Duke Kunshan University

The Mandelbrot set and the Julia sets

MATH 303 Project Report

Chenyu Wang, Edward Yue, Lunji Zhu, Jinchen Zhao

March. 11th. 2021

 Abstract
This article explored some basic properties of the Mandelbrot Set and Julia sets from
the prospective of iterative function schemes and complex dynamics, with some discussions
on the relationship between iterative functions and fractals. We first gave a brief history of
fractals and dynamical systems, and introduced some basic concepts about iterative function
schemes. Then, we focused on the properties of the Julia Sets and the Mandelbrot Set, from
the perspective of dynamical system and topology. Our investigations may shed light on the
link between the geometrical properties of fractal shapes and the properties of dynamical
systems. It also gave us a brief view of the basics of discrete dynamical systems, which we
did not discuss much in the lectures. At the very end of the report, we discussed possible
ways we can go if we want to know more about the topic, such as the Measure Theory.

1 Introduction
Discrete dynamical system is a vital branch in the field of dynamical systems as a whole. Not
only its resemblance and differences with continuous dynamical systems are worth investigating,
but also they are closely related to one of the most beautiful topics in math, fractals. This report
is based on our investigation on the properties of the Julia Sets and the Mandelbrot Set mainly
from the prospective of dynamical systems. We would introduce the historical background of
the link between dynamical systems and fractals, and provides some general knowledge of dy-
namical systems. Then, after gaining a broad view, we focus specifically on the discrete complex
dynamics. Topological and dynamical discussion are given on the properties of Mandelbrot Set
and Julia Sets in the case of f (z) = z 2 + c, which is the main part of our investigation. Finally,
we would discuss possible paths for future investigations. Through understanding the concepts
and learning about the theorems and their proofs, we have not only gained a better perception
of discrete dynamical system, but also broadened our horizons by taking a glimpse of the links
between dynamical systems and topology.

2 Literature Review
2.1 History of Fractals, the Julia Sets and the Mandelbrot Set
According to Falconer (2003), just as the biological definition of “life” for biologists, mathemati-
cians still have not reached a total consensus on the definition of Fractals, since every attempt for
rigorous definition would unavoidably exclude some cases that are actually widely acknowledged
to be archetypal “Fractals”.[2] Nonetheless, the concept and discussion on Fractals arise from a
very early age. According to Pickover (2009), the history of discussion on Fractals can be traced
back to 17th century.[7] The Philosopher and Mathematician Gottfried Wilhelm (von) Leibniz
have primarily discussed the self-similarity of iteration. However, it was not until the year of
1872 that the famous mathematician Karl Weierstrass has proposed the first definition of what
we are now calling Fractals (Trochet, 2009).[8] In 1918, a young French mathematician Gaston
Julia, who devised the formula of the Julia sets, wrote his famous work A Note on the Iteration
of Rational Functions. His work creatively discovered the link between self-similarity and com-
plex function and iterative functions (Julia, 1918).[4] In 1918, the German mathematician Felix
Hausdorff wrote his work Dimension and outer measure, where he proposed the so-called Haus-
dorff Dimension which can define the dimension of any subset of a measure space, like fractals
(Hausdorff, 1918).[3] His work laid a vital foundation for Mandelbrot’s proposal of the definition
of fractals. The terminology of fractal, which comes from the Latin word fractus (means “frag-
mented” and “broken”), is settled by Benoit B. Mandelbrot, a mathematician who is mostly

1
famous for his accomplishments and discovery in Fractal Dynamical Systems.[1] He starts to
work on self-similarity since 1960s. In the year 1967, his work How Long Is the Coast of Britain?
Statistical Self-Similarity and Fractional Dimension was published on Nature. This article used
the length of the coastline of Britain as a introductory case to discuss the self-similar curve with
a dimension between 1 and 2 (Mandelbrot, 1967).[6] In the year of 1975, Mandelbrot publicized
his famous work Fractals: Form, Chance and Dimension. In this book, he proposed the termi-
nology f ractals to indicate an object whose Hausdorff Dimension is larger than its topological
dimension. Thanks to the development of computer, now we can easily iterate and visualize the
complex function z 7→ f (z; c) = z 2 + c to gain an intuitive sense of its properties and beauty.

2.2 Fractals and Iteration Functions

Our report will be centered on the properties of the Julia Sets and the Mandelbrot Set, which
are derived from the properties of the iteration of the function f (z) = z 2 + c and are them-
selves f ractals. To extend our discussion to a more general scope, it is necessary to give a brief
introduction of the link between fractals and iterative functions before we dig into the specific
discussion of the Mandelbrot set and the Julia sets.

2.2.1 Iterative Functions

In our MATH 303 course, we are basically talking about dynamical systems. However, we are
mostly focusing on continuous dynamical systems, which are characterized by Differential Equa-
tions. Meanwhile, an iterative function system is also a kind of discrete dynamical systems, which
shares some basic properties with the continuous case however with some difference. Iterative
functions can serve as the construction formula for fractal shapes.
First we define a special kind of functions, which are vital for the construction of many kinds of
fractals.

Definition 2.2.1 (Contraction Mapping): f : D 7→ D is a contraction on D if |f (x)−f (y)| ≤

c|x − y|, where c ∈ (0, 1).

The intuitive interpretation of contraction mapping is that: if we put two different points into this
mapping, after times and times of iteration, the points would inevitably get closer and closer to
each other. Then, we give the definition of Iterative Function Scheme based on contraction maps.

Definition 2.2.2 (Iterative Function Scheme): An iterative function scheme is a finite

set of contraction mappings: {S1 , S2 , S3 ...Sn }.

Though we are restricting the iterative function scheme to contraction mappings, broadly speak-
ing they are not necessarily be contractions. The reason why we wish to use this constraint is
that the scheme of contraction are more closely related to our discussion of invariant sets and
attractors, which we are to introduce instantly.

Definition 2.2.3 (Invariant Set): An invariant

S∞ set with respect to an iterative function scheme
is a compact subset of D, that suffices F = i=1 Si (F ).

Basically, invariant set refers to the set of points who are still in the set after iterated by any
iterative function in the iterative function scheme. The example of Cantor Set may help one to

2
understand the concept.

Example 2.2.4 (Cantor Set): The Cantor Set C is the invariant set of the iterative func-
tion scheme {S1 , S2 }, where S1 (x) = x1 , and S2 (x) = x3 + 23 . One way to verify this fact is that
if we try to plug in any point in Cantor Set to any iterative function in the iterative function
scheme, we would still get a point that is in the Cantor Set.

Now we state the theorem of Existence and Uniqueness of Invariant Set without proof, since
that’s not our focus in the project.

Theorem 2.2.5 (Existence and Uniqueness of Invariant Set): Consider the iterative
function scheme, given by a family of contractions, that is

|Si (x) − Si (y)| ≤ ci |x − y| (x, y ∈ D) (1)

where for every i, ci ∈ (0, 1). Then there exist a unique non-empty compact set F suffices
∞
[
F = Si (F ) (2)
i=1

Furthermore, if we define a function S on the family of non-empty compact subsets of D, such

that for all non-empty subsets of D (Denoted by E) satisfies
∞
[
S(E) = Si (E) (3)
i=1

and denotes S k to be the kth iteration of S, then for every E and i, Si (E) ⊂ E, and
∞
\
F = S k (E) (4)
k=0

The significance of this theorem is that it constructs a correspondence between the iterative
function scheme and its invariant sets. More specifically, it triggers the questions that how can
we find the invariant sets of a given iterative function scheme and that what can we say about
the corresponding iterative function scheme given the properties of the invariant set. A useful
relation to explore the correspondence is
m
[ m
[
d( Si (A), Si (B)) ≤ ( max ci )d(A, B) (5)
i∈[1,m]
i=1 i=1

Basically, this relation states that due to the “invariance” of the invariant set and the nature
of compact sets, the points are basically getting closer and closer to the points in the invariant
sets. This is basically true in the real case. An important implication is that if we first put the
whole domain to the iterative function scheme, as the iteration goes on, the output set would
be a better and better approximation of the invariant set, and that’s exactly the way computer
get the simulation of invariant set of an iterative function scheme. Sometimes, the points in the
invariant sets may have different properties.

Definition 2.2.6 (Attractor): An attractor is a set of invariant points, in whose neighborhood

(called Basin of Attraction) the orbits converges to the points.

3
Remark: The elements of the attractors may not necessarily be a point, it can be a set of
points that is an periodical orbit. In this case, we may understand the “distance” of one point
to the set as the distance between it to the closest point of the set.

Definition 2.2.7 (Repellor): A repellor is a set of points whose nearby points iterates away.

Remark: Roughly speaking, the attractor of f is the repellor of f −1 . One may notice that in
this case, the function f −1 cannot be an contraction given f is a contraction. And for an iterative
k k−1
T∞ kthe invariant set is often an attractor, since f (D) ⊂ f
function scheme, (D) ⊂ · · · f (D) ⊂ D
and F = k=1 f (D).

2.2.2 How does iterative function scheme relates to Fractals ?

First, iterative functions can “generate” fractal shapes. In this case, the iterative functions are
self-similar, that is |Si (x) − Si (y)| = ci |x − y| where ci ∈ (0, 1). Noticeably, the invariant sets
are self-similar, that is, a fractal shape.
Second, in our discussion, the Julia Sets and the Mandelbrot Set are sets that “dipicts” some
properties of an complex iterative function f (z) = z 2 + c. The shapes of the Julia Sets and
the Mandelbrot set are fractals too, which somehow originated from the iterative nature of the
functions that they relates to.

3 Motivation
Fractal patterns often occur in dynamic systems. An iterative function graph {f k } is said to be
a discrete dynamic system. The Julia set J(f ), the set of all points z in C such that f displays
sensitive dependence at z, can be derived from the iteration of a complex function. An amazing
aspect in the theory of the Julia sets is that we do not have to construct a complicate function
f to find examples where J(f ) is an interesting and complicated set. Instead, by simply letting
fc (z) = z 2 + c for some complex constant c, the corresponding Julia set J(f ) might construct
amazing fractal patterns on the complex plane.

4 Research Questions
1. What are the definitions of the Julia sets and the Mandelbrot set?
2. What are the geometric properties of the Julia sets and the Mandelbrot set?
3. What is the structure of the Julia sets of quadratic functions?
4. What is the relationship between the structure of the Julia sets of quadratic functions and the
Mandelbrot set?

4
5 The Julia Sets and Fractal Geometry
5.1 Definition
Definition 5.1.1 The filled in Julia set of the function f is defined as

K(f ) = z ∈ C : f k (z) 9 ∞

(6)

Definition 5.1.2 The Julia set of the function f is defined to be the boundary of K(f ) i.e.

(J) = ∂K(f ) (7)

Definition 5.1.3 The Fatou set of f is defined to be the compliment of K(f ) i.e.

F (f ) = C\K(f ) (8)

Definition 5.1.4 We call a subset F of D an attractor for f if F is a closed set that is invariant
under f (i.e. with f (F ) = F ) such that the distance from f k (x) to F converges to zero as k
tends to infinity for all x in an open set V containing F. The largest such open set V is called
the basin of attraction of F .

Definition 5.1.5 A set B is a dense subset of A if B ⊂ A ⊂ B̄, i.e. if there are points of B
arbitrarily close to each point of A.

Definition 5.1.6 If f (w) = w we call w a fixed point of f, and if f p (w) = w for some integer
p > 1 we call w a periodic point of f ; the least such p is called the period of w. We call
w, f (w), . . . , f p (w) a period p orbit.

5.2 Fundamental Properties

In this section we will prove some general results relating to the structure of the Julia sets. We
will first show that they are compact subsets of C.

Theorem 5.2.1. The Julia set J of a function fc (z) = z 2 + c is compact for all c ∈ C.

Proof. We first show that J is bounded. Choose r = max(|c|, 3) and let |z| ≥ r. Now we have:
 2
z  = |fc (z) − c| ≤ |fc (z)| + |c| (9)

and so we have that:

|fc (z)| ≥ z 2  − |c| ≥ 3|z| − |z| = 2|z|
 
(10)
Now we can deduce that if |z| ≥ r then

fcn (z) ≥ 2n |z| → ∞ (11)

and so J (fc ) ⊆ B(0, r) Now it remains to show that J is closed. Let z ∈ C\K (fc ) . Hence, there
exists an m ∈ N such that fcm (z) ≥ r and using the continuity of fcm we have that there exists
a ball B(z, δ) for some small δ such that fcm (z) ≥ r for all z ∈ B(z, δ). Hence, C\K (fc ) is open
and hence K (fc ) is closed. Since, J (fc ) = ∂K (fc ) we have that J (fc ) is closed. And hence, J
is compact. 

5
The fact that the Julia sets are closed and bounded (compact) is a useful property. For example,
in determining whether or not a particular point z is in K (fc ) we may conclude that z ∈/ K (fc )
if fcm (z) ∈
/ B(0, r) for some m. This is particularly useful when using a computer to plot the
Julia sets. We will now show that J (fc ) is symmetric about the origin.

Theorem 5.2.2. z ∈ J (fc ) ⇐⇒ −z ∈ J (fc ).

Proof. First note that we have fc (z) = z 2 + c = (−z)2 + c = fc (−z). Hence, fck (z) =
fck−1 (fc (z)) = z 2 + c = (−z)2 + c = fck−1 (fc (−z)) = fck (−z). And we have that fck (z) →
∞ ⇐⇒ fck (−z) → ∞. Hence, we can deduce that K (fc ) and hence J (fc ) is symmetric about
the origin. 

It can easily be shown that the Julia sets are invariant under f and f −1 (Falconer, 2003: 218).[2]
This is a useful property and demonstrates the fact that the Julia sets are repellers, in that they
themselves are invariant under f but points just outside J will be mapped away from J either
to ∞, if in the Fatou set, or towards some fixed point of fc (z) in the filled in Julia set.

Theorem 5.2.3. For every integer p, J(f p ) = J(f ).

Proof. Apply the same strategy as we did in Theorem 5.2.1, we could assert that f k (z) → ∞ if
and only if (f p )k (z) = f pk (z) → ∞. Therefore, f and f p shares the same filled in Julia set and
hence the same Julia set. 

The proofs for the following theorems inevitably involves much knowledge about complex analy-
sis. Therefore, we skipped some of them and focused more on the implication of these theorems.

Theorem 5.2.4. Except for at most one exception, the following conclusions holds for all c ∈ C:
(a) If U is any open set that intersects with J(f ), then there are infinite number of k such that
f −k (z) intersects with U .
S∞
(b) If z ∈ J(f ), then J(f ) is the closure of k=1 f −k (z).

This theorem is the foundation of the methods for computers to draw the Julia sets.

Theorem 5.2.5. Suppose w is the attractor of f , then ∂A(w) = J(f ). The statement also holds
for w = ∞.

Proof. Suppose z ∈ J(f ). Then for every integer k, f k (z) ∈ J(f ), hence it cannot converge
to the attractor of f and we have z ∈ / A(w). However, according to Theorem 5.2.4, for any
neighborhood U of z, there exists some k such that f k (U ) A(w) 6= ∅, i.e., there exists a point
T
x arbitrarily closed to z such that f k (x) → w. Therefore, z ∈ A(W ) and z ∈ ∂A(w). The proof
of inverse part has to be omitted here since it requires some tools from complex analysis. 

As a typical example of this theorem, the Julia set of f (z) = z 2 is a unit circle, which is the
boundary of A(0) and A(w). See Figure 1.
In the following part we will discuss chaos in dynamic systems. Chaos describes a phenomenon
of some dynamic systems whose apparently random states of disorder and irregularities are
actually governed by underlying patterns and deterministic laws that are highly sensitive to initial

6
 Figure 1: Julia Set of f (z) = z 2

conditions. Chaos Theory was originally put forward by American meteorologist Edward Lorenz
(Lorenz, 1963).[5] In one of his simulation of weather by computer, he accidentally discovered that
a tiny change of initial condition for weather system would saliently affect the computation result.
Whereafter, he reduced his original model to a three-variable first-order differential equations,
which is defined as follows:
dx
= σ(y − x) (12)
dt
dy
= x(ρ − z) − y (13)
dt
dz
= xy − βz (14)
dt
As Figure 2 implies, the solutions of this system shows sensitive dependence on initial conditions.
Similar to the description of chaos in continuous dynamic systems, a definition for a mathematical
description of sensitive dependence on initial conditions has also been developed for discrete
dynamic systems.

Definition 5.2.6 Suuppose F is the attractor or repeller of f . f is said to be chaos if:

(1) There exists x ∈ F , such that the orbit f k (x) is dense in F .

(2) The periodical points of f in dense in F .
(3) For every x ∈ F , there exists δ > 0 such that for arbitrarily closed points x, y and any
integer k, we have |f k (x) − f k (y)| ≤ δ.

7
 (a) x(0) = 0, y(0) = 1, z(0) = 1 (b) x(0) = 0.0001, y(0) = 1, z(0) = 1

Figure 2: Two Solutions with Tiny Difference of Initial Conditions

Based on previous properties of the Julia sets, we could eventually come up with the chaos
property of the Julia sets.

Theorem 5.2.7 Suppose f is a complex polynomial function. f is chaos on the Julia set J(f ).

Proof. Omitted. 

6 The Mandelbrot Set

A given complex constant c will determine the structure of J (fc ) It turns out that the location
of c in the Mandelbrot set tells us a lot about this structure. Named after Benoit Mandelbrot
(1924 − 2010), the Mandelbrot set is an extremely complicated set. It is defined as follows:

Definition 6.1. The Mandelbrot set M is defined as:

M = {c ∈ C : J (fc ) is connected} (15)

Given the link to the connectedness of the Julia sets it is hardly surprising that the location
of c in M is linked to the structure of J (fc ) . It is not immediately obvious, however, just how
strongly the two are linked. The definition given above is usually awkward to use directly and
so the following theorem is useful:

Theorem 6.2. c ∈ M ⇐⇒ fcn (0) → ∞.

Proof. Omitted. (See Falconer, 2003: 225-227).[2] 

This theorem is commonly referred to as the Fundamental Theorem of the Mandelbrot set.
From the definition of the Mandelbrot set we see that the Julia set corresponding to a point
outside M will be disconnected and inside M will be connected. The link goes much deeper than
that. The Mandelbrot set is split into obvious chambers (see Figure 3). The large heart shaped
region containing the origin is referred to as the main cardioid with the smaller circular regions

8
commonly referred to as buds. Each of these buds has smaller buds growing out from them and
then at the far reaches of M we see tiny hairs extending out into the complex plane. Julia sets
corresponding to points in each of these distinctive areas have very different structures.

Figure 3: A Plot of the Mandelbrot Set

7 Julia Sets of Quadratic Functions

In this section, we will focus on the structure of Julia sets to the polynomial fc (z) = z 2 + c.
In particular, we will discuss how the structure of the Julia set J(fc ) changes as the parameter
c varies across the complex plane. We will follow the discussion of connection between the
Mandelbrot set and the Julia sets, and the significance of the different parts of the Mandelbrot
set M will start to become obvious.
The attractive periodic points of fc are important to the structure of J(fc ). If w is an attractive
periodic point of a polynomial f then there is a critical point z such that f k (z) is attracted to
the periodic orbit containing w. Since the only critical point of fc is 0, fc can have at most one
attractive periodic orbit. Moreover, if c ∈ / M then fck → ∞, so fc has no attractive periodic
orbit.
We can classify fc by the period p of its attractive orbit. The values of c corresponding to
different p can be identified as different regions of the Mandelbrot set M (see Figure 3).
First, consider the case where c lies outside M , so fc has no attractive periodic points. J(fc )
is totally disconnected and can be expressed as the disjoint union J = S1 (J) ∪ S2 (J), where S1
and S2 are the two branches of fc−1 on J This means that J is the attractor of the IFS {S1 , S2 }.
We study this situation in more detail when c is large enough to allow some simplification.
√
Theorem 7.1. Suppose |c| > 41 (5 + 2 6) then J(fc ) is totally disconnected, and is the attractor
of the contractions given by the two branches of fc−1 = ±(z − c)1/2 for z near J.

Proof. Let C be the circle |z| = |c| and D its interior |z| < |c|. Then

fc−1 (C) = {(ceiθ − c)1/2 : 0 ≤ θ ≤ 4π} (16)

9
Figure 4: The periods of the attractive orbits of fc for c in various parts of the Mandelbrot set.

which is a a figure of eight with self-intersection point at 0, with the loops on either side of
a straight line through the origin. Since |c| > 2 we have fc−1 (C) ⊂ D since if |z| > |c| then
|fc (z)| ≥ |z 2 | − |c| ≥ |c|2 − |c| > |c|. The interior of each of the loops of fc−1 (C) is mapped by fc
in a bijective manner onto D. If we define S1 , S2 : D → D as the branches of fc−1 (z) inside each
loop, then S1 (D) and S2 (D) are the interiors of the two loops.
Let V be the disc {z : |z| < |2c|1/2 }. We have chosen the radius of V so that V just contains
fc−1 (C), so S1 (D), S2 (D) ⊂ V ⊂ D. Hence S1 V, S2 V ⊂ V with S1 (V ) and S2 (V ) disjoint. We
have, for i = 1, 2,

|z1 − z2 |
|Si (z1 ) − Si (z2 )| = |(z1 − c)1/2 − (z2 − c)1/2 | = (17)
|(z1 − c)1/2 + (z2 − c)1/2 |

Hence, if z1 , z2 ∈ V , taking least and greatest values,

1 |Si (z1 ) − Si (z2 )| 1

(|c| + |2c|1/2 )−1/2 ≤ ≤ (|c| − |2c|1/2 )−1/2 (18)
2 |z1 − z2 | 2
√
The upper bound is less than 1 if |c| > 14 (5 + 2 6, in which case S1 and S2 are contractions on
the disc V . There is a unique non-empty compact attractor F ⊂ V satisfying

S1 (F ) ∪ S2 (F ) = F. (19)

Since S1 (V ) and S2 (V ) are disjoint, so are S1 (F ) and S2 (F ), implying that F is totally discon-
nected.
Of course, F is none other than the Julia set J = J(fc ). One way to see this is to note that, since

10
 Figure 5: Proof of Theorem 7.1.

S∞
V contains at least one point z of J, we have J = closure( k= fc−k (z) ⊂ V , since fc−k (V ) ⊂ V .
J is a non-empty compact subset of V satisfying J = fc−1 (J) or, equivalently, J = S1 (J) ∪ S2 (J).
Thus J = F , the unique non-empty compact set satisfying (19). 

Next, we turn to the case where c is small. If c = 0 then J(fc ) is the unit circle. It is reasonable
to expect that as c moves away from 0, the circle will be distorted into a simple closed curve. In
fact, this is the case given that fc retains an attractive fixed point, i.e., if |f 0 | < 1 at one of the
roots of fc (z) = z. This happens if c lies inside the main cardioid of the Mandelbrot set given
by z = 12 eiθ (1 − 12 eiθ )(0 ≤ θ ≤ 2π). We consider |c| < 14 for convenience.

1
Theorem 7.2. If |c| < 4 then J(fc ) is a simple closed curve.

Proof. Omitted. (See Falconer, 2003: 230-231).[2] 

Another case we will consider is when fc has a attractive periodic orbit of period 2. This occurs
if |c + 1| < 41 , that is, if z lies inside the circular disc of M next to the cardioid. Since fc2 is a
polynomial of degree 4, fc has two fixed points and two period-2 points. Let w1 and w2 be the
points of the attractive period-2 orbit. The basin of attraction for wi includes a region bounded
by a simple closed curve Ci surrounding wi , for i = 1, 2, and Ci ⊂ J(fc2 ) = J(fc ). The curves Ci
are mapped onto themselves in a two-to-one fashion by fc2 , which implies that there is a fixed
point of fc2 on each Ci . The period-2 points are strictly inside the Ci , so there is a fixed point of
fc on each Ci ; since the Ci are mapped onto each other by fc , the only possibility is for C1 and
C2 to touch at one of the fixed points of fc . The inverse function fc−1 is two-valued on C1 (That
means, one value in the domain may corresponds to two values in the range, which contradicts to
the definition of function actually, however we still use ‘function’ to call it for convenience). One

11
of the inverse images is C2 . However, the other branch of fc−1 (C1 ) is a further simple closed curve
enclosing the second value of fc−1 (w1). We may continue to take inverse images in this way to
find that J(fc ) is made up of infinitely many simple closed curves which enclose the pre-images
of w1 and w2 of all orders and touch each other in pairs at “pinch points”.
The same idea is used to analyze the situation where fc has an attractive periodic orbit of period
p > 2.The region of the period-p points that are drawn into the attractive orbits are bounded by
simple closed curves which touch each other at a common point. The Julia set consists of these
fractal curves together with all their pre-images under fk .
The most complex Julia sets J(fc ) corresponds to c at the boundary of M . If c is on the boundary
of the cardioid or a bud of M then fc has an indifferent periodic point (where (f p )0 (w) = 1). If c
is at a “neck” where a bud touches a parent region then J(fc ) consists of a series of “tendrils”. If
c lies elsewhere on the boundary of the cardioid, the Julia set may include “Siegel discs”, which
is infinitely many curves bounding open regions. Moreover, if c is on the “hairs” of M and the
critical point 0 is periodic under iteration by fc then J(fc ) is a “dendrite”.
We use Figure 6 to summarize our discussion in this section.

Figure 6: Julia sets J(fc ) for c at various points in the Mandelbrot set.

8 Conclusion and Discussion

In a iterative function system view, we study the geometric properties of the Julia sets and
the Mandelbrot set. In particular, we demonstrate the change of structure of the Julia sets
of quadratic functions as the parameter c varies, and it brings out the connection between the
Mandelbrot set and the Julia sets. It is surprising to see that the iteration of a complex function
with rather simple formula can lead to a very complex chaotic behaviour with fractal essence.
Due to our limited knowledge, we are not able to examine some very important topics. For
instance, the fractal dimension is an important property which in a sense yields the size of
fractals. But the discussion of fractal dimension requires a deep understanding of measure theory
which we currently do not obtain. So for further study, we may focus on the theory of complex
analysis and measure which are fundamentals for learning fractal geometry.

12
References
[1] The Editors of Encyclopaedia Britannica. Fractal. Encyclopedia Britannica. 2017.
[2] Kenneth Falconer. Fractal geometry: mathematical foundations and applications second edi-
tion. John Wiley & Sons, 2003.
[3] Felix Hausdorff. “Dimension and outer measures”. In: Mathematische Annalen 156.3775
(1918), pp. 157–179.
[4] Gaston Julia. “Mémoire sur la permutabilité des fractions rationnelles”. In: Journal de
Mathématiques Pures et Appliquées. Vol. 1. 1918, pp. 47–246.
[5] Edward Lorenz. “Deterministic non-periodic flow”. In: Journal of atmospheric sciences 20.2
(1963), pp. 130–141.
[6] Benoit Mandelbrot. “How long is the coast of Britain? Statistical self-similarity and frac-
tional dimension”. In: science 156.3775 (1967), pp. 636–638.
[7] Clifford A Pickover. The math book: from Pythagoras to the 57th dimension, 250 milestones
in the history of mathematics. Sterling Publishing Company, Inc., 2009.
[8] Holly Trochet. A history of fractal geometry. MacTutor. 2009.

13