2 Holomorphic Functions: the Cauchy-Riemann equa-

tions and examples

Lecture 5
Denition 2.1. Let D be a domain in C, f : D C and let a D. If
lim
0
f(a + ) f(a)

exists (here C) then it is denoted by

df
dz
(a) or f

(a) and called the complex

derivative of f at a.
The complex derivative satises many of the same properties as the usual
derivative for functions dened on open subsets of R.
(i) (cf)

(a) = cf

(a) if c C;
(ii) (f + g)

(a) = f

(a) + g

(a);
(iii) (f.g)

(a) = f

(a)g(a) + f(a)g

(a);
(iv)

f
g

(a) =
f

(a)g(a) f(a)g

(a)
g(a)
2
;
(v) (f g)

(a) = f

(g(a))g

(a).
Example 2.1. Let f(z) = z. Then
f(a + ) f(a)

=
a + a

= 1.
We usually write
d(z)
dz
= 1.
Example 2.2. Let g(z) = z
n
where n N. Then
(a + )
n
a
n

=
a
n
+ na
n1
+
n(n1)
2
a
n2

2
+ +
n
a
n

= na
n1
+
n(n 1)
2
a
n2
+

n
3

a
n3

2
+ +
n1
na
n1
Hence g

(a) = na
n1
.
12
Example 2.3. Let h(z) = z. Then
a + a

=
re
i
re
i
= e
2i
if = re
i
.
Let r 0 so 0. We now see that the general limit
lim
0
a + a

,
where C, does not exist as the limit above depends on . Therefore, the function
h(z) = z does not have a complex derivative at any a C.
Denition 2.2. Let D be a domain in C and let f : D C. If
df
dz
(a) exists for all
a D then f is said to be holomorphic on D.
NOTE: To say that a function is holomorphic is to say that the complex
derivative exists at all points of an open connected set its domain. It makes
sense to say that a function is holomorphic near a, that is, on some B

(a).
Let f(z) = f(x + iy) := f(x, y) using the correspondence C R
2
. We write
f
x
(a) := lim
0
R
f(a + ) f(a)

,
when this limit exists.
Note that if
df
dz
(a) exists than so does
f
x
(a) and
df
dz
(a) =
f
x
(a).
We write
f
y
(a) := lim
0
R
f(a + i) f(a)

when this limit exists. If

df
dz
(a) exists then
df
dz
(a) = lim
0
R
f(a + i) f(a)
i
=
1
i
f
y
(a).
So
df
dz
(a) = i
f
y
(a).
13
Now we decompose the function f into real and imaginary parts,
f(z) = f(x + iy) := u(x, y) + iv(x, y),
and assume that f

(z) exists for all z in some domain D. Then

f

(z) =
f
x
(z)
= lim
0
f(x + + iy) f(x + iy)

= lim
0
u(x + , y) + iv(x + , y) (u(x, y) + iv(x, y))

=
u
x
(x, y) + i
v
x
(u, y).
Similarly,
f

(z) = i
f
y
(z) = i

u
y
(x, y) + i
v
y
(x, y)

.
Equating real and imaginary parts (and using more compact notation):
u
x
= v
y
and u
y
= v
x
.
These are known as the CauchyRiemann equations (C-R equations) and
we have just proved:
Theorem 2.1. A holomorphic function f(x + iy) = u(x, y) + iv(x, y) satises the
CauchyRiemann equations
u
x
= v
y
u
y
= v
x
.
A more modern formulation goes as follows. Note rst that
df
dz
=
1
2

f
x
i
f
y

.
Denition 2.3. Dene

z
:=
1
2

x
i

y

z
:=
1
2

x
+ i

y

.
14
The CauchyRiemann equations as stated in the theorem come from equat-
ing the real and imaginary parts of the equation (valid for holomophic equa-
tions)
f
x
= i
f
y
that is,
f
x
+ i
f
y
= 0.
So the CauchyRiemann equations may be formulated as:
f
z
= 0.
Theorem 2.2. Let D be a domain and let f(x+iy) := u(x, y)+iv(x, y), be a function
such that u
x
, u
y
, v
x
, v
y
all exist and are continuous on D. If they satisfy the Cauchy
Riemann equations then the function f is holomorphic.
We will not give a proof of this theorem in these lectures but a proof may
be found in Priestley p59.
A more general characterization of complex differentiablity and holomor-
phy can be given in terms of the real derivative. Let us consider z = (x, y) as a
vector in D R
2
and f(z) = (u(x, y), v(x, y)) as a function f : D R
2
. Then f
is called real-differentiable at z
0
D if there exists a matrix A R
22
(called
Jacobi matrix) so that
f(z) = f(z
0
) + A (z z
0
) + r(z, z
0
)
where r(z, z
0
) is a remainder term so that r(z, z
0
)/z z
0
0 as z z
0
. In
this case, all rst partial derivatives exist, and A is necessarily given by
A =

u
x
u
y
v
x
v
y

.
Note, however, that existence of partial derivatives alone does not imply real-
differentiability. Observe that the matrix-vector multiplication A(zz
0
) can be
expressend in terms of a complex multiplication a(zz
0
) where a = a
1
+ia
2
C
only if A has the form
A =

a
1
a
2
a
2
a
1

which is equivalent to the CauchyRiemann equations. Then we have a =

f
z
and f is complex differentiable at z
0
. We deduce that if f : D C is real-
differentiable on D and satises
f
z
= 0, then f is holomorphic.
15
Example 2.4. Let f(z) =

n=0
a
n
z
n
be a function given by a power series with
radius of convergence R, then f : B
R
(0) C and we will show that f is holomorphic
on the domain B
R
(0).
Lemma 2.1. The series

n=0
a
n
z
n
and

n=1
na
n
z
n1
have the same radius of con-
vergence.
Proof. Let R be the radius of convergence of

n=0
a
n
z
n
and R

be the radius of
convergence of the derived series

n=1
na
n
z
n1
.
Let |z| < R and |z| < r < R. Then

|a
n
r
n
| converges and |a
n
r
n
| 0 so
there exists M > 0 such that |a
n
r
n
| < M for all n N {0}. Now
|na
n
z
n1
| = |a
n
r
n
|.
n
r
.

z
r

n1
M.
n
r
.

z
r

n1
.
Applying the ratio test: as n ,
n+1
r

z
r

n
n
r

z
r

n1
=
n + 1
n

z
r

z
r

< 1.
Therefore,

n
r

z
r

n1
is convergent and

|na
n
z
n1
| converges by compari-
son. Hence, R

R.
Note that |a
n
z
n
| |na
n
z
n1
||z| if n 1. If

|na
n
z
n1
| converges then so
does

|na
n
z
n
| and so does

|a
n
z
n
| by comparison. Therefore, R

R and
hence R = R

.
Lecture 6
Theorem 2.3. Let f : B
R
(0) C be a function given by a power series,
f(z) =

n=0
a
n
z
n
,
of radius R. Then f is holomorphic on the domain B
R
(0).
Proof. Let
f(z) =

n=0
a
n
z
n
,
g(z) =

n=1
na
n
z
n1
h(z) =

n=2
n(n 1)a
n
z
n2
.
16
All these series have the same radius of convergence. Let z B
R
(0), 0 |z| < < R
and C be such that || < |z|. Let
I() =
f(z + ) f(z)

g(z).
We will show that I() 0. This will prove that f

(z) exists and equals g(z).

Note
|I()| =

n=0
a
n

(z + )
n
z
n

nz
n1

and
(z + )
n
z
n

nz
n1
=

z
n
+ nz
n1
+

n
2

z
n2

2
+ +
n
z
n

nz
n1

n
2

z
n2
+

n
3

z
n3

2
+ +
n1

k=2
n!
k!(n k)!
z
nk

k1

= n(n 1)||

k=2
1
k(k 1)
(n 2)!
(k 2)!(n k)!
z
nk

k2

= n(n 1)||

n2

l=0
1
(l + 2)(l + 1)

n 2
l

z
n2l

by putting k = l + 2
n(n 1)||
n2

l=0

n 2
l

|z|
n2l
||
l
= n(n 1)|| (|z| +||)
n2
= n(n 1)||
n2
,
which shows that
|I()| ||

n=0
|a
n
n(n 1)
n2
|.
Therefore, I() 0 as 0 and the proof is complete.
Similarly, if f(z) =

n=0
c
n
(z a)
n
, a power series about the point a, con-
verges for |z a| < R then f is holomorphic on B
R
(a).
This gives us many examples of holomorphic functions.
17
Example 2.5. Let c C and dene
exp(cz) = e
cz
=

n=0
c
n
n!
z
n
.
The function exp(cz) is holomorphic on C.
Example 2.6. The function
sin z =
1
2i

e
iz
e
iz

n=0
(1)
n
z
2n+1
2n + 1!
is holomorphic on C.
Example 2.7. The function
cos z =
1
2i

e
iz
+ e
iz

n=0
(1)
n
z
2n
2n!
is holomorphic on C.
Example 2.8. The function
tan z =
sin z
cos z
is holomorphic on the domain C \ {z C : cos z = 0} = C \ {(n +
1
2
) : n Z}.
Example 2.9. The function
sinh z =
1
2

e
z
e
z

is holomorphic on C.
Example 2.10. The function
cosh z =
1
2

e
z
+ e
z

is holomorphic on C.
Example 2.11. The function
tanh z =
sinh z
cosh z
is holomorphic on C \ {(n +
1
2
)i : n Z}.
18
A key function is given by the extension of the logarithm function from
the positive real axis to domains in C. To describe this multifunction we need
to note that exp is periodic in C of period 2i.
We know that e
z
.e
w
= e
z+w
by manipulation of power series.
SKETCH:

n=0
1
n!
z
n
.

n=0
1
n!
w
n
=

k=0

n+m=k
1
n!
1
m!
z
n
w
m

k=0
1
k!

n+m=k
k!
n!m!
z
n
w
m

k=0
1
k!
(z + w)
k
,
(the interchange of limits at is justied because the series converge abso-
lutely).
Then e
z+w
= e
z
if e
w
= 1. Put w = x + iy. Then e
w
= e
x+iy
= e
x
.e
iy
, and
e
w
= 1 implies e
x
= 1 so x = 0, also e
iy
= cos y + i sin y = 1 implies y = 2n for
some n Z. Thus e
w
= 1 w = 2ni. It follows that
e
z+2ni
= e
z
.
We have shown that the function f(z) = e
z
has period 2i and so it is not a
one-to-one function.
However, e
z
is one-to-one when restricted to the domain
D = {z C : 0 < Imz < 2}
and
exp(D) = {z C \ {0} : 0 < arg z < 2} =:

D.
The inverse of exp |
D
is a branch of the multifunction we denote by log.
Lecture 7
To become invertible the function exp needs to be restricted to a suitable do-
main. Say
D

= {z C : < Im z < + 2i}.

Then
exp(D

) = {z C \ {0} : < arg z < + 2} =:

D

and exp |
D

: D

is invertible. By the Inverse Function Theorem for holo-

morphic functions (set on a problem sheet), the inverse is also holomorphic
giving a branch of the multifunction log. Let us put this into a denition:
19
Denition 2.4. Suppose G C \ {0} is a domain and f : G C a holomorphic
function so that exp(f(z)) = z for every z G. Then f is said to be a branch of the
logarithm on G.
Branch point and branch cut: Branches of the logarithm are usually de-
noted by log : G C. In practice, however, we need to specify, which partic-
ular branch we are actually considering. For this purpose it is enough, e.g., to
specify the function value in one point in G. There is no branch of the logarithm
that contains the origin, which is called branch point of the logarithm. This
does not mean, however, that a branch of the logarithmexists on every domain
that excludes the origin: in fact, there is no branch of log on G = C \ {0}. One
would need e.g. to cut out a half line starting from the origin, i.e. a branch
cut letting G =

D

as above. These are the only branches that we will use in

this course, but there are may more domains that have a logarithm, including
what is called simply connected subdomains of C \ {0}.
Relation with the arg function and geometric interpretation: Recall that
exp(log z) = z for any branch of log and if z = re
i
= e
log r+i
, where r = |z| and
is the argument of z dened (mod 2i). We see that log z = log r+i with the
multivalued nature of log expressed in terms of the multivalued nature of the
argument function. The graph of z arg z is a subset of C R R
3
which
looks like a helicoid (a bit like a spiral path). To use arg z to dene a function
we need to restrict its domain and take part of the graph over that domain
this is known as taking a branch.
Example 2.12. Let
G = {z C \ {0} : 0 < arg z < 2}.
Take log z = log |z| + i arg z with the restriction 0 < arg z < 2i to the argument of
z. With this branch, log i = i/2, log(1) = i, log(i) = i3/2.
Example 2.13. If, when dening the function log, we restrict arg z to lie in (, )
then log i = i/2, log(1) is not dened and log(i) = i/2.
Example 2.14. We take a branch of the logarithm and calculate:
d
dz

e
log z

=
d
dz
(z)
d
dz
(log z)e
log z
= 1
d
dz
(log z) =
1
z
20
Complex Powers
We use log to dene complex powers by z
w
= e
(log z)w
or more precisely:
Denition 2.5. Suppose G C\ {0} is a domain on which a branch of the logarithm
is dened. Then the function z e
(log z)w
is said to be a branch of the power
function z
w
on G.
Therefore complex powers are multifunction unless the exponent w is an
integer. Indeed, if w = n Z then
z
w
= e
(log |z|+i arg z+i2m)n
= e
(log |z|+i arg z)n
.e
2mni
= e
(log |z|+i arg z)n
.
So z
n
is a function not a multifunction. But consider
z
1
2
= e
(log |z|+i arg z)
1
2
= e
(log |z|)
1
2
.e
1
2
i(arg z)
Now
e
(log |z|)
1
2
= e
log |z|
1
2
= |z|
1
2
as in real analysis. However,
e
1
2
i(arg z+2n)
= e
1
2
i arg z
.e
ni
and e
ni
= 1 if n is an even integer, e
ni
= 1 if n is an odd integer. So the
multifunction z
1
2
takes two values for each z.
If we dene f
1
(z) = z
1
2
by restricting arg z to lie in (0, 2) then as arg z
varies from 0 to 2, e
1
2
i arg z
varies from 1 to 1. If we now dene f
2
(z) = z
1
2
by
restricting arg z to lie in (2, 4) then as arg z varies from 2 to 4, e
1
2
i arg z
varies
from 1 to 1. Thus we have two holomorphic branches on D and f
1
= f
2
.
Notice that neither is continuous across the positive real axis but there is a
jump of a factor of 1.
21