂U f(w)
w−z − 2πif(z) as ε → 0.
1 ∂f 1
The function w 7→ w−z is integrable on U and therefore so is ∂w w−z . An application of
the dominated convergence theorem gives that
Z Z
∂f 1 ∂f 1
→
Uε ∂w w − z U ∂w w − z

which proves the result.

Corollary 3. Let U ⊂ C be a bounded domain with piece-wise regular boundary given positive
orientation. Let f be holomorphic in a neighbourhood of U. Then for each z ∈ U:
Z
1 f(w)
f(z) = dm(w).
2πi ∂U w − z

2 Holomorphic functions are analytic

We now come to the proof of one of the central facts of the subject. Complex-analytic and
holomorphic are equivalent concepts.

Theorem 4. Let f ∈ H(D(a, R)) and let

Z
1 f(w)
cn := dw, r < R.
2πi C(a,r) (w − z)n+1
P
Then the cn ’s are independent of r and the power series cn (z − a)n converges to f(z)
for z ∈ D(a, R).

Proof. For r < R and |z − a| < r, Cauchy’s integral formula gives:

Z
1 f(w)
f(z) = dw.
2πi C(a,r) w − z
1
We will now expand the Cauchy kernel z−w as a power series about the point a. This is
easily done as we did for proving that rational functions are analytic

1 1 1 X (w − a)n
= = w−a

=− .
w−z (w − a) − (z − a) z−a − 1 (z − a) (z − a)n+1

2
 3 The principle of analytic continuation

|z−a|n
Now fix z with |z − a| < r. Then |f(w)| · |w−a|n+1 6 |f(w)||ζ|
n where |ζ| < 1 (independent of w)
P (z−a)n
and therefore by the Weirstrass M-test, it follows that f(w) (w−a) n+1 converges uniformly
f(w)
on C(a, r) to w−z . Thus, we can interchange the integral and summation in the following:
Z
1 f(w)
f(z) = dw
2πi C(a,r) w − z
Z X XZ
1 (z − a)n (z − a)n
= f(w) = f(w) dw
2πi C(a,r) (w − a)n+1 C(a,r) (w − a)n+1
P
This shows that the power series cn (z − a)n converges to f(z) on D(a, r). Any two power
series development of f around the point a must be equal. This shows that the coefficients cn
P
are independent of r. By taking r → R, it follows that the radius of convergence of cn (z−a)n
is at least R.

This proves that holomorphic functions are complex-analytic and hence infinitely differen-
tiable in both the real and complex sense. Another consequence of the above theorem is the
following extension of Cauchy’s integral formula.

Corollary 5. Let f ∈ H(D(a, R)) then for each 0 < r < R, we have
Z
(n) 1 f(w)
f (a) = .
2πi C(a,r) (w − a)n+1

Proof. This follows immediately from Taylor’s theorem and comparing coefficients.

The following theorem can be viewed as a converse of Cauchy’s theorem.

Theorem 6 (Morera). Let f be a continuous function on the open set U. Then f ∈ H(U) iff
R
∂4 f(z)dz = 0 for any closed 4 ⊂ U.

Proof. One implication follows from Cauchy’s theorem. On the other hand, if it is true that
R
∂4 f(z)dz = 0 for any triangle, then f locally has anti-derivatives. This means that f is
holomorphic.

3 The principle of analytic continuation

We will now study some of the aspects of holomorphic functions that arise out of complex-
analyticity. One main distinguishing feature of analytic functions as opposed to C∞ functions
is that analytic functions are more rigid. This is made precise in this section. We first define
the order of a holomorphic function at a point.

Lemma 7. Let U ⊂ C be a domain, a ∈ U and let f ∈ H(U). Then the following conditions
are all equivalent:

(i) f(n) (a) = 0 ∀n.

(ii) f(z) = 0 in a neighbourhood of a.
(iii) f ≡ 0 on U.

3
 3 The principle of analytic continuation

Proof. Obviously (c) =⇒ (b), (b) =⇒ a. It suffices to prove that (a) =⇒ (c). To do this
we use a connectedness argument. Let

A := {z ∈ U : f(z) = 0 in some neighbourhood of z}.

Clearly, A is an open set. Let zm ∈ A be such that zm → z ∈ U. Then f(n) (zm ) = 0 by the
definition of A and consequently by passing to limits, f(n) (z) = 0. But this means that the
Taylor expansion of f centred at 0 has all coefficients 0 and this means that f vanishes on some
disk around z and hence z ∈ A. This proves that A is closed in U and consequently A = U
proving that (a) =⇒ (c).

Definition 8. Let U ⊂ C be a domain and f ∈ H(U). Suppose f 6≡ 0 and f(a) = 0 a ∈ U. Then

we can find a smallest m > 0 such that f(m) 6= 0. This m is called the order of f at a. We
also say that f vanishes to order m at the point a.

If f vanishes to order m at the point a then it follows that the Taylor series of f around the
point a is of the form
f(z) = cm (z − a)m + . . . , cm 6= 0.
Thus, we can write
f(z) = (z − a)m g(z),
in a neighbourhood of a, where g(z) = cm + cm+1 (z − a) + . . . . If for some z, the series
above converges then so does the series for f as (z − a)n is an analytic function with radius of
convergence ∞. Conversely, if f(z) converges for some z then so does the series for g(z). This
shows that the function g is analytic and therefore holomorphic on the disk of convergence
D(a, R). Moreover, shrinking this disk, we may assume that D(a, R) ⊂ U and that g does not
take the value 0 in the disk D(a, R). Now, define

Z(f) := {z ∈ U : f(z) = 0},

If f 6≡ 0 then the above shows that around any point a ∈ Z(f), we can find a disk around which
f(z) = (z − a)n g(z) and g has no zeroes in D(a, R). This proves that on D(a, R) the function
f takes the value 0 only at a. Consequently we have proved the

Theorem 9. Let f be holomorphic and not identically 0 on a domain U. Then Z(f) is a

closed and discrete subset of U. Consequently, on any compact subset of U, f has only
finitely many zeroes and only countably many zeroes on U.

We can summarize everything we have established above with the following theorem.

Theorem 10 (Principle of analytic continuation). Let f, g be holomorphic on a domain U.

Then f ≡ g iff any one of the following equivalent conditions holds:

1. For some point a ∈ U, f(n) (a) = g(n) (a), n = 0.1, . . . .

2. For some indiscrete set A ⊂ U, f|A = g|A .
3. There is an open set V ⊂ U such that F|U = G|V .

We now give an application of the above results.

4
 4 The open mapping theorem

Definition 11. A domain U ⊂ C is said to be symmetric about the real-axis if z ∈ U whenever

z ∈ U.

Theorem 12 (Schwarz reflection principle). Let U ⊂ C be symmetric about the real-axis and
f ∈ H(U) and f takes real values on R ∩ U. Then f(z) = f(z).

Proof. Let g(z) = f(z). It is easy to see that g is also holomorphic on U. But on R ∩ U, g
agrees with f and consequently by the principle of analytic continuation g ≡ f.

4 The open mapping theorem

Let f : U → C be holomorphic and nowhere vanishing. We have shown that f has a continuous
branch of the logarithm iff f 0 /f has an anti-derivative. We have also proved that on any star-
shaped domain, any holomorphic function has an antiderivative. But from the results above,
f 0 /f is a holomorphic function and consequently we get:

Proposition 13. On a convex open set, any nowhere vanishing holomorphic function admits
a continuous branch of the logarithm and therefore a continuous branch of the argument.
This implies that the function also admits a branch of the n-th root function. Every
holomorphic function on a domain locally admits a continuous branch of logarithm,
argument and n-th root.

Now we will study the local behaviour of holomorphic functions. Suppose f(a) = b and the
function f(a) − b vanishes to order m at a. This means that

f(z) − b = (z − a)m g(z),

in a disk D := D(a, R) and g has no zeroes in D. From the previous proposition, we can find
a function h ∈ H(D) such that hm = g. This means that

f(z) = ((z − a)h(z))m := f1 (z)m . (4.1)

We see that the function f1 (z) satisfies f1 (a) = 0 and f10 (a) = h(a) 6= 0. By the inverse
function theorem, it follows that f1 is a local-homeomorphism. We can also see this directly.
By translating and scaling, we may assume that a = 0 and g10 (0) = 1. Thus the power-series
expnasion of f1 is of the form
z + f2 (z),
where f20 (0) = 0. Fix 0 < ε < 1, we can find a r < R such that for z ∈ D(0, r), |f20 (z)| < ε. For
z, w ∈ D(0, r), by the fundamental theorem of complex calculus, we see that

|f2 (w) − f2 (z)| 6 ε|z − w|.

The above inequality shows that

(1 − ε)|z − w| 6 |f1 (z) − f1 (w)| 6 (1 + ε)|z − w|, z, w ∈ D(0, r).

This shows that the map f1 is injective on D(0, r). We will now show that every ζ with |ζ|
sufficiently small has a preimage in D(0, r). If f1 (z) = ζ then z is a fixed point of ζ − f2 (z).

5
 5 Maximum principle and applications

We define a sequence as follows: let z0 = 0 and zn+1 = ζ − f2 (zn ). Now, |zn+1 − zn | =

|f2 (zn ) − f2 (zn−1 )| 6 ε|zn − z − n − 1| and we get

|zn+1 − zn | 6 εn |z1 − z0 | = εn |ζ|.

This shows that the sequence (zn ) is Cauchy and therefore converges Now,

X
n
1
|zn | 6 |zi − zi−1 | 6 |ζ| .
1−ε
i=1

If |ζ|leqr(1 − ε), it follows that we can find z ∈ D(0, r) such that f1 (z) = ζ. This proves that
f1 is a local-homeomorphism of some neighbourhood V of 0 onto a disk D(0, s). The inverse
map f−1 is automatically a holomorphic map by previous results. From (4.1), it follows that
the image f(V) is exactly D(b, sm ). This proves the

Theorem 14 (Local mapping theorem). Let f be holomorphic in neighbourhood of the point

a ∈ C and suppose that f(a) = b with multiplicity m. Then we can find a neighbourhood V
of a and a disk D(b, δ) such that f(v) = D(b, δ) and furthermore each point ζ ∈ D(b, δ)\{b}
has exactly m-preimages under f in V.

Remark 15. f is locally injective at a iff f 0 (a) 6= 0. Only one implication is true for real
variable functions and this implication follows from the inverse function theorem. Note that
the function ez has non-vanishing derivative at each point of C but is not injective globally.
Observe from the proof that δ = sm and

Theorem 16 (The open mapping theorem). Let u ⊂ C be a domain and let f ∈ H(U) be
non-constant. Then f is an open map.

Proof. Let V ⊂ U be open. Then f has finite multiplicity at each point of V. From the
previous theorem it follows that f(V) contains some disk D(b, δ).

5 Maximum principle and applications

Theorem 17 (Maximum principle). If f is holomorphic on a domain U and is non-constant
then |f| cannot have a local maximum on U.

Proof. If f attains a local maximum at a ∈ U then f cannot be an open map near a.

Corollary 18. Let U ⊂ C be a bounded domain and f ∈ C( U) ∩ H(U). Then the maximum of
|f| on U is attained on ∂U.

Proof. If the maximum is attained at a ∈ U then a is local maximum of |f|.

If f ∈ H(U) and has no zeroes then 1/f ∈ H(U) and applying the maximum principle for
1/f, we conclude that if m 6 |f| 6 M on ∂U then m 6 |f|leqM on U.
Using Cauchy’s integral formula, we can get nice estimates on the derivatives of a holomor-
phic function.

6
 5 Maximum principle and applications

Theorem 19 (Cauchy’s inequalities). Let f ∈ H(D(a, R)) and |f(z) 6 M whenever z ∈ C(a, R)
then
n!
|f(n) (a)| 6 M n , n = 0, 1, . . .
R
Proof. By Cauchy’s integral formula
Z 2π Z 2π
n! f(a + Reiθ ) n!
f(n) (a) = iReiθ dθ = f(a + Reiθ )e−inθ dθ,
2πi 0 Rn+1 ei(n+1)θ 2πRn 0

from which the result follows immediately.

Theorem 20 (Liouville). An entire function that is bounded is constant.

Proof. Apply Cauchy inequalities on larger and larger disks.

Theorem 21 (The fundamental theorem of algebra). Let P(z) = a0 + a1 z + · · · + an zn be a

polynomial whose degree n > 1. Then P has a zero in C.

Proof. Consider f(z) = 1/P(z). If P has no zeroes then f is well-defined and holomorphic on
C. But  an−1 a0 
P(z) = zn an + ) + ··· + n ,
z z
and it is clear that as |z| → ∞, |P(z)| → ∞. This means that the function f is bounded and
thus by Liouville’s theorem is constant which is absurd as the degree of P is higher than 1.

Now, we shall see an interesting consequence of the maximum principle that has a number
of interesting applications.

Theorem 22 (Schwarz lemma). Let D be the open unit disk in the complex plane C centered
at the origin and let f : D → D be a holomorphic map such that f(0) = 0. Then,
|f(z)| 6 |z| ∀z ∈ D and |f 0 (0)| 6 1.
Moreover, if |f(z)| = |z| for some non-zero z or |f 0 (0)| = 1, then f(z) = az for some
a ∈ C with |a| = 1.

Proof. Consider the function  f(z)

z if z 6= 0
g(z) =
f 0 (0) if z = 0.
Then g ∈ H(D) as g is rational with nowhere zero denominator on D \ {0} and f 0 (0) exists. On
the closed disk D(0, r), the maximum principle implies that we can find an zr ∈ C(0, R) such
that |g(z)| 6 |g(zr )| in D(0, r). This means that |g(z)| 6 |f(z r )| 1
|zr | 6 r on D(0, r). Taking r → 1
shows that |g(z)| 6 1 and this delivers the first part of the result.
Moreover, if |f(z)| = |z| or |f 0 (0)| = 1 then |g(z)| = 1 for some z ∈ D and as |g(z)|leq1 on D,
the maximum principle shows that g ≡ 1 on D.

Let U ⊂ C be a domain. We define

Aut(U) := {f : U → U : f ∈ H(U), f is bijective}.

7
 5 Maximum principle and applications

If f ∈ Aut(U) then by the local mapping theorem it follows that f 0 does not vanish at any
point of U. Consequently, f−1 ∈ Aut(U) as well. From this, it is easy to show that Aut(U) is
group under composition of functions. Note also that the group Aut(U) acts naturally on the
domain U by (g, z) 7→ g(u).

Theorem 23. The automorphism group of the unit disk is the set of linear fractional trans-
formations of the form
z−a
φa,α := eiα , a ∈ D, α ∈ R.
1 − az

Proof. Let z = eiθ . Then

eiθ − a i(α−θ) e
iθ − a
φa,α (eiθ ) = eiα = e ,
1 − ae−iθ e−iθ − a

which proves that |φa,α (eiθ )| = 1. But φa,α (a) = 0 and consequently φa,α maps the unit
disk onto itself. This proves that φa,α ∈ Aut(D).
Now let f ∈ Aut(D) and let f(0) = a. Then g := φa,0 ◦ f satisfies g(0) = 0. Therefore
|g(z)| 6 |z|. It is clear that |g 0 (0)| = 1 and consequently by Schwarz lemma g(z) = eiθ z is just
a rotation and hence
φa,0 ◦ f(z) = eiθ z
which proves that f(z) = φ−1 iθ −1
a,0 (e z). But an easy computation shows that φa,0 = φ−a,0 and
thus
eiθ z + a
f(z) = = φ−eiθ a,eiθ .
1 + aeiθ z