Simulating the Natural Process of Growth through Fractals

and
Measuring It’s Complexity Through Fractal Dimension

Edward Bujak

2006 RET – University of Pennsylvania

Preface

During the 2006 Summer I was an NSF Fellow in a NSF-sponsored Research Experience
for Teachers (RET) program at the University of Pennsylvania Laboratory for Research
on Structure of Matter (LRSM) working with Prof. Ritesh Agarwal, MSE on 'Synthesis
and Characterization of Cadmium Sulfide (CdS) Nanowires.
(www.lrsm.upenn.edu/outreach/#ret). We developed processes to grow CdS nanowires
to control the morphology and other characteristics such as scaling, uniformity,
composition, purity, and distribution. CdS has phenomenal luminescent properties and
emits visible green light and hence is used for Light-Emitting Diodes (LEDs). With
nanowires we desire a nanostructure that is crystalline (aligned structure of similar atoms)
and very thin with an aspect ratio of 1000:1. The best “wire” is a line segment; the
straighter the better. For your curiosity there are many nano-scale shapes that exhibit
desirable properties: nano-rods (like capsules), nano-tubes (single walled and multi-
walled), nano-checkerboards, and many others.

In the lab we also grew other nanostructures where our desired shape was more fibrous or
dendritic like a branching tree. In this case we wanted maximum surface area in a fixed
volume (mathematically this is more important than it appears!).

This paper/presentation is an introductory mathematical study into growth simulation

using mathematical models; particularly fractals with simple branching and iteration rules
using computer software to automate the repetitive process. This will be taught as a few
lessons within a unit.

This should probably be the first lesson using fractals and might be prefaced by some
visually appealing youtube.com, video.google.com, or similar videos which will catch the
students’ attention and interest.

This unit will investigate patterns, symmetry, beauty (order in chaos) with a very simple
iterative process to created structures. Some fractal terms; such as dimension, self-
similarity, iteration (recursion), need to be understood.

2006 Penn RET - Bujak Page 1 of 22

Brief History (taken from http://www.math.umass.edu/~mconnors/fractal/hist.html)

Historically, the revolution was forced by the discovery of mathematical structures that
did not fit the patterns of Euclid and Newton. Many of the fractals go back to classical
mathematics and mathematicians of the past like Georg Cantor, Guisseppi Peano, Helge
Von Koch, Waclaw Sierpinski and many others. However, reputable scientists and
mathematicians called these structures a "gallery of monsters." Jules Henri Poincaré
deemed many of Cantor's creations "pathological." Cantor's Dust, for example,
constructed in 1877, seemed to jump dimensions. It is constructed by chopping up one
dimensional line segments - but in the end it contains only zero dimensional points
without length or width.

Peano's curve is a space filling curve

These monster curves were seen as shapes intended to show deviation from the norm
rather than to typify the normal. Their profound meaning had been somewhat forgotten
until Benoit Mandelbrot took them out in the 1960's, dusted them off and saw these
pathological shapes as an important clue to a new mathematics of natural forms such as
clouds, trees, and mountain ranges. Mandelbrot is often called the father of fractal
geometry. He coined to word fractal from the Latin frangere meaning to break into
irregular fragments and published his book, The Fractal Geometry of Nature.

Introduction

We are mesmerized by the regularity of shapes; the predictability of patterns. There is

math, an equation, behind much of nature. We try to set out to create the pattern in nature.
Patterns can be explained.

Some natural fractal objects are: mountain ranges, snowflakes, clouds, crystals, leaves,
leaf veins, trees, tree branching, blood vessels, lungs, sea shells, and many others.

Fractals can be characterized by just a few properties:

• Scale invariance
• Self-similarity – part looks like the whole (self-organizing)
• Dimension

Fractals belong to a non-Euclidean way of looking at the universe which may have a non-
integer dimension. They are geometric shapes or patterns that help describe the forces of
growth.

In 1975 the French mathematician Benoit Madelbrot defined fractals as objects that do
not lose their detail or their proportions when they are magnified or shrunk.
This property is reminiscent of phi ( Φ , ϕ , the Golden Mean, the Golden Section, or
1.618...) where the same essential proportion is retained every time you cut the line, the

2006 Penn RET - Bujak Page 2 of 22

rectangle, or other shape. In fact, the qualities of both fractals and phi are concerned with
growth.

Features of Fractals

There are two different types of fractal, the geometric fractal and the random fractal.
The snowflake is an example of a geometric fractal that grows (in the simplest terms) by
the addition of equilateral triangles in specific patterns. Random fractals are computer
generated, in both modeling and games.

Fractal geometry can lead to convincing images of natural growth phenomena, such as
coastlines, ferns, and tree bark. They can also emerge from climate and manmade
phenomena, such as stock price graphs or economic predictions, which show self-
similarity.

Some ferns are classic natural examples of a fractal, with each section (pinna) of leaf
being miniature replication of the whole leaf. A single pinna if magnified looks like a
whole leaf. In addition, in some species, their buds unfold in the shape of a logarithmic
spiral (Fibonacci number – Golden Mean). This means that nature does not have to
redesign the leaf at every stage of its growth, but the initial design keeps on replicating.
This is efficiency.

The other important feature of fractals is scaling. In a fractal the degree of irregularity or
fragmentation is identical on all scales. Fractals do not get smoother as the magnification
brings you closer; they simply continue to generate new irregularities that are
commensurate with the speed at which you zoom in on them.

Order not chaos

Fractals are associated with the mathematics of chaos, but they are in fact, very ordered –
just millions of interlocking self-replicating, interlocking objects. They only look
chaotic yet are governed by a definite geometry.

Because fractals contain repeating elements, they can be created using mathematics.

Usually simple growth specification is repeated iteratively by a generator (rule) with an

initiator and is called an IFS – Iterated Function System. You can do IFS with
Geometer’s Sketchpad® with the Transform then Iterate menu choices.

IF time permits or students want to investigate (or differentiate) them have some play the
chaos game.

2006 Penn RET - Bujak Page 3 of 22

Creating Fractals by hand

Cauliflower Fractal – separate handout, need ruler, compass, paper, and pencil.

Dimension - background

Fractals are called fractals because they may a non-integer dimension. We know 1-D,
2-D, 3-D, and maybe even 4-D, but there are objects below dimension one, in between 1
and 3, and beyond.

Mention methods to measure dimension such as simple box counting dimension,

similarity dimension, or mass dimension (these are very labor intensive, but time well
spent especially if cooperative and/or inquiry-based exercises are desired). Derive and
explain the dimension formula and validate that it does work for geometric objects that
you know: a line should have D=1, a planar shape (such as a rectangle or a square) should
have D=2, a prism (such as a cube) should have D=3, etc. Now measure the dimension
of a Koch curve with a simple box counting algorithm by using successively smaller
grids (on transparencies laid over the image).

The following table is very useful in trying to grasp the notion of a fractal (fractional)
dimension. Make sure students understand the “native” dimension for a 0-D object (a
point) is simply counting the points, that the “native” dimension for a 1-D object is a line,
and so forth. Be sure to explain that a 1-D object has no area or volume, and so forth. Be
sure to explain that a 3-D object is measured by volume and that it has an infinite number
of cross-sectional slices that make up this volume, and so forth. Some of these are hard
to explain or conceptualize, like dimensions above 3 which are not outlined here, but note
that the chart can be extended to any integer dimension.

Ways to measure dimension-n compact objects:

Dimension
0 1 2 3
Count finite ∞ ∞ ∞
Measure

Length 0 finite ∞ ∞

Area 0 0 finite ∞

Volume 0 0 0 finite

So why is a line one-dimensional and the plane two-dimensional? Note that both of these
objects are self-similar. We may break a line segment into 4 self-similar intervals, each
with the same length, and each of which can be magnified by a factor of 4 to yield the
original segment. We can also break a line segment into 7 self-similar pieces, each with

2006 Penn RET - Bujak Page 4 of 22

magnification factor 7, or 20 self-similar pieces with magnification factor 20. In general,
we can break a line segment into N self-similar pieces, each with magnification factor N.
Note: It is important to realize that “similar” is in the geometric sense of the word; the
pieces must have the save proportional shape, the pieces need to look alike, but can be
scaled (in the tree branching lab, the left and right branches are scaled differently)

A square is different. We can decompose a square into 4 self-similar sub-squares, and

the magnification factor here is 2. Alternatively, we can break the square into 9 self-
similar pieces with magnification factor 3, or 25 self-similar pieces with magnification
factor 5. Clearly, the square may be broken into N2 self-similar copies of itself, each of
which must be magnified by a factor of N to yield the original figure. See figure below.
Finally, we can decompose a cube into N3 self-similar pieces, each of which has
magnification factor N.

A square may be broken into N2 self-similar pieces,

each with magnification factor N

Now we see an alternative way to specify the dimension of a self-similar object: The
dimension is simply the exponent of the number of self-similar pieces with magnification
factor N into which the figure may be broken.

Dimension – mathematically

Fractal dimension is a measure of how "complicated" a self-similar figure is. In a rough

sense, it measures "how many points" lie in a given set. A line is larger than a point. A
planar is "larger" than a line. A solid is “larger” than a planar shape.

The simplest dimension is the box counting dimension. If time permits, do a “classic”
coastline counting exercise with successively smaller unit squares. Use transparent graph
paper to overlay the map. It can be empirically shown that:
d
⎛1⎞
N (r ) = ⎜ ⎟
⎝r⎠
where:
r = “radius” of side length of unit box used to count along the edge
N(r) = number of boxes with side length r along edge your counted boxes along
d = dimension

2006 Penn RET - Bujak Page 5 of 22

To isolate the dimension, logarithms are necessary:
⎛1⎞
log N ( r ) = d log ⎜ ⎟
⎝r⎠
log N ( r )
d=
⎛1⎞
log ⎜ ⎟
⎝r⎠

This is like “y = mx + b” for a log-log plot, where the independent x = log (1/r) and
dependent y = log N(r). The linear plot’s slope will be the fractal dimension d. To
convince you that this works, apply it to a line (d=1), a concave planar shape (d=2), and a
concave solid (d=3). Is a hypercube really 4 dimensions? Who said the 4th dimension has
to be time?

When shapes are self-similar (the basic pattern iterates), then a similarity dimension can
be used which is a special case of the box counting dimension, but is much easier.

log N ( r )
d=
⎛1⎞
log ⎜ ⎟
⎝r⎠

where:
r = scaling factor (or magnification factor)
N(r) = number of copies (or # of self-similar pieces)

There are other measures of fractal dimensionality; such as the mass dimension, which is
a power law relationship between mass and radius:
M = k rd
log M = d log r + log k
like “y = mx + b”
What does the y-intercept (log k) mean … is it the (linear) density of the material?
Should it always be the Origin (I would hope not)? Can it be non-zero?

This is skipped here, but if time permits, a great hands-on would be (jelly) bean counting,
to calculate dimensionality. Is packing an issue? Assuming the shape could be scaled
would the size of the similar objects affect the dimension? Packing of atoms is extremely
important in materials science the structure of the atoms affect the materials properties.
Examples of standard crystal lattice structures are SC (simple cubic) unit cell, FCC (face-
centered cubic) unit cell, BCC (body-centered cubic) unit cell, and hexagonal closest
packing unit cell, and others. Note that the packing efficiency is not the dimension!

2006 Penn RET - Bujak Page 6 of 22

Now let’s look specifically at the Koch Curve in detail.

N(r) = # S = 1/r = Ln =
n = Stage =
Picture Line Segment Length
Iteration
Segments Length
0
⎛ 1⎞ 0

0 1=4 0 1= ⎜ ⎟ ⎛4⎞
1= ⎜ ⎟
⎝ 3⎠ ⎝3⎠

1
1 ⎛ 1⎞ 1

1 4=4 1 =⎜ ⎟ 4 ⎛4⎞
=⎜ ⎟
3 ⎝ 3⎠ 3 ⎝3⎠

2
1 ⎛ 1⎞ 2

2 16 = 4 2 =⎜ ⎟ 16 ⎛ 4 ⎞
=⎜ ⎟
9 ⎝ 3⎠ 9 ⎝3⎠

3
1 ⎛ 1⎞ 3

3 64 = 4 3 =⎜ ⎟ 64 ⎛ 4 ⎞
=⎜ ⎟
27 ⎝ 3 ⎠ 27 ⎝ 3 ⎠

4
1 ⎛ 1⎞
4 256 = 44 =⎜ ⎟
81 ⎝ 3 ⎠

n
⎛ 1⎞
n 4n ⎜ ⎟
⎝ 3⎠

log N ( r ) log 4n
d= =
⎛ 1 ⎞ log 3n
log ⎜ ⎟
Using the similarity dimension formula: ⎝r⎠
n log 4 log 4
d= = ≈ 1.26
n log 3 log 3

2006 Penn RET - Bujak Page 7 of 22

 The length of the Koch curve is greater than each Ln, so greater than every number. That
is, the Koch curve has infinite length.

Here's an interesting corollary to the infinite length of the Koch curve.

Not only does the Koch curve itself have infinite length, but measured along the curve,
the length of any pair of points in the Koch curve is infinite.

To make these computations concrete, note that if the original L0 is 1 meter, then
n = Stage = S = 1/r = Segment Ln =
Picture N(r) = # Line Segments
Iteration Length Length
24
−12 ⎛ 1⎞
24 81,474,976,710,656 = 424 3.5 x 10 m  ⎜ ⎟ 1 km
⎝ 3⎠
115,792,089,237,316,195,423, 128
570, 985,008,687,907,853,269, −62 ⎛ 1⎞
128 984,665,640,564,039,457,584, 8.5 x 10 m  ⎜ ⎟ 1 light
007,913,129,639,936 ⎝ 3⎠ year
= 4128

In summary the Koch curve has:

∞ length ⇒ d > 1
and
0 area ⇒ d < 2

which agrees with our integer dimension chart (first figure). The Koch curve dimension
is between 1 and 2. It has a fractal dimension d ≈ 1.26

Advanced students might want to study a “Koch” curve with non-equal segment lengths,
not all one-third. How and when does this affect the dimensionality? Does the shape
fold back upon itself and when? The “regular” Koch curve will not fold upon itself?
Why? Do you think that not folding upon oneself is desirable and when/why?

2006 Penn RET - Bujak Page 8 of 22

So what is the dimension of the Sierpinski triangle (or Sierpinski gasket)?

n = Stage = N(r) = # dark S = 1/r = Segment

Picture
(Iteration) Triangles Length

0
⎛ 1⎞
0 1 = 30 1= ⎜ ⎟
⎝ 2⎠

1
1 ⎛ 1⎞
1 3 = 31 =⎜ ⎟
2 ⎝ 2⎠

2
1 ⎛ 1⎞
2 9 = 32 =⎜ ⎟
4 ⎝ 2⎠

3
1 ⎛ 1⎞
3 27 = 33 =⎜ ⎟
8 ⎝ 2⎠

4
1 ⎛ 1⎞
4 81 = 34 =⎜ ⎟
16 ⎝ 2 ⎠

n
⎛ 1⎞
n 3n ⎜ ⎟
⎝ 2⎠

log N ( r ) log 3n
d= =
⎛ 1 ⎞ log 2n
log ⎜ ⎟
Using the similarity dimension formula: ⎝r⎠
n log 3 log 3
d= = ≈ 1.58
n log 2 log 2

2006 Penn RET - Bujak Page 9 of 22

More about Dimension

More examples can be found at

http://www.math.umass.edu/~mconnors/fractal/sierp/sierp.html where there is also a
great collection of worksheets for calculating dimension
http://www.math.umass.edu/~mconnors/fractal/fractal.html.

Growing Fractals - Hands-On Virtual Manipulative

Use Geometer’s Sketchpad® (computer software application) to create the classic

Sierpinski triangle (play the chaos game-very time consuming, but demonstrates order
from chaos/randomness), fractal trees, and the classic Koch curve (or snowflake) fractal -
separate handout with detailed instructions. Use an initial pattern (rule) and iterate. Note
that the Koch snowflake is simply the Koch curve placed on each of the three edges of a
triangle. Also note and experiment with segmenting the Koch curve in non equal 3-
segments. Experiment with the Koch snowflake not based on an equilateral triangle –
does not look like a snowflake, why are snowflakes symmetric and regular?

Explore starting with a square, dividing it into 9 equal smaller sub-squares and cutting
out the middle. If you iterate this pattern you will end up with a Sierpinski carpet.

Extension on fractal trees

Dynamic broccoli tree or forest showing iteration with a simple branching pattern.

2006 Penn RET - Bujak Page 10 of 22

In The Essence of Chaos, pages171-173, a fractal branching is outlined. We can use the
pattern of the Geometer’s Sketchpad® exercise “A Tree Fractal”
Begin with a vertical line segment; this is the trunk. From the top of the trunk,
draw line segments extending horizontally to either side, each six-tenths as long
as the trunk; these are the limbs. From the end of each limb , draw a segment
extending upward and one extending downward, each six-tenths as long as a limb.
Continue, alternating between horizontal and vertical segments, until lines are as
close together as the width of pencil marks. You will obtain something like
Figure 59. If you carry out the process for an infinite number of steps, you would
have a fractal, with dimension 1.356. The leaves, that is, the point to which the
branches eventually converge, also form a fractal with dimension 1.356.

2006 Penn RET - Bujak Page 11 of 22

The ratio of limb length to the trunk length need not be six-tenths. If you vary the ratio,
the following can be measured using the simple box-counting technique to find the
dimension:

Limb:Trunk ratio Dimension

1
< <1.0
2
1
1.0
2
0.6 1.356

1
2.0
2
1
> >2.0
2

Utilize box counting and come up with dimensions for each of the suggested fractal trees
and see if they agree with the above table.

So fractal objects can have integer dimensions, just like Euclidean geometric shapes.

So there is what appears to be a 2-D object that can have a dimension greater than 2.

2006 Penn RET - Bujak Page 12 of 22

Extensions

A great hands-on (time-intensive) is to create a Sierpinski tetrahedron which to all

appearances is 3-D, but in fact is exactly 2-D ! This can be done with cutting standard
small office envelopes and folding and joining the tetrahedrons in a Sierpinski “pattern.”
This can also be used to explain geometric sequences; here 1,4,16,64,256,…

An even more intense object is the Menger Sponge

which has d ≈ 2.73

2006 Penn RET - Bujak Page 13 of 22

Folding and cutting construction paper to make fractal lines, stairs, and other “shapes”

Advanced IFS: Scale, Reflect, Rotation, Translate – use specific notation used with
fractal geometry

Iterative representations of sequences and series

Explore how complex (imaginary) numbers are utilized in the Julia Sets and the more
specific Mandelbrot set. Need technology tools for this mathematically intensive and
repetitive computations.

Explore the synthetic creation and calculation of dimension for other objects:
Is a Moebius strip one dimensional and when you cut it did the dimension change? Why
or why not? Investigate “wallpaper groups” that tessellate with only 17 known “patterns”
– it’s about finding the basic shape that iterates and how it iterates through scaling(s),
rotation(s), translation(s), reflecting/mirroring. Investigate topology … try to explain the
Klein bottle.

Cantor Set (the Cantor middle-thirds set has d ≈ = 0.63)

Julia Sets (plural)
Mandelbrot Set (singular)
Mapping of set (particularly Madelbrot set of is mapping of a very special set of Julia
Sets)

Use math software which becomes necessary; especially when complex numbers is
introduced with the numerous iterative calculations in Julia Sets and Mandelbrot Set.
• Maple – computer
• Mathematica - computer
• Excel – computer
• Graphing calculator - handheld

Cellular automata (fractal grids) – especially with software like Mathematica

Space filling curves, such as the Hilbert curve (use IFS, actually easy with Geometer’s
Sketchpad®)

2006 Penn RET - Bujak Page 14 of 22

“Fractal” Toys

1. http://www.constructiontoys.com/store/fractiles.php

2. www.zometool.com

Websites:

1. http://classes.yale.edu/fractals/ - Fractal Geometry by Michael Frame, Benoit

Mandelbrot, and Nial Neger – really phenomenal site! – many links and resources.

2. http://www.dynamicgeometry.com/ - Key Curriculum Press Resource Center for

Dynamic Geometry. Search on fractal for phenomenal Geometer’s Sketchpad®
examples and files.

3. http://www.geom.uiuc.edu/ - The Geometry Center – search on fractals.

4. http://id.mind.net/~zona/mmts/geometrySection/fractals/tree/treeFractal.html -
Java-based tutorial on “Fractal Tree” – nice

5. http://math.bu.edu/people/bob/ - Bob Devaney’s site, author of 2 books in

References – amazing Java Applets and other dynamic systems resources

6. http://www.charlottemathtutor.com/htmlfile/fractal.htm - fractal links

Websites with dynamic images or virtual manipulatives

1. http://math.bu.edu/DYSYS/ - portal to The Dynamical Systems and Technology

Project at Boston University

2. http://www.shodor.org/master/fractal/ – fractal modeling tools

3. http://id.mind.net/~zona/mmts/geometrySection/fractals/tree/treeFractal.html -
fractal trees

4. http://math.bu.edu/DYSYS/chaos-game/chaos-game.html - the chaos game

5. http://didio.wikidot.com/ - Dr. Richard A. DiDio, from La Salle University,

presented the chaos game at a colloquium at Arcadia University on November 14,
2006.

6. http://www.scienceu.com/geometry/fractals/ - fractal color scheme chooser

(actually only the Mandebrot set)

2006 Penn RET - Bujak Page 15 of 22

Websites with fractal movies and presentations

1. http://www.unitedstreaming.com/ video on “Patterns, Symmetry, and Beauty”

(26:00 minutes)

2. http://www.youtube.com/results?search_query=fractal many great videos on

fractals. Note: the images are better appreciated when the viewer has a little
understanding of fractals.

3. http://www.veoh.com/ many videos on fractals.

4. http://video.google.com/ many videos on fractals.

2006 Penn RET - Bujak Page 16 of 22

References

1. An Eye for Fractals - A Graphic & Photographic Essay, Michael McGuire,

Addison-Wesley Publishing Company, Inc., 1991
- Easy book to read with many photographs and illustrations, introductory
math

2. Introducing Fractal Geometry, Nigel Lesmoir-Gordon, Will Rood, and Ralph

Edney, Totem Books, ISBN 1840461233, www.iconbooks.co.uk
- Easy book to read with many illustrations, introductory math

3. Fractals – A Tool Kit of Dynamics Activities, Jonathan Choate, Robert L.

Devaney, Alice Foster, Key Curriculum Press, 1999, ISBN 1-55953-355-2,
www.keypress.com, http://www.keypress.com/x8192.xml
- Blackline Activity Masters, great for classroom activities, part of a series
that includes 4 “activities” books: Iteration, Fractals, Chaos, and The
Mandelbrot and Julia Sets.

4. Iteration – A Tool Kit of Dynamics Activities, Jonathan Choate, Robert L.

Devaney, Alice Foster, Key Curriculum Press, 1999, ISBN 1-55953-354-4,
www.keypress.com, http://www.keypress.com/x8192.xml
- Blackline Activity Masters, great for classroom activities, part of a series
that includes 4 “activities” books: Iteration, Fractals, Chaos, and The
Mandelbrot and Julia Sets.

5. Sacred Geometry – A Tool Kit of Dynamics Activities, Stephen Skinner, Sterling

Publishing Co., New York, 20006, ISBN 1-4027-4129-4, www.sterlingpub.com
- Colorful “coffee table book”, examines math, beauty, symmetry in nature

6. The Essence of Chaos, Edward N. Lorenz, Seattle: University of Washington

Press, 1993, ISBN 0-295-97514-8
- A great book on chaos from the discoverer of Lorenz curves, includes a
nice discussion about fractality and dimensions.

7. Chaos – Making a New Science, James Gleick, New York: Penguin Books, 1987,
ISBN 0-14-009250-1
- national bestseller, probably the “popular” book that started a great deal of
interest in chaos, very readable, mostly text, hardly any math, but
worthwhile

Appendices

Worksheets and handouts for labs follow.

2006 Penn RET - Bujak Page 17 of 22

2006 Penn RET - Bujak Page 18 of 22
2006 Penn RET - Bujak Page 19 of 22
2006 Penn RET - Bujak Page 20 of 22
2006 Penn RET - Bujak Page 21 of 22
2006 Penn RET - Bujak Page 22 of 22