Complex Analysis
Table of Contents
Definitions
A multivalued function is a function which assigns to each point a set of values rather than a single value to each point, as a regular function would.
We define the open ball (or open disk since the complex plane is 2-dimensional) in the same way we do for the reals:

and the closed disk

Open and closed sets in the complex plane then follow the same definition as for open and closed sets in the 2-dimension real case.

Functions
The *complex exponential function is the function defined by


and

Let be non-zero. We define the multivalued function
by

We call any element of
a logarithm of
.
The principal branch of the logarithm function is the function defined by

for non-zero , where
is the principal value of the argument function.
Or more general
The branch cut of a complex function is defined to be the point in the principal branch where the function is no longer holomorphic.
Theorems
Holomorphic functions
Notation
is almost always denoted
, i.e. the unique
in the range
is the complex plane without origin => can be equipped with multiplication operation to form an (Abelian) group
- complex-valued function of a real variable means
, i.e. input is real and we can decompose the output into a real part and an imaginary part
Differential Geometric view on arg
function
Let
defined by
, which is then surjective and continuous
Apparently, we call a complex manifold such as the one above a Riemann surface, and I honestly believe this is something I ought to return to once I have a better understanding of what complex manifolds are.
Observe then that this defines a fibre bundle, since for any two points we have the fibres be isomorphic, i.e.
.
is the plane defined by Cartesian product
since
and
. Thus,
is in fact a topological manifold (probably equipped with the standard topology, but I haven't checked). Further we observe that
is also a topological manifold. Thus, we our fibre bundle as claimed earlier.
Complex-valued functions
Suppose we have a complex-valued function of a real value

We then say that is continuous at
if and only if both
and
are continuous at
.
If is instead a complex-valued function of a complex variable, let
, then

We then say that is continuous at
if and only if for any
there exists
s.t.

where and
are real valued functions, and this is thus equivalent of
and
being continuous at the point
.
Complex differentiability and holomorphicity
A function in a neighborhood
of
is differentiable (everywhere in
) if and only if its real and imaginary parts
and
are continuously differentiable and obey the Cauchy-Riemann equations (everywhere in
):

This can be seen by considering the "normal" definition of differentiability at a point, and observing the definition changes depending on which direction we approach the point from (vertical or horizontal direction).
Important: it's crucial that the function be differentiable in a neighborhood of
not just at the point
!
We say that the complex-valued function is holomorphic at
if it is differentiable everywhere in some neighbourhood
of
.
We say that a function is entire if it is holomorphic in the whole complex plane.
Let .
Then is harmonic if it satisfies the Laplace equation:

Let be open, and let
be harmonic.
We say a harmonic function is the harmonic conjugate of
if the complex-valued function

is holomorpic on .
Solving differential equations
This techinque can be used to solve differential equations on "ugly" domains.
Consider the following:
Polynomials and rational functions
Complex powers
Let with
. Then we define
th power of
by

Unless stated otherwise, a complex power is defined in the principal branch.
Let with
. Then
- if
there is exactly one value of
- if
where
are coprime, with
there are exactly
values of
- if
or
, there are infinitely manu values of
Let . Then the
values

where

are the roots of unity.
Graphing complex functions
- Hard to visualize as 2D + 2D becomes 4D
- Good idea to treat each of the different 3D plots separately
Let be open, and let
.
We say is conformal if
preserves angles: i.e. if the angle between the images under
of two straight lines in
are equal to the angle between the two straight lines themselves.
By applying this definition to tangents of differentiable curves, more generally we can say the same about the angles between curves at certain points.
We're saying the push-forward of , is angle-preserving.
Therefore, any diffeomorphism between and
is angle-preserving.
Consider . For
,
. Then we consider the surface

and the function
defined by
.
- This is the projection of a bundle; further,
forms a fibre bundle (right?)
Let be open, and let
be holomorphic.
Then preserves angles at every
where
.
This theorem is useful for establishing the image of a function (if it's a polytope (?)) since we can simply compute the mapped values at the edges (i.e. curves from vertex to vertex) and, knowing that the angles are preserved, immediately know how the edges between the mapped vertices look.
Möbius transformations
A Möbius transformation is a function of the form

where are such that
.
Observe that is not defined on the entirety of
, which leads us to defining the extended complex plane.
We'll often consider the case , if nothing else is specified.
That is, a Mobius transformation is defined on .
The extended complex plane is the set , where
is just some object
.
We extend the usual arithmetic operations in the following way: for and non-zero
,

Consider the coordinates describing
. We identify the complex plane with the plane defined by
, and a complex number
with the point
.
The Riemann sphere is the unit sphere in
defined by

and we consider the "north pole" to be the point .
The Riemann sphere therefore has two charts:
- For all points in the complex plane, the chart is the identity map from the sphere (with
removed) to the complex plane.
- For
, the chart neighborhood is the sphere (with the origin removed), and the chart is given by sending
to
and all other points
to
.
Let such that the three points
,
, and
are colinear.
It is clear that

thus we define also and thereby consider
as being defined on the extended complex plane, i.e.
.
The map is evidently bijective, so it has an inverse
. This function
is the sterographic projection.
Better description:
The unit sphere in is the set of pints
such that
.
Let be the "north pole", and let
be the rest of the sphere.
The plane (xy-plane) runs through the center of the sphere; the "equator" is the intersection of the sphere with this plane.
For any point on
, there is a unique line through
and
, and this line intersects the plane
in exactly one point
. We define the stereographic projection of
to be this point
in the plane.

and

Stereographic projection maps a circle to either a circle or a straight line (a "circline").
Makes some people say that straight lines are circles of infinite radius.
A translation is a Möbius transformation of the form
which corresponds to the matrix
A rotation is a Möbius transformation of the form
so that
for some
, which corresponds to the matrix
A dilation is a Möbius transformation of the form
which corresponds to the matrix
An inversion is a Möbius transformation of the form
which corresponds to the matrix
We say that a Möbius transformation fixes the point of infinity if
.
Translations, rotatons, and dilations fix the point at infinity, while inversions do not.
Let be a Möbius transformation. Then
is a composition of a finite number of translations, rotations, diluations, and, if and only if
does not fix the point at infinity, one inversion.
Möbius transformations map circlines to circlines.
Let be three distinct points. Then there exists a unqiue Möbius transformation
such that

This is useful because we by simply knowing how the Möbius transformation maps the three different points, we can tell what it does to circles or lines.
Let be distinct points.
The cross-ratio of the four points is the image of
under the Möbius transformation which sends
to
.
Complex integration
Complex integrals
Let be a interval, and
of the form

Then is integrable if its real and imaginary parts
are integrable in the usual (real) sense, and we define the integral of
by

It will usually suffice to observe that continuous functions are integrable.
Example

where denotes the arc-length of
.
is by def. closed and bounded, i.e. a compact set
is continuous
Hence is in fact bounded i.e. finite.
Contour integrals
Let be distinct.
Then a (parametrized) curve connecting
and
is a continuous function
![\begin{equation*}
\begin{split}
\gamma : [t_0, t_1] \to \mathbb{C}, & \quad t_0, t_1 \in \mathbb{R}, \quad t_0 < t_1 \\
\text{such that} & \quad \gamma(t_0) = z_0, \quad \gamma(t_1) = z_1
\end{split}
\end{equation*}](../../assets/latex/complex_analysis_8369f820d71cf1365de5edc841af4833e39672d1.png)
Writing and
, we decompose
into a real and imaginary parts for continuous real functions
, so
![\begin{equation*}
\gamma(t) = x(t) + i y(t), \quad \forall t \in [t_0, t_1]
\end{equation*}](../../assets/latex/complex_analysis_8f5f60e424b927d41f41a4abf91d949195644368.png)
We say the curve is regular if
is continuously differentiable and
forall
.
A curve from
to
in
is a contour if it is a finite union of regualr curves, which together joint
with
, i.e. there exists
![\begin{equation*}
\gamma_i : [t_0^i, t_1^i] \to \mathbb{C}, \quad i = 1, \dots, n
\end{equation*}](../../assets/latex/complex_analysis_a3cc1eabc7d380b9116d61b465045deddae09378.png)
such that
![\begin{equation*}
\gamma(t) =
\begin{cases}
\gamma_1(t) & \quad \text{if } t \in [t_0, t_1] \\
\gamma_2(t) & \quad \text{if } t \in [t_1, t_2] \\
& \vdots \\
\gamma_n(t) & \quad \text{if } t \in [t_{n - 1}, t_n]
\end{cases}
\end{equation*}](../../assets/latex/complex_analysis_e9b33701e751d9a052a3d068700b49cf8ee72cfd.png)
such that is continuous.
For a continuous function , we define the contour integral of
along
by

Let be a curve in
, and let
be continuous. Then


Let . We will say that
is a domain if
is open and every two points in
can be connected by a contour which lies wholly on
.
Let be a domain, and
be continous. Then the following are equivalent:
has an antiderviative
on
for all closed contours
in
- all contour integrals
are independent of path
, and depend only on the endpoints.
Cauchy's Integral Theorem
Let be a contour in
.
Then is simple if it has no self-intersections, except possible at the endpoints, i.e.
for all distinct
, unless
and
and
is a closed contour.
A loop is a simple, closed contour.
Let be a loop in
. Then
defines two regions in the
, with
as their common boundary:
- a boundary domain, the interior of
- an unbounded domain, the exterior of
Let be a loop in
. We say
is positively-oriented if as we move along the curve in the direction of parametrization, the interior is on the LHS.
Otherwise state, all loops should be assumed to be positively-oriented.
is said to be simply connected if the interior of every loop on
lies wholly in
.
Let and
be a loop in
which does not pass through
. Then

Since , we clearly have two cases:
: integral is zero due to Cauchy Integral theorem
: Consider the following figure:
Since
this figure basically "encapsulates" all possible loops which have
outside!
We break the integrand , aswell as the differential
into their real and imaginary components:

In this case we have

By Green's Theorem, we may then replace the integrals around the closed contour with an area integral throughout the domain
that is enclosed by
as follows:

and for imaginary part,

However, being the real and imaginary parts of a function holomorphic in the domain ,
and
must satisfy the Cauchy-Riemann equations there:

We therefore find that both integrads (and hence their integrals) are zero:

Which gives us the result

as wanted.
Let be a loop,
, and
be holomorphic inside and on
. Then

Let be a loop,
be holomorphic inside and on
, and
lie inside
.
Then is infinitively differentiable at
and, for all positive integers
,

Liouville's Theorem and its applications
Let be a domain,
and
be such that
,
be holomorphic on
, and
be such that
for all
.
Then for all , we have that

Maximum Modulus Principle
and
be such that the closed disc
, and
be holomorphic on
.
Then

Let:
be holomorphic on a domain
s.t.
Fix , then

where is the circle of radius
centred at
(Cauchy Integral formula).
is parametrized by
, given by

So, using the definition of the contour integral

Suppose

i.e. maximum value is attained at the center of the circle. Then,

hence

which, since is clearly constant, we can use the linearity of the integral to get

which is a non-negative continuous function, which implies that the expression inside the integral is zero for all , i.e.
![\begin{equation*}
|f(z_0) | - \Big| f \big( z_0 + r \exp(it) \big) \Big| = 0, \quad \forall t \in [0, 2 \pi]
\end{equation*}](../../assets/latex/complex_analysis_276a7754dcd149492473dfadbc8373a2f3ff4b01.png)
which is true for every point in is of this form, for some
, and some
.
Hence is constant on
.
Let , and
be holomorphic and bounded on
, i.e.

for some .
If achives its maximum at
, then
is constant on
.
Series Expansions for Holomorphic Functions
Stuff
Let be a power series.
Then s.t.
converges if
converges uniformly on
for all
diverges if
is holomorphic on its disc of convergence
, where
is the radius of convergence.
To proof holomorphicity of convergent power series, one could do as follows:
Fix . Then
for some
. Series converges uniformly on
, thus partial sums are holomorphic, which implies series is holomorphic at
.
Let be open, and
.
Then we say that is analytic if at everypoint
,
can be expressed as a convergent power series.
Suppose is holomorphic on
. Then the Taylor series

converges to on
, and converges uniformly on
for all
where

for any loop with
.
It's REALLY important to realize that this Taylor series is only on some disk !!!
Therefore, if we know is holomorphic on some open subset
, then if we want to talk about the Taylor series on any point
, still can only say something about this on some disk
!!
The Taylor series at some point is NOT necessarily the same as the Taylor series at some other point
.
Let be the circle of radius
centered at
where
. By CIF we know that for
we have

So

where we've used the fact that

thus

hence we could recognize the fraction above as the convergent series. Further, due to the series being convergent, we can interchange the summation and integration

which is just the Taylor series!
Taylor series of a (if it exists) is unique, i.e. if
is holomorphic on
and

A anulus is


Suppose is holomorphic on the anulus
.
Then

for any loop with
for all
, and the series converges uniformly on
for all
.
We say is a singularity if
is not holomorphic at
.
Suppose has a singularity at
.
If there exists s.t.
is holomorphic on the punctured disk
,
has a Laurent series centred at
, valid on this disk

Then ONE of the following is true:
for all
(
is a removable singularity)
such that
but
for all
(
is a pole of order
)
infinitively many negative
s.t.
(
is an essential singularity)
If you have a removable singularity and then consider the integral along some loop around .
Then the integral is the integral along some loop on which is holomorphic, hence it's zero.
We can therefore redefine to take on the value
at the singularity point
, and we got ourselves a holomorphic function on the non-punctured disk centered at
!
is a zero of holomorphic function
if
.
Further, we say it's a zero of order if

but

Then

Let ,
be a neighbourhood of
,
be holomorphic on
, and such that
for a sequence of distinct points
which converge to
.
Then is identically zero on some disc centered at
.
Let be holomorphic at
be a zero of
of order
.
Then has
- has a pole of order
at
if
- has a pole of order
(if
) at
, if
is a zero of order
of
, OR
is a removable singularity if
Analytic continuation
Suppose is holomorphic on a domain
with
and
.
Then

Power series above might have radius of convergence

is a holomorphic on with
, i.e.

One might then as does on
?
Yes, is an analytic continuation of
to
where

and is well-defined.
Let be domain, and
be holomorphic.
We say that a holomorphic function is an analytic continuation of
if

Let be a domain,
, and
be holomorpic on
and such that

for a sequence of distinct points which converge to
. Then

Let be a domain,
, and
be holomorphic on
and such that

for a sequence of distinct points which converge to
. Then

Let be holomorphic at
, where
is zero of order
. Then
- if
is not zero of
, then
has a pole of order
at
- if
is a zero of order
of
, then
has a pole of order
at
if
, and has a removable singularity at
otherwise.
Cauchy Residue
Theorem
Let ,
be holomorphic on the punctured disc
for some
, with an isolated singularity at
, and
be a loop inside
, with
. Then

where is the coefficient of the
term in the Laurent expansion of
centred at
,

Let and
be holomorphic on the punctured disc
, for some
, with an isolated singularity at
.
Then the residue of at
, is

where is the term in the Laurent series of
centered at
.
Let , and
be holomorphic on the punctured dist
for some
, with removable singularity at
. Then

Let and
be holomorphic on the punctured disc
, for some
, with a pole of order
at
. Then

Let and
,
be holomorphic on
for some
, such that
has a simple zero (
) at
, while
. Then, defining
we have

Let be a loop, and
be holomorphic inside and on
except for finitely many isolated singularities
. Then

Let be a domain.
A function is meromorphic on
if for all
, either
has a pole at
or
is holomorphic at
.
Application: trigonometric integrals
Integrals of the form

for ration function , can often be evaluated by considering a contour integral of appropriate function around the unit circle centered at
.
On with
we have

and

Defining , we therefore have that

If parametrized by
defined by
, then

It basically comes down to rewriting and
with
, which then often provides us with a rational function of which it's substantially esaier to obtain the singularities, thus the residue, hence the integral around
.
Improper Integrals
We define the Cauchy principal value of the integral

as

Let be a rational function, where

ad such that
. Then

where and
are the semicircular contours from
to
in the upper and lower half-plane, respectively.
Let
be a domain
,
be meromorphic on
with a simple pole at
be the circular arc parametrized by
for
for some
.
Then

Let , and let
be a loop with
. Then
