PATTERI zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA NlBOOHK

zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQ

DrI
FRACTALS ., ART,D and NATURE zyxwvutsrqponmlkjihgfe
This page is intentionally left blank
POTTER! BOOH
FRACTALS, ART, and NATURE

Editor
CliffordflPichoverzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLIFB

IBM Thomas J. Watson Research Center

World Scientific zyxwvutsrqponmlkjihgfedcbaZYXW
Singapore»New Jerser • London • Hong Kong
Published by
World Scientific Publishing Co. Pte. Ltd.
P O Box 128, Farrer Road, Singapore 912805
USA office: Suite IB, 1060 Main Street, River Edge, NJ 07661
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

Library of Congress CataloginginPublication Data

The Pattern Book : fractals, art, and nature / editor, Clifford A.
Pickover. p. cm.
Includes index.
ISBN 981021426X
1. Computer graphics. 1. Pickover, Clifford A.
T385.P376
745.4dc20 9447114
CIP

British Library CataloguinginPublication Data

A catalogue record for this book is available from the British Library.

First published 1995

Reprinted 1997,2000

Copyright © 1995 by World Scientific Publishing Co. Pte. Ltd.

All rights reserved. This book, or parts thereof, may not be reproduced in anyform or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to
photocopy is not required from the publisher.
Printed in Singapore by UtoPrint
Clifford A. Pickover
The Pattern Book:
Fractals, Art, and Nature

Introduction

"Art and science will eventually be seen to be as closely connected as

arms to the body. Both are vital elements of order and its discovery. The
word 'art' derives from the IndoEuropean base 'ar', meaning to join or
fit together. In this sense, science, in the attempt to learn how and why
things fit, becomes art. And when art is seen as the ability to do, make,
apply or portray in a way that withstands the test of time, its connection
with science becomes more clear."

Sven Carlson, Science News (1987)

This book will allow you to travel through time and space. To facilitate your
journey, I have scoured the four corners of the earth in a quest for unusual people
and their fascinating patterns. From Mozambique, to Asia, to many Eu ropean
countries, the contributors to The Pattern Book include worldfamous cancer
researchers, littleknown artists, and eclectic computer programmers. Some of the
patterns are ultramodern, while others are centuries old. Many of the patterns are
drawn from the universe of mathematics. To start you on the journey, I will first
provide some relevant background material on computers, pattern, science, and art.

The line between science and art is a fuzzy one; the two are fraternal
philosophies formalized by ancient Greeks like Pythagoreas and Ictinus. To day,
computer graphics is one method through which scientists and artists reunite these
philosophies by providing scientific ways to represent natural and artistic objects. In
fact many of this book's patterns were generated on small computers using simple
algorithms. Other (equally interesting) patterns were generated by human hands, and
these patterns often illustrate ornaments of both modern and ancient civilizations.
Sometimes these patterns consist of
V
vi C. A. Pickover

symmetrical and repeating designs, for example, Moorish, Persian, and other motifs
in tiled floors and cloths.
This book serves as an introductory catalog to some of the many facets of
geometrical patterns, and you are urged to explore the ideas in greater depth than
can be presented in this compendium. Perhaps I should attempt to define "pattern"
before proceeding. You can find many definitions when consulting a dictionary, for
example, "an artistic or mechanical design" or "a natural or chance configuration".
The patterns in this book have such a great diver sity that colleagues have debated
whether the shapes should really be called "patterns" at all. However, I take the
broad view, and include visually inter esting shapes and themes from all areas of
human, natural, and mathematical realms. Although the emphasis is on computer-
generated patterns, the book is informal, and the intended audience spans several
fields. This book might be used by students, graphic artists, illustrators, and
craftspeople in search of vi sually intriguing designs, or anyone fascinated by
optically provocative art. In addition, the book may be used by scientists, artists,
laypeople, programmers and students. In the same spirit as Gardner's book,
Mathematical Circus, or Pappas' book, The Joy of Mathematics, The Pattern Book
combines old and new ideas — with emphasis on the fun that the true pattern lover
finds in doing, rather than in reading about the doing! The book is organized into
three main parts: Representing Nature (for those patterns which describe or show
real physical phenomena, e.g., visualizations of protein motion, sea lil lies, etc.),
Mathematics and Symmetry (for those patterns which describe or show
mathematical behavior, e.g., fractals), and Human Art (for those patterns which
are artistic works of humans and made without the aid of a computer, e.g., Moslem
tiling patterns). I provide a comprehensive glossary to help ease readers into
technical or unfamiliar waters.

When deciding how to arrange material within the three parts of The Pat tern
Book, many divisions came to mind — computer and noncomputer gen erated
forms, science and art, nature and mathematics. However, the line between all of
these categories becomes indistinct or artificial, and I have there fore randomly
arranged the patterns within each part of the book to retain the playful spirit of the
book and to give the reader unexpected pleasures. Some patterns could easily be
placed in either of the three main sections of the book.
The reader is forewarned that some of the presented material in this book's
catalog of shapes involves sophisticated concepts (e.g., "The Reversible
GreensbergHastings Cellular Automaton" by Drs. P. Tamayo and
The Pattern Book: Fractals, Art, and Nature vii

H. Hartman) while other patterns (e.g., "Satanic Flowers" by Dr. Harold J.

McWhinnie) require little mathematical knowledge in order to appreciate or
construct the shapes. Readers are free to pick and choose from the smorgas bord of
patterns. Many of the pattern descriptions are brief and give the reader just a flavor
of an application or method. Additional information is often in the referenced
publications. In order to encourage reader involvement, compu tational hints and
recipes for producing many of the computerdrawn figures are provided. For many
readers, seeing pseudocode will clarify the concepts in a way which mere words
cannot.
Currently, I know of no book which presents such a large range of patterns and
instructions for generating the patterns. There are, however, numerous books
available that publish patterns in specific categories. Most are inexpen sive
paperbacks published by Dover Publications, and many are reprints of nineteenth
century books. I think you will enjoy these. Some are listed in the reference section.

Before concluding this preface, I should point out that today scientists and artists
seem to have a growing fascination with symmetry and repetition in design. On the
topic of art, there are the modern isometric designs of John Locke and the
geometrical ornaments of Russian artist, Chernikow (where simple forms create
complex interweavings),

and a variety of popular art deco designs. Also "controlled accident" has found its
place in many areas of the modern arts (O'Brien, 1968). For example, Dadaist and
Surrealist painters such as Miro, Masson, and Arp capitalized on the elements of
chance, and the works they created provide challenges for the mind as well as the
eye. In the area of science, researchers are intrigued by the way nature often
expresses itself in terms of repeating symmetries — and the cross section of plants,
phase transitions, standing waves on metal plates, muscle striations, snow crystals,
and dendritic ice are just a few examples.
viii C. A. Pickover

From the branching of rivers and blood vessels, to the highly convoluted sur face of
brains and bark, the physical world contains intricate patterns formed from simple
shapes through the repeated application of dynamic procedures. Questions about the
fundamental rules underlying the variety of nature have led to the search to identify,
measure, and define these patterns in precise scientific terms.

One final observation on patterns in nature. Our physical world around us often
seems chaotic, exhibiting a limitless and complex array of patterns. How ever, you
should note that our world is also actually highly structured. From an evolutionary
standpoint, biological themes, structures, and "solutions" are repeated when
possible, and inanimate forms such as mountains and snowflakes are constrained by
physical laws to a finite class of patterns. The apparently intricate fabric of nature
and the universe is produced from a limited variety of threads which are, in turn,
organized into a multitude of combinations. You will see some of these threads
throughout this book.

The World of Fractals and Chaos

Many of the patterns in this book come from the exciting mathematical fields of
fractal geometry and chaos. This section is intended as a brief intro duction to these
fields.
These days computergenerated fractal patterns are everywhere. From squiggly
designs on computer art posters, to illustrations in the most seri ous of physics
journals, interest continues to grow among scientists and, rather surprisingly, artists
and designers. The word "fractal" was coined in 1975 by IBM scientist Benoit
Mandelbrot to describe a set of curves rarely seen before the advent of computers
with their ability to perform massive numbers of cal culations quickly. Fractals are
bumpy objects which usually show a wealth of detail as they are continually
magnified. Some of these shapes exist only in abstract geometric space, but others
can be used to model complex natural shapes such as coastlines and mountains.

Chaos and fractal geometry go handinhand. Both fields deal with intricately
shaped objects, and chaotic processes often produce fractal patterns. To ancient
humans, chaos represented the unknown, the spirit world — menacing, nightmarish
visions that reflected man's fear of the irrational and the need to give shape and
form to his apprehensions. Today, chaos theory is a growing field which involves
the study of a range of phenomena exhibiting a sensitive dependence on initial
conditions. This means that some natural systems, such
 The Pattern Book: Fractals, Art, and Naturezyxwvutsrqi

as the weather, are so sensitive to even small local fluctuations that we will never be
able to accurately predict what they will do in the future. For certain mathematical
systems, if you change a parameter ever-so-slightly, the results can be very
different. Although chaos seems totally "random", it often obeys strict mathematical
rules derived from equations that can be formulated and studied. One important
research tool to aid the study of chaos is computer graphics. From chaotic toys with
randomly blinking lights to wisps and eddies of cigarette smoke, chaotic behavior is
irregular and disorderly. Other examples include certain neurological and cardiac
activity, the stock market, and some electrical networks of computers. Chaos theory
has also often been applied to a wide range of visual art.

So extensive is the interest in fractals and chaos that keeping up with the
literature on the subject is rapidly becoming a full-time task. In 1989, the world's
scientific journals published about 1,200 articles with the words "chaos"
or "fractal(s)" in the title. The figure here shows the number of papers with zyxwv
F i g u r e 1. Chaos and fractal article explosion. A review of the world scientific literature between 1973
and 1990 shows the number of chaos and fractal articles rising dramatically between the years 1982 and
1990. (Figure from Computers and the Imagination by C. Pick-over, ©199 1 St. Martin's Press. All rights
reserved.)
x C. A. Pickover

titles containing the words "chaos" or "fractal(s)" for the years 19751990, the 1990
values estimated from data for JanuaryJune 1990.

The Best of the Best

After the patterns in this book were compiled, I invited six distinguished judges
to help select the "best" patterns in the book. Naturally selecting from such a diverse
collection was not an easy or welldefined task. The judges selected patterns that
they considered beautiful, novel, and/or scientifically interesting. I thank the
following individuals on the "board of judges" for singling out their favorite
patterns for special recognition in the book.

1. Professor Akhlesh Lakhtakia, Pennsylvania State University. Editorin Chief,

Speculations in Science and Technology; Milestone Editor, Selected Papers on
Natural Optical Activity; Coauthor, Timeharmonic Electromag netic Fields in
Chiral Media; member, The Electromagnetics Academy.
2. Dr. Roger Malina, Editor of Leonardo, a journal of the International Society for
the Arts, Sciences, and Technology.
3. Ivars Peterson, Author of The Mathematical Tourist and Islands of Truth, and
mathematics and physics editor at Science News.
4. Chris Severud, President of Bourbaki Software, Inc.
5. Dawn Friedman, Chemistry Department, Harvard University. Naturalist,
sciencewriter, theoretical chemist, futurist.
6. Phil LoPiccolo, Editor, Computer Graphics World.

The First Place Prize was awarded to Steven Schiller of Adobe Systems,
California. His "Gaussian Fraction" pattern was judged the best because of its
aesthetic quality, mathematical interest, and novelty. One of the more poetic judges
exclaimed: "The pattern is a perfect daydream generator. It reminds me of Dirac's
ocean of negativeenergy electrons, with stray points fountaining from the surface
like solar prominences, and unseen quantum events producing shimmers deep
undersea."
There was a threeway tie for the second place prize. The Second Place Prize
was awarded to Earl Glynn of Kansas for his "Spiraling Tree/Biomorphic Cells"
pattern, Dr. Ian Entwistle of the UK for his pattern, "Serpents and Dragons: A
mapping of f(z) —» sinh(z) + c in the complex plane", and to the late Ernst Haeckel
for the "SeaLilies" pattern. Of the "Serpents and Dragons", one judge noted:
"Among dozens of stunning patterns based on Julia sets, this was simply the most
beautiful. Beauty and order seem to swim
 The Pattern Book: Fractals, Art, and Nature xi

upwards out of a chaotic sea like vigorous fish, growing and uncurling like fern
fronds reaching for the light." Of the sealilies, another judge noted that "these
beautifully intricate forms owe as much to the artist's eyes as they do to the natural
shapes of the organisms themselves; this is typical of the precision and artistry
found in much 19thcentury illustration".
The remaining prizes go to Stefan Muller, Theo Plesser, and Benno Hess of
Germany for their patterns on "Rotating Spiral Waves in the Belousov Zhabotinskii
Reaction", A. K. Dewdney of Canada for his "Informal Tessela tion of Cats", John
MacManus of Canada for his "Jungle Canopy", and Henrik Bohr and Soren Brunak
of Denmark for their "Patterns of Protein Conforma tions". Of the rotating spiral
wave patterns, one judge noted: "Ever since I first heard of and then actually
demonstrated for myself the existence of oscillating chemical reactions, I have
found both the chemistry and the mathe matics of these patterns irresistible. The
spirals have wonderfully complicated symmetries." Another judge remarked:

"Naturally, as a chemist, I am delighted to see a beautiful chemi cal

pattern, its spiral forms enhanced by an ingenious and skillful piece of
laboratorycumvideo work. But I chose this pattern for another reason as
well. The aim of this book, other than to delight, must be to elucidate.
What happens in the murky swirling fluid inside a reaction vessel is the
ancient mystery of chemistry, darker and more forbidding than the
linked equations of thermodynam ics or the fortress walls of quantum
matrices. Many students of chemistry are reduced to following recipes
and pouring ingredients, hoping the magic will work as promised. The
discovery of struc ture, pattern, beauty, inside that murky flask, the
casting of light into the darkness, must encourage us all."

Of Dewdney's cats, one judge noted:

"I'm sure I won't be the first or the last to pick this one. I admire Dr.
Dewdney's unspoken point: if you are going to tile a plane with a
species, by all means choose one that is naturally graceful, flexible, and
incapable of uncomfortable angularities."
xii C. A. Pickover

For Further Reading

"Some people can read a musical score and in their minds hear the
music ... Others can see, in their mind's eye, great beauty and structure
in certain mathematical functions ... Lesser folk, like me, need to hear
music played and see numbers rendered to appreciate their structures."

Peter B. Schroeder, Byte Magazine (1986)

As you will see in many patterns from this book, mathematical formulas can
sometimes be used to simulate natural forms. For example, computer graphics
provides a way to represent biological objects. For an excellent book on techniques
for simulating nature, see Rivlin (1986). Researchers have ex plored the use of rules
based on the laws of nature, such as logarithmic spirals for sea shells (Kawaguchi,
1982) or tree branching patterns determined from the study of living specimens
(Aono, 1984). Other papers describe the genera tion of plant leaf vein patterns
(Kolata, 1987) and woodgrains (Yessios, 1979). Bloomenthal (1985) describes
methods for simulating tree bark, leaves, and limbs. Other sophisticated approaches
to botanical structure generation exist, for example, beautiful "particle systems"
consisting of trajectories of particles influenced by the pull of gravity (Reeves,
1985). See also (Viennot et al., 1989; Prusinkiewicz et al., 1988). For references on
symmetry in historical orna ments, see Audsley (1968) and Rozsa (1986). Audsley's
book includes illustra tions of ancient Egyptian patterns from the painted ceiling of
various tombs, interlaced Celtic designs typical of those used to illuminate
manuscripts, and various Japanese ornaments. For a fascinating collection of
Persian designs and motifs, see Dowlatshahi (1979). Symmetrical ornaments, such
as those presented in The Pattern Book, have persisted from ancient to modern
times. The different kinds of symmetries have been most fully explored in Arabic
and Moorish design. The later Islamic artists were forbidden by religion to repre
sent the human form, so they naturally turned to elaborate geometric themes. To
explore the full range of symmetry in historic ornaments, you may wish to study the
work of Gombrich who discusses the psychology of decorative art and presents
several additional examples of fivefold symmetry.

The following reference list includes books and papers describing patterns in a
range of scientific and artistic fields.
 The Pattern Book: Fractals, Art, and Nature xiii

Allen, J. (1988) Designer's Guide to Japanese Patterns. Chronicle Books: San

Francisco.
Audsley, W. (1968) Designs and Patterns from Historic Ornament. Dover: New
York.
Barnard, J. (1973) The Decorative Tradition. The Architectural Press: London.
Barratt, K. (1980) Logic and Design in Art, Science, and Mathematics. Design
Press: New York. (The pictures alone will stimulate readers to experiment
further.)
Blossfeldt, K. (1985) Art Forms in the Plant World. Dover: New York. Doczi, C.
(1988) The Language of Ornament. Portland House: New York. Doczi, G. (1986)
"Seen and unseen symmetries", Computers and Mathematics
with Applications 12B:3962.
Dowlatshahi, A. (1979) Persian Designs and Motifs. Dover: New York.
Durant, S. (1986) Ornament. McDonald and Company: London.
Gardner, M. (1969) "Spirals", in The Unexpected Hanging. Simon and Schuster:
New York.
Gardner, M. (1970) Mathematical Circus. Penguin Books: New York.
Glazier, R. (1980) Historic Ornament. Coles: Toronto.
Gombrich, E. (1979) The Sense of Order: A Study in the Psychology of Deco rative
Art. Cornell University Press: New York.
Grunbaum, B., Grunbaum, Z. and Shephard, G. (1986) "Symmetry in Moor ish and
other ornaments", Computers and Mathematics with Applications
12B:641653. Hargittai, I. (1989) Symmetry 2: Unifying Human Under standing.
Oxford: New York.
Hargittai, I. and Pickover, C. (1992) Spiral Symmetry. World Scientific: New
Jersey.
Harlow, W. (1976) Art Forms from Plant Life. Dover: New York.
Hayes, B. (1986) "On the bathtub algorithm for dotmatrix holograms",
Computer Language 3:2125.
Justema, W. (1976) Pattern: A Historical Perspective. New York Graphic Society:
Massachusetts.
Lockwood, E. and Macmillan, R. (1978) Geometric Symmetry. Cambridge
University Press: New York.
Lewis, P. and Darley, G. (1986) Historic Ornament: A Pictorial Archive.
Pantheon Books: New York.
Makovicky, E. (1986) "Symmetrology of art: coloured and generalized symme
tries", Computers and Mathematics with Applications 12B:949980.
xiv C. A. Pickover

Mamedov, K. (1986) "Crystallographic patterns", Computers and Mathematics with

Applications 12B:511529.
Mandelbrot, B. (1983) The Fractal Geometry of Nature. Freeman: San Francisco.

Moon, F. (1987) Chaotic Vibrations. John Wiley and Sons: New York.
Peterson, I. (1988) The Mathematical Tourist. Freeman: New York.
Pickover, C. (1990) Computers, Pattern, Chaos, and Beauty. St. Martin's Press:
New York.
Pickover, C. (1991) Computers and the Imagination. St. Martin's Press: New York.

Pickover, C. (1992) Mazes for the Mind. St. Martin's Press: New York. Pickover, C.
(1994) Chaos in Wonderland: Visual Adventures in a Fractal
World. St. Martin's Press: New York.
Pickover, C. (1995) Keys to Infinity. Wiley: New York. Postle,
D. (1976) Fabric of the Universe. Crown: New York.
Racinet, A. (1988) The Encyclopedia of Ornament. Portland House: New York.
Reitman, E. (1989) Exploring the Geometry of Nature. Windcrest Books:
Pennsylvania.
Reichardt, J. (1969) Cybernetic Serendipity: The Computer and the Arts.
Prager: New York.
Rucker, R. (1982) Infinity and the Mind. Bantam: New York.
Shaw, A. (1984) The Dripping Faucet as a Model Chaotic System. Aerial Press:
California.
Stevens, C. (1989) Fractal Programming in C. M and T Books: California. (This
book is a dream come true for computer programmers interested in fractals.)

Steinhaus, H. (1983) Mathematical Snapshots, 3rd ed. Oxford University Press:

New York. (Topics include tesselations, soapbubbles, maps, screws, spiders,
honeycombs, and platonic solids.)
"Symmetries and Asymmetries" (1985) Mosaic 16 (An entire issue on the subject
of fractals, symmetry and chaos. Mosaic is published six times a year as a
source of information for scientific and educational communities served by the
National Science Foundation, Washington DC 20550).
Tufte, E. (1983) The Visual Display of Quantitative Information. Graphics Press:
Connecticut.
Niman, J. Norman, J. and Stahl, S. (1978) "The teaching of mathematics through
art" (a report on the conference March 2021, 1978) Metropolitan
 The Pattern Book: Fractals, Art, and Nature xv

Museum of Art and Hunter College of the City University of New York. pp. 1-
55.
O'Brien, J. (1968) How to Design by Accident. Dover: New York.
Pappas, T. (1990) The Joy of Mathematics. Wide World Publishing: California.
Peachey, D. (1985) "Solid texturing of complex surfaces", Computer Graphics
(ACM SIGGRAPH) 19(3): 279286.
Perlin, K. (1985) "An image synthesizer", Computer Graphics (ACM SIGGRAPH)
19(3): 287296.
Postle, D. (1976) The Fabric of the Universe. Crown Publishers Inc.: New York.

Rozsa, E. (1986) "Symmetry in Muslim arts", Computers and Mathematics with

Applications 12B:725750.
Viennot, X., Eyrolles, G., Janey, N. and Arques, D. (1989) "Combinatorial analysis
of ramified patterns and computer imagery of trees", Computer Graphics (ACM-
SIGGRAPH) 23(3): 3140.
Zvilna, J. (1986) "Colored symmetries in spacetime", Computers and Mathe matics
with Applications 12B:895911.

Acknowledgments

Some of the figures in the book come from the Dover Pictorial Archive of
modern and ancient art. This series constitutes a collection of the world's greatest
designers from Ancient Egypt to Art Deco designs. For more infor mation, write to
Dover Publications, 31 East 2nd Street, Mineola, New York 11501.

The opening quotation by Sven G. Carlson on art and science appeared in his
letter to Science News (Vol. 132, 1987, p. 382).
This page is intentionally left blank
Contents

The P attern Book: Fractals , Art , a nd Nature v

C. A. Pickover

Part I: Representing Nature

Evolution of the Solar a nd Pl anetary Vortices 3

K. L. Cope

Rotating Spiral Waves in the Belousov-Zhabotinskii Reaction 4

S . C. Miiller, T. Pl esser and B . Hess

Patterns of Protein Conformations 6

H. Bohr and S . Brunak

Blind Watchmaker Biomorphs 9

R . Dawkins

Apple Tre e Pattern 12

P . Brodatz

Spring Til es t he Planes of Sunlight : Decu ssate and Tri cussate

Phyllotaxy in New Growth 14
D . Friedman

Traj ectories of a Neural Network Quantizer in Rhythm Space 18

P. Desain and H. Hon ing

An Internally Mechanistic Nucleon 22

J. B oivin

Wood Pattern 24
P . Brodatz

Moi re P attern 26
H. Giger
XV
ll
xviii Contents

Fractal Phyllotactic Spirals: Broccoli "Minaret" 28

D. Friedman

Puppy Representation of DNA Nucleotide Sequences 30

U. Melcher

Stone Wall Pattern 32

K. Miyata

Novel Representations, Pattern, Sound 36

A. Peevers

Haeckel's SeaLilies 38
C. A. Pickover

Diffusion Limited Aggregation 40

D. Platt

RNA Structure Based on Prime Number Sequence 42

M. Sluyser and E. L. L. Sonnhammer

Genetic Patterns as shown by Sequence Logos 44

T. D. Schneider

Part II: Mathematics and Symmetry

Gaussian Fractions 49
S. Schiller

Lattice Design 1 52
/. 0. Angell

Goldbach's Comet 55
H. F. Fliegel and D. S. Robertson

JCurve (OLFRAC) (14 Iterations) 57

G. Jagoda

Carpet (13 Iterations) 58

G. Jagoda
 Con ten ts xi x

3D-Cubes 60
G. Jagoda

The-End 62
G. Jagoda

The Ikeda Attractor 64

C. A. Pickover

Star-NLPI-5m 66
J . Rangel-Mondragon and S . J . Abas

Whirlpools 68
I. Astrahan

Fern- Wolf (Fili cinae Lupus) 70

I . Astrahan

Magic Tessaract 72
J . R . Hendricks

The Rever sibl e Gr eenberg-Hastings Ce llula r Automaton 74

P. Tamayo and H. Hartman

Cy be rnetic Rapids 77
I . A strahan

Pinwheels 79
C. W . Hend erson

Circlefest 81
C. W . Hend erson

XOR Size 83
C. W . Henderson

Trig Gem 85
C. W. Hend erson
xx Contents

Locked Links 87
G. Ricard

Exponential Tunnel 90
G. Ricard

4
A Generalization of the Regular Tiling (4 ) 93
H. Okumura

A Pattern by Fujita Configurations 96

H. Okumura

Star Trails 99
K. J. Hooper

An Iteration Map 103

D. Scruton

A SelfSimilar Structure Generated by a BrokenLinear Function 104

M. Szyszkowicz

Roots of Algebraic Polynomials 106

M. Szyszkowicz

Mosaics 108
R. Telgdrsky

Pattern of Euler's Formula 112

M. Szyszkowicz

SelfAccelerating Version of Newton's Method 114

M. Szyszkowicz

Voronoi Fractal 116

K. Shirriff

The Logistic Map in the Plane 118

M. Szyszkowicz
 Contents xxi

Modified Logistic Map in the Plane 120

M. Szyszkowicz

Newton's Method in the CPlane 122

M. Szyszkowicz

Iterations with a Limited Number of Executions 124

M. Szyszkowicz

The Starry Night — Iterates of Tan(z) 126

S. D. Casey
2
From Asymmetry to Symmetry — Iterates of Exp(z) and Exp(z ) 129
S. D. Casey

A Variation on a Curve of Mandelbrot 132

S. D. Casey and N. F. Reingold

An Asymmetric Sierpinski Carpet 135

S. D. Casey and N. F. Reingold

The "Computer Bug" as Artist — Opus 3 138

S. D. Casey

The Last Word in Fractals 140

E. M. Richards

TopsyTurvy Fractal Growth 142

E. M. Richards

Symmetrized DotPatterns and Cellular Automata 144

W. K. Mason
z
"Carpet of Chaos": Mapping the Bilateral Symmetric z + c in
the Complex z and c Planes 147
/. D. Entwistle

"Spidermorphs": Mappings from Recursion of the Function

z z4 4
z^z »+z z+ z z+c 151
Ian D. Entwistle
xxii Contents

"Fractal Turtle" and "Elephant Star": Multiple Decomposition

5
Mappings of f(z) : z —> z + c in the Complex Plane 155
/. D. Entwistle

4
"Floral Table": A Mapping of the Function z —► z z + c in
the Complex z Plane 159
/. D. Entwistle

"Entrapped Lepidoptera": An Alternative Mapping of a Julia Set

4
for the Function z —> z — z + c 163
/. D. Entwistle

SelfMapping of Mandelbrot Sets by Preiteration 166

S. D. Calahan and J. Flanagan

Unrolling the Mandelbrot Set 169

S. D. Calahan

Parameter Space of Mandelbrot Sets from Variation of Coefficients 172

S. D. Calahan

Lizards 176
N. Chourot and V. Wright

FourDimensional Space Flowers 178

K. Miyazaki and M. Shiozaki

The Dolphin Head Bifurcation Diagram 180

/. Lengyel, I. R. Epstein and H. Rubinstein

Embellished Lissajous Figures 183

B. Brill

Islamic Pattern 194

S. J. Abas

Swirl 197
P. D. Bourke
 Contents xxiii

Circlegraph 7.11 199

C. Cassin

Flow 2.8 200

C. Cassin
3
Botanical Biomorph Arising from the Iteration of z —► z + ^ 201
J. D. Corbit

Transient Microstructure 203

W. H. Cozad

Engel Curves 205

D. A. Engel

"Spinning Chaos": An Inverse Mapping of the Mandelbrot Set 208

/. D. Entwistle

"Wings of Chaos": Mappings of the Function sinh(z) + c in the

Complex Plane 211
/. D. Entwistle

"Islands Among Chaos": Mappings of the Transcendental Function

cosh(z) + c in the Complex Plane 215
/. D. Entwistle

Alternative Methods of Displaying the Behavior of the Mapping

2
z+z +c 221
I. D. Entwistle

"Serpents and Dragons": A Mapping of

f(z) —> sinh(z) + c in the Complex Plane 225
/. D. Entwistle

"Star of Chaos": A Multiple Decomposition Mapping of

7
f(z) : z —> z + c in the Complex Plane 228
/. D. Entwistle

Complex Branchings 231

M. Frame and A. Robucci
xxiv Contents

Nevada Sets 233

P. Gibone

"Handmade" Patch Quilt Generated by LSystem 235

/. Glickstein

Spiraling Tree/Biomorphic Cells 238

E. F. Glynn

Fractal Hurricane 240

E. F. Glynn

Intertwined 242
E. F. Glynn

2
Mandelbrot Iteration z —> z~ + c 244
E. F. Glynn

2
Julia Iteration z —► z~ + c 246
E. F. Glynn

Bessel Fireworks and Bessel Parallelogram 248

E. F. Glynn

Fractal Moon 250

E. F. Glynn

Isogonal Decagons 251

B. Griinbaum

Analytic Computer Art 254

J. Jacobson

Euler's Triangle 256

W. J. Jones

Euler's Crossing I and II 258

W. J. Jones
 Contents xxv

Stripes 260
W. J. Jones

Outer Space 262

W. J. Jones

Parquet 264
W. J. Jones

TESS Patterns 266

W. J. Jones

Skew Squares 268

W. J. Jones

Fractal Gaskets 269

H. Jones

An Octahedral Fractal 273

H. Jones

A Packing with Icosahedral Symmetry 1 276

H. Lalvani and N. Katz

Hard Lines, Soft Circles 278

H. Lalvani

Tesspoly 3, into Tesspoly and Sectus 280

H. Lalvani

Tetrahedron 283
W. Li

Binary Matrix Symmetry 285

S. de Vos and R. G. Lauzzana

TriHadamards 288
R. G. Lauzzana and S. de Vos
xxvi Contents

Fractal Limit — A Fractal Pattern Emulating M. C. Escher 290

J. E. Loyless

Fractal Limit — A Perspective View 292

J. E. Loyless

Jungle Canopy 295

J. MacManus

A "Y" Hexomino Tiling 297

T. Marlow

Beauty in Functions 299

D. S. Mazel

Logistic 3 302
L. K. Mitchell

Inflation Rules 3 304

L. K. Mitchell

Hommage a Durer 307

V. Molnar and F. Molnar

The Center of Dewdney's Radiolarian 309

D. H. Moore

2D Feigenbaum 311
J. Peinke, J. Parisi, M. Klein and 0. E. Rossler

Biomorphic Mitosis 313

D. Stuedell

Triangular Numbers and the Distribution of Primes 315

J. Nugent

Bistable Tiling Patterns with Converging/Diverging Arrows 317

T. V. Papathomas
 Contents xxvii

Vivid Depth Percepts from Simple GrayLevel Line Patterns 319

T. V. Papathomas

Mandelbrot Set Spiral Tile Mosaics 322

K. W. Philip

Evolution of Spirals in Mandelbrot and Julia Sets 331

A. G. D. Philip

Hyperbolic Tilings 339

U. Quasthoff

Products of Three Harmonics 342

C. A. Reiter

Pascalian Cellular Automata 344

D. E. Smith

Inverse Mandelbrot Art Prints 347

F. G. Szabo

Bio Art Prints 350

F. G. Szabo

Rotation 353
F. G. Szabo

Wallpaper from GCONTOUR Procedure 354

M. Szyszkowicz

Pattern Generated by XOR 356

M. Szyszkowicz

Representation of a Surface on the Plane 358

M. Szyszkowicz

Composition of Color Pieces 360

M. Szyszkowicz
xxviii Contents

Patterns Composed with Squares 362

M. Szyszkowicz

Eucledian Geometry using IFS 365

Daniel T. H. Tan, G. N. Toh, L. C. Liong and S. K. Yek

The Galactic Virus Attack 368

D. Walter

Sound Ornaments 370

W. Weber

Part III: Human Art

An Informal Tesselation of Cats 375
A. K. Dewdney

Op Art 1 376
J. Larcher

Extension(s) of a Reconstructed Tamil RingPattern 377

P. Gerdes

Japanese Optical and Geometrical Art 1 380

H. Ouchi

Greek Landscape — A Visual Diary 382

J. Frank

Persian Designs and Motifs 384

A. Dowlatshahi

Symmetric Celtic Sycophancy 386

A. Mackay

Horoscope 388
E. Schoen

Bizarre and Ornamental Alphabets 389

C. B. Grafton
 Contents xxix

Ambiguous Art 391

P. Hettich

Japanese Diaper Ornaments 394

W. and G. Audsley

Satanic Flowers 396

H. J. McWhinnie

How to Create Celtic Anthropomorphic Ornaments (BeardPullers) 398

G. Bain

Clowns 400
S. Kreuzer

Celtic Plants Emerging from Pots 402

G. Bain

Interactive Patterns Between Light and Matter 403

Y. Nakano

Art Deco Design 1 407

M. Loeb

Periodic Pattern of Class 19 409

F. Fernandez

Turtle Moon Artists' Logo and Chop Mark 411

W. Tait

Glossary 413

Index 425

Proceeding pages (pp. xxxxxxvi)

Some of the winning designs: Gaussian Fractions; Haeckel's SeaLilies; Serpents and Drag ons; Patterns
of Protein Conformations; Spiraling Tree/Biomorphic Cells; Jungle Canopy (landscape); An Informal
Tesselation of Cats.
Kevin L. Cope
Evolution of the Solar and
Planetary Vortices

This pattern represents the emergence of heavenly bodies from primitive ce

lestial vortices as imagined by Gabriel Daniel, a late seventeenthcentury com
mentator on Descartes. More than a few controversialists of the earlymodern era
made their careers by reviewing, revising, and interpreting Descartes' re markable
vortex theory. These overly ambitious scribblers applied a plethora of methods to
the French physicistphilosopher, trying everything from science to satire. Descartes
had argued that the universe was a plenum, an infinitely large container completely
filled with extended matter and uncorrupted by empty spaces. To explain the
differences in the densities of objects in a uni verse allegedly packed as tightly as
possible, Descartes conjectured that his cosmos might be pockmarked with
"vortices". Energetic, unstable, and even dangerous, these whirlpools might draw
some collections of matter into tighter quarters than others. Turbulence reigns
supreme in a Cartesian universe. Ac cording to vortecticians, matter begins as a
plasmalike whirlpool, condenses into a furious sun, calms into a jiggling "terrela" (a
luminous, gaseous, whirling, or incomplete planet like Jupiter or Saturn), and, at
last, settles into a plan etary lump. The pattern shown here portrays the dynamic,
even terrifying interference of the great, originative vortices that were eventually to
become the components of our solar system. Gabriel Daniel was himself a rather un
stable fellow. After writing several hundred pages of satire against Descartes'
theories, he apparently developed an affection for the great geometer, consol ing
himself that Descartes had never really died, but had only been carried up into some
higher vortex by means of his own whirling tobacco smoke [1].

Reference
1. G. Daniel, A Voyage to the World of Cartesius (Thomas Bennett, 1694)
(translated by Taylor, T.).
3
Stefan C. Miiller, Theo Plesser and Benno Hess
Rotating Spiral Waves in the
BelousovZhabotinskii Reaction

Describe here is a macroscopic chemical pattern occurring on a 1 mm scale in a

chemical excitable solution which is maintained under conditions far from
thermodynamic equilibrium. The BelousovZhabotinskii reaction has become the
most prominent chemical model example of an excitable medium displaying spatio-
temporal selforganization. In this reaction due to the nonlinear interac tion of
complex reaction kinetics with molecular diffusion, spiralshaped waves of chemical
activity are readily observed [1, 2]. For quantitative pattern anal ysis a
computerized, videobased twodimensional spectrophotometer has been designed
which yields digitized highresolution data of the spatiotemporal evolution of the
patterns [3, 4]. Picture A of the figure shows a complex wave pattern in a thin
solution layer of this reaction, as observed in transmitted light of appropriately
selected wavelength and recorded by a video camera.
In order to characterize the properties of the core regions of the displayed
spirals, a digital overlay technique is applied to a series of stored images. The result
of this overlay is depicted in picture B of the figure, showing a black spot close to
each of the spiral tips of picture A. These spots indicate the sin gular features of the
spiral cores in that, at these particular sites, the chemical state of the medium
remains constant in time, whereas at all points outside the core region the medium
undergoes an oscillatory transition between two states of maximum and minimum
excitation. By applying threedimensional perspective techniques [5] the structure of
the cores can be visualized in detail (picture C of the figure), thus emphasizing their
singular properties.

References
1. R. J. Field and M. Burger, eds., Oscillations and Travelling Waves in Chemical
Systems (John Wiley, 1986).
2. J. Ross, S. C. Miiller and C. Vidal, "Chemical waves", Science 240 (1988) 460-
465.
3. S. C. Miiller, Th. Plesser and B. Hess, "Twodimensional spectrophotometry of
spiral wave propagation in the BelousovZhabotinskii reaction. I. Experiments
and digital data representation", Physica D 24 (1987) 7186.
4
 Rotating Spiral Waves in the BelousovZhabotinskii Reaction 5

4. S. C. Miiller, Th. Plesser and B. Hess, "Twodimensional spectrophotometry and pseudocolor

representation of chemical reaction patterns", Naturwissenschaften 73 (1986) 165179.

5. S. C. Miiller, Th. Plesser and B. Hess, "Threedimensional representation of chemical gradients",

Biophys. Chem. 26 (1987) 357365.
 Figure 1. (A) Snapshot of a spiral wave pattern in the BelousovZhabotinskii reaction. (B) Digital overlay
of six consecutive images covering one revolution of the spirals of image (A).
(C) Threedimensional perspective representation of the composite image (B) seen from a view point
located below the upper edge of (B).
Henrik Bohr and S0ren Brunak
Patterns of Protein Conformations

The figure visualizes an ensemble of threedimensional conformations of the

protein, Avian Pancreatic Polypeptide (APP), obtained by a method of
optimization which is based on an analogy of the Traveling Salesman Problem.
The dynamical behavior of complex systems like proteins is in general ex
tremely hard to simulate on a computer. It has been suggested that the reason might
be that many complex systems actually function as computers, per forming
computations which cannot be completed in fewer logical steps than the systems
are using themselves. In other words, the computations are irre ducible and the
computational results cannot be obtained by the use of short cuts, because they do
not exist [1].
In the case of the dynamics of proteins, where parts of the proteins move
relatively to each other, the question is: What is the protein computing while it
moves, and what is the final result of the computation?
The result of the protein folding process is clearly a topology. The process looks
for good neighbors for the various parts, and the system ends up in a
conformational state with high stability in which parts with an affinity for each
other (or a dislike of the solvent) are brought together, and where mutually
repulsive parts seek positions with the highest possible distance between them.
Conformational substates of a protein are partly determined by the posi tions of
the sidechains [2]. The multiple configurations of these sidechains and their
stability is the subject of this study [3]. A change in the position of one sidechain
can have global effects on the positions of other sidechains. We have considered a
model where only the nearest neighbor interactions are included. This assures that
the dynamics of the sidechain configurations are collective. We are then left with
the problem of choosing a suitable neigh bor topology for the threedimensional
configuration of sidechains, being the result of a global optimization of interaction
energy throughout the protein.
In the classical optimization problem of the traveling salesman, a similar
situation arises: how to choose — for each city — two neighboring cities which
minimizes the overall length of a round trip between the cities. In this analogy, the
(moving) protein sidechains take the place of the cities, and the cost function
becomes the sum of the interaction between neighboring sidechains.
6
 7

F i g u re [Link]
Polypeptide comprising 36 amino acids. The cones represent degrees of freedom for each side-chain.
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
8 H. Bohr & S. Brunak

The actual optimization procedure starts with the sidechains in random positions on the
cones (representing the degrees of freedom for each sidechain) and a random choice for the
nearest neighbor topology. The optimization pro ceeds as simulated annealing and the tour
becomes stable at a certain temper ature where all the sidechains have relaxed accordingly
into a conformation of low global energy. The result of the optimization is presented in the
fig ure, where a sample of one hundred subconformations for APP is shown with the
distribution over sites in the cone state spaces. The weight of each site is not shown. This
picture of possible conformational substates is in reasonable agreement with existing data
[4].

References
1. S. Wolfram, "Complexity engineering", Physica 22D (1986) 385.
2. A. Ansani et al, "Protein states and protein quakes", Proc. Natl. Sci. 82 (1985) 5000.

3. H. Bohr and S. Brunak, "A travelling salesman approach to protein conformation",

Complex Systems 3 (1989) 9.
4. I. Glover et al, "Conformational flexibility in a small globular hormone: Xray analysis of
avian pancreatic polypeptide at 0.98 angstroms resolution", Biopoly mers 22 (1983) 293.
Richard Dawkins
Blind Watchmaker Biomorphs

These patterns, called "biomorphs", are all generated by the same recur sive tree
algorithm, familiar from computer science textbooks and most easily understood with
reference to the simple tree in the middle of the figure:

procedure Tree(i, y, length, dir: integer; dx, dy: array [0..7] of integer);
{Tree is called with the arrays dx and dy specifying the form of the tree, and the
starting value of length. Thereafter, tree calls itself recursively
with a progressively decreasing value of length}; var
xnew,ynew: integer;
begin if dir < 0 then dir: = dir + 8; if dir >=8 then dir:=dir — 8; xnew:=i +
length * dx[dir]\ ynew:=y + length * dy[dir]\
MoveTo(x, y); LineTo(xnew, ynew);
if length > 0 then {now follow the two recursive calls, drawing to left and
right respectively}
begin
tree(xnew, ynew, length — 1, dir — 1) {this initiates a series of inner calls}
tree(xnew, ynew, length — 1, dir + 1)
end
end {tree};

The only difference between the biomorphs (with some additions noted below) is in the
quantitative parameters fed into the procedure as the arrays dx and dy. The program was
written as a demonstration of the power of Darwinian evolution by artificial selection. The
basic tree algorithm constitutes the "embryology" of the organisms. The quantitative
paramters are thought of as "genes", passed from "parent" to "child" in a sexual reproduction.
In every generation, a parent biomorph is displayed in the centre of the screen, surrounded by
a litter of its own offspring which may, with some random probability, have mutated. A
human then chooses which one to breed from. It glides to the center of the screen and
"spawns" a new generation of mutant progeny. The process continues until, after a few dozen
generations of this selective breeding, a radically different shape has evolved to the taste of
the chooser.

Many of the biomorphs in the figure were generated by a slightly extended version of the
program [1], described in Appendix [2] to the American edition
9
10 R. Dawkins

of my book, The Blind Watchmaker. As well as the nine genes of the original
version, the later version has additional genes controlling "segmentation" and
symmetry in various planes [3].

"When I wrote the program, I never thought that it would evolve

anything more than a variety of treelike shapes. I had hoped for
weeping willows, cedars of Lebanon, Lombardy poplars, seaweeds,
perhaps deer antlers. Nothing in my biologist's intuition, nothing in my
20 years' experience of programming computers, and noth ing in my
wildest dreams, prepared me for what actually emerged on the screen. I
can't remember exactly when in the sequence it first began to dawn on
me that an evolved resemblance to some thing like an insect was
possible. With a wild surmise, I began to breed, generation after
generation, from whichever child looked most like an insect. My
incredulity grew in parallel with the evolving resemblance ... I still
cannot conceal from you my feeling of exul tation as I first watched
these exquisite creatures emerging before my eyes. I distinctly heard
the triumphal opening chords of "Also sprach Zarathustra" (the "2001"
theme) in my mind. I couldn't eat, and that night "my" insects swarmed
behind my eyelids as I tried to sleep ... There are computer games on
the market in which the player has the illusion that he is wandering
about in an underground labyrinth, which has a definite if complex
geography and in which he encounters dragons, minotaurs or other
mythic adversaries. In these games the monsters are rather few in num
ber. They are all designed by a human programmer, and so is the
geography of the labyrinth. In the evolution game, whether the
computer version or the real thing, the player (or observer) obtains the
same feeling of wandering metaphorically through a labyrinth of
branching passages, but the number of possible pathways is all but
infinite, and the monsters that one encounters are undesigned and
unpredictable. On my wanderings through the backwaters of Biomorph
Land, I have encountered fairy shrimps, Aztec temples, Gothic church
windows, aboriginal drawings of kangaroos, and, on one memorable
but unrecapturable occasion, a passable caricature of the Wykeham
Professor of Logic" (from The Blind Watchmaker, pp. 5960).
Blind Watchmaker Biomorphs 11

References
1. This extended program, called The Blind Watchmaker is available for the Apple
Macintosh computer, from W W Norton and Co., 500 Fifth Avenue, New York 10110,
USA. The unextended version is available for the IBM PC from the same address. In the
UK, both versions can be obtained from Software Production Associates, P.O. Box 59,
Leamington Spa CV31 3QA, UK.
2. R. Dawkins, The Blind Watchmaker (W W Norton 1986).
3. R. Dawkins, "The evolution of evolvability", Artificial Life, ed. C. Langton (Addison
Wesley 1989) pp. 201220.

&

Figure 1. Twentyone "biomorphs", distant "cousins" of one another, generated by the

program "Blind Watchmaker".
Phil Brodatz
Apple Tree Pattern

Wood patterns are infinitely varied, offering myriad opportunities for novel
designs. Shown here is a photograph of a crosssection of an apple tree burl. Infrared
film with A25 filter is used to eliminate the gray tones and to accent the harder
wood lines.

Reference
1. P. Brodatz, Wood and Wood Grains (Dover, 1971).
12
 13

WW73 Cross section of apple tree burl.

Infrared film with A25 filter.
Dawn Friedman
Spring Tiles the Planes of Sunlight:
Decussate and Tricussate Phyllotaxy in
New Growth

Plant forms are characterized by arrangements of leaves around a stem:

alternating, whorled, or spiraling. The various arrangements, or phyllotaxies, form
patterns with different degrees of visual symmetry. Though these patterns may not
be explicitly considered by gardeners and artists, a plant derives much of its beauty
and its visual "feel" — formal or free, ornate or simple — from its phyllotaxy.
When the phyllotaxy along a single axis is duplicated over and over, spectacular
patterns can be formed.
The plants in these figures are members of a ubiquitous garden group: mound-
forming perennials. As perennials, each spring, they grow back from their roots to
fill an everlarger share of sunlit space. As mounding plants, rather than growing a
main stem and reaching for the sun with that single "limb", they make dozens or
hundreds of small stems radiating upward in all directions. Each stem is an
independent phyllotactic axis, unfurling leaves in its set pattern until they overlap
the leaves of its neighboring stems. The shoots grow taller, the plant larger, but the
surface of the mound is always a hemisphere crammed with leafy phyllotaxies.
Whether the sun shines from the east or west, there is a leaf to intercept the light.
Every possible plane of incident sunlight is tiled with leaves.

Figure 1(a) shows a catmint (Nepeta species) with the characteristic phyl lotaxy
of the mint family: decussate, or whorls of two. Each stem produces pairs of leaves
on opposite sides of the stem, and each new pair is staggered, ro tated 90 degrees
relative to the one before. In Fig. 1(b), another member of the mint family, Veronica
Latifolia, tiles a mound with its decussate stems. The multiplication of right angles
in this phyllotaxy, emphasized by long, arrowlike leaves, seems to contradict the
rounded form of the plant as a whole.
Figure 2(a) shows the tricussate phyllotaxy of Sedum Seiboldii. The leaves may
seem to spiral at first glance, but in fact they appear in whorls of three, staggered
with respect to their predecessors by a 60degree rotation. It is the propeller tilt of the
whorl members that gives a spiral effect. Figure 2(b) shows
14
 Spring Tiles the Planes of Sunlight ... 15

the entire sedum plant, a compact mound of more than one hundred tricussate stems.
These round whorls of three rounded leaves create a hemisphere of circles. The alert
reader will discover a stem with fourleafed whorls near the center of the plant. Like
the fourleafed clover, there is always an exception to the rule.

All four photographs are by the author, taken at Princeton University and at the
Mary Flagler Cary Arboretum, Millbrook, New York.
16

(a)

(b)
Figure 1.
 17

(a)

(b)
Figure 2.
Peter Desain and Henkjan Honing
Trajectories of a Neural Network
Quantizer in Rhythm Space

Described here is a pattern showing different trajectories in a space of all

possible rhythms of four notes. This space is traversed by a quantization system that
seeks a metrical interpretation of a performed rhythm. The durations of the notes are
adjusted to an equilibrium state in which many of the adjacent induced time
intervals have a small integer ratio. The system is implemented in a connectionist,
distributed way. A network of cells — each with very simple individual behavior —
is used, in which the cells that represent adjacent intervals interact with each other.

The three degrees of freedom are mapped to two dimensions by normalizing the
total length of the rhythm. Each point (x, y) represents a rhythm of three interonset
invervals x : y : 1 — x — y in a net of interacting cells. Plotting the trajectories of
different rhythms exhibits the behavior of the network and the stable attractor points
in this twodimensional space. They are positioned on straight lines that represent
rhythms with an integer ratio of two durations or their sums (x = y, x + y = z, 2x = y,
etc.). A graphical front end to the system was used to produce the figures.

Figure 1 shows the space in which the system is given an initial state in an
interactive way (clicking with a mouse at a certain point in the rhythm space),
making it possible to explore the space, evaluate the performance of the quantizer,
search for maximally ambiguous rhythms, etc.
Figure 2 shows an automated run through a large systematic set of possible
rhythms. One can see relatively large areas around the simple rhythms and
relatively small areas around more complex rhythms.
The system differs from other methods of the quantization of musical time in
that it combines the following three characteristics: it is context sensitive, has no
musical knowledge and exhibits graceful degradation. It was first de scribed in [1]
together with a micro version of the connectionist quantizer coded in CommonLISP.
Methods for studying its behavior appeared in [2]. Nonconnectionist methods for
quantization of musical rhythms are described in [3] using a system with tempo-
tracking, in [4] using an expert system, and in [5] using a divisive search method
with backtracking. Clark [6] gives an
18
 Trajectories of a Neural Network Quantizer in Rhythm Space 19

idea of all the processes that are involved in adding expressive information to a musical score
in performance.

Note that the resemblance to actual microscopic pictures of neurons is purely accidental.

R e f e r e n c es
1. P. Desain and H. Honing, "Quantization of musical time: A connectionist approach",
Computer Music Journal 13, 3 (1989) 5666.
2. P. Desain, H. Honing and K. de Rijk, "A connectionist quantizer", Proc. 1989 Int.
Computer Music Conf., San Francisco, California (Computer Music Association, 1989)
pp. 8085.
3. R. B. Dannenberg and B. MontReynaud, "An online algorithm for realtime ac
companiment", Proc. 1987 Int. Computer Music Conf., San Francisco, California
(Computer Music Association, 1987) pp. 241248.
4. J. Chowning et a/., "Intelligent systems for the analysis of digitized acoustical
signals", CCRMA Report No. STANM15, Stanford, California (1984).
5. H. C. LonguetHiggins, Mental Processes (MIT Press, 1987).
6. E. Clarke "Levels of structure in the organization of musical time", Contemporary Music
Review 2 (1987) 212238.
20
Figure 1. Exploring the quantization of rhythm space interactively.
21
Figure 2. Systematic behavior mapping of the Quantizer in rhythm space.
Jacques Boivin
An Internally Mechanistic Nucleon

Represented here is the first visualization of a nucleon structure developed

according to the principles of the Heart Single Field Theory. In this illustration,
each small circle represents a spherical space containing an electronlike compo nent
closely bound to its neighbors, their number totalling 1836 to form a pro ton. These
components are "electronlike" only in terms of that aspect of their structure which
is responsible for their mass; that aspect of their structure, normally responsible for
"charge", has been altered into an intranucleonic weave. This model was meant as
an example of a type of theoretical construct that could be sought within the context
of a coherent allencompassing theory on the nature of the universe. Uncertainty,
simultaneity and dualities (such as "wave/particle") may not be considered as
absolute features of elementary reality for the purpose of such models; rather, the
behavior of precise funda mental structures must give rise to apparent uncertainty,
instantaneity and duality.

Briefly stated, the Heart Single Field Theory holds that all events in the universe
derive from the structural configurations of aggregates of Hearts, a Heart being the
sole building block of the universe and existing in only one state, a basic self -
consistent field. Great numbers of these identical elementary units join end to end to
form strings that organize themselves along a natural progression of complexity.
Dimensional configuration determines the proper ties of each particular entity; a
string of Hearts with both ends unattached becomes a traveling helix (a "photon")
while loops of Heartstrings form into a hierarchy of shapes, such as toroidal coils,
which combine in various ways to generate all existent particles and fields.
This "neoclassical" vision conceives of reality as ultimately amenable to
conscious understanding without resorting to counterintuitive or compart mentalized
abstractions. Quarks, gluons, gravitons and their ilk are postu lated to exist in only
one area of the universe: the cranial cookie canisters of twentiethcentury physicists.
Currently, the only version of the Heart Single Field Theory in print is the schematic
early version from which this "nucleon structure" is taken [1]; a much more
elaborate version has been promised for many years but a number of mundane
preoccupations have been holding it up [2].
22
 An Internally Mechanistic Nucleon 23

R e f e r e n c es
1. J. Boivin, The Heart Single Field Theory: Some Speculations on the Essential Unity of the Universe
(selfpublished). Reprinted in Speculations in Science and Technology 3, 2 (1980) 185204.

2. J. Boivin, Hearts Not Quarks (Tentative title) (to appear).

Figure 1. Nucleon structure, o = electron (total: 1836). • = center of nucleon (location of stabilizing
particle in neutron).
Phil Brodatz
Wood Pattern

Wood patterns are infinitely varied, offering myriad opportunities for novel
designs. Shown here are photomicrographs of cross sections for various woods
including Urnday, Orey wood, and Paldo or Guinea wood.

Reference
1. P. Brodatz, Wood and Wood Grains (Dover, 1971).
24
25
Top to bottom: Ceiba, Balsa, and Freijo.
Hans Giger
Moire Pattern

The pattern described belongs to the Moire patterns first investigated by Lord
Raleigh in 1874.
The Moire phenomenon is based on purely geometrical principles in as much as
the image processing in the eye of the observer need not be taken into account: Two
textures with black and white components are geometrically superposed by forming
the union or the intersection of the black pointsets.
If, for instance, each of the textures L\ and L2 are the black contours and its
white interspaces of a geographical map, the superposition of these maps as films
form a Moire L\ U L2 with its typical Moire fringes.
Harthong and the author independently and with different methods have proven
a generalized form of the following theorem: If the two contour maps L\ and L2 of
two landscapes are superposed, the Moire fringes of the result ing Moire L\ U L2
can be interpreted as the contour map of the landscape determined by the
differences of the heights of each pair of points with the same projection point on
the map, the first point laying on L\ and the second on Li

From this interpretation results the following corollary: The Moire fringes of the
Moire L\ U L2 with L = L\ = L2, i.e., of the selfsuperposition of a contour map, in
the case of a "small",

1) translation is the contour map of the directional,

2) rotation relative to a given turning point is the contour map of the rotational,

3) radial stretching relative to a given fixed point is the contour map of the radial
derivative of the function describing the landscape.

The three effects are demonstrated with the map of the landscape given by
cylindrical coordinates z(r, ip) = r ■ tp, 0 < r, 0 < <p < 2n. (Figs. 1, 2, 3).
Moire 1) direction ip = 0, resembles the sourcefield of an electrical charge; Moire
2) turning point r = 0, is the contour map of the surface of a circular
cone;
Moire 3) fixed point r = 0, reproduces the contour map of the given landscape.
26
Moire" Pattern 27

References
1. J. Harthong, "Le Moire", Advances in Applied Mathematics 2 (1981) 2475.
2. H. Giger, "Moires", Comp. & Maths, with AppL, 12B, 1 & 2 (1986) 329361.
Dawn Friedman
Fractal Phyllotactic Spirals:
Broccoli "Minaret"

Fractals have the property of producing complexity from simple, iterated rules.
In this variety of broccoli, the rule of selfsimilarity operates on a spiral pattern to
create three levels of nested spiraling florets. The effect is as elab orate and fanciful
as Moorish architecture, giving the plant its unbroccolilike name of Minaret. But
the fractal generation of rich complexity from simple rules appears everywhere in
nature. What is most remarkable about Broc coli Minaret is that the fundamental
spiral pattern on which the fractal rules operate is itself the product of a few simple
rules.
Each variety of plant produces its leaves, flowers, and shoots in character istic
patterns called phyllotaxies (leaf arrangements). Shoots may appear in opposite
pairs along the stem, or in a zigzag ladder pattern; they may form whorls of three,
four, or more; or they may climb along the stem in a spi ral helix. To explain these
patterns, it has long been assumed that existing shoots exert a repulsive influence,
ensuring that new shoots will be placed at a distance from old ones. The
biochemical or biophysical mechanism of this influence is still unknown. But
recently it has been shown that a single set of equations describing the behavior of
the repulsive effect can generate each of the phyllotactic patterns seen in plants,
from the alternating leaves of a leek to the tight spiral of broccoli florets seen here.
The strength and range of the repulsive forces in a particular plant variety determine
the precise pattern which will be seen.

Mathematical rules, therefore, govern both the basic spiral in Broccoli Minaret
and its elaboration in three nested levels. A computer program can generate a
perfect replica of the pattern in this photograph — even while the biochemistry
behind the living pattern remains a mystery.
■2S
Broccoli Minaret is reproduced here by written permission ofzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJohnny'sSelectedSeeds.
to
Photograph: Johhny's Selected Seeds. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Ulrich Melcher
Puppy Representation of
DNA Nucleotide Sequences

Described here is a pattern showing a useful representation of the sequence of

nucleotides in the DNA of a plant virus. The DNA is the genetic material of the
virus and encodes all functions necessary for the virus to reproduce itself in host
plants. A DNA strand is a linear polymer of four nucleotides: adenosine, guanosine,
thymidine, and cytidine monophosphates. The information content of DNA lies in
the order of the nucleotides. To depict the order, the upper case Roman character for
the first letter of the English name of a nucleotide usually represents that nucleotide.
Such representations are difficult to scan for inter esting features, inefficient in the
amount of space they occupy and in some type fonts lead to mistaken reading of "G"
and "C". In this pattern [1], the top line of the sequence row has circles for
nucleotides containing purine bases ("A" and "G"), while circles on the bottom line
of the sequence row indicate pyrimidine nucleotides ("T" and "C"). The middle line
of the row contains an additional circle for "G" and "C" residues. Most DNA
molecules, including that of this virus, contain two antiparallel DNA strands. The
sequence of one strand is complementary to that of the other, according to the rules
that "A" pairs with "T" and "C" with "G". The Puppy representation shows the
sequence of both strands; reading the diagram upside down gives the
complementary strand. This feature also means that small selfcomplementary
sequences (often tar gets for enzyme recognition), such as the "stairstep" sequence
TCGA, have recognizable symmetric appearances. Single circles represent
nucleotides that form weaker base pairs in doublestranded DNA, while pairs of
circles show the stronger base pair formers. Thus, circle density of a region of
sequence is an indicator of the thermal stability of that region. The sequence shown
in the figure is that of the Cabbage S isolate of cauliflower mosaic virus [2] and is
8,024 nucleotides in length. Visual inspection of the representation allows
identification of many sequence features. A particularly striking feature is the comb-
like region of repeated alternating "C" and "T" residues. Visual inspec tion led to the
identification of a region that, in the RNA transcript of this DNA, is probably
internally basepaired.
30
 Puppy Representation of DNA Nucleotide Sequenceszyxwvutsrq31

References
1. U. Melcher "A readable and space-efficient DNA sequence representation: Appli-cation to
caulimoviral DNAs", Comput. Appl. Biosci. 4 (1988) 93-96.
2. A. Pranck et al., "Nucleotide sequence of cauliflower mosaic virus", Cell 21 (1980) 285-294.
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

, , ,
, ,
Tr 1 , , , r, rr, r, r
,1 B,, ,,, ,,r , v r , ,
,, , , , , r ,,
Va'"ii^^\T.""J__*."^ n""Li"""aD~~au™«.'.~~B.""iuB'»~ "»~/i": i~* _^i i."v"iju."~,*.~«•.*j ^_i"jj^"a~rji\."»JBjYoi\"u'ma V. o^B/i^i«^__VVna.. ^""j^ji Y"i"ix~Yj _V.V_i i B"LjiiL Lij"uu
t*ij»j».t"»j»j»*i**.'ir.o.B.*i*[Link].
tijiji.'[Link].
li j Tut.^nvn.
IjTi ... —I ..—. T I*xi_M.ITI'M .B . —
. . — !i T II *xi__ [Link]'M.■"■""'tiirri'....*...'*PI*MV"■M■.tirra*I'*»'.—Ml. . . PI 1* . . * * * *.' " . * * * . *.

A T * _ _ p . V A . * _ _ * r _ i . B * _ 1 _ ut P _ . . p J _ l P _ _ .*__,—BI _ _.UI i i.*_—.*«. |i_*__ i_V %»_1t1< P*P_

— . •J i. to.-*— V r r
a M V^V^-i":
ji,tT*jj rr ■AA *__j|Tjirl. B' J J ll rpJJ r JJ ll rJJ__ . * JJJ. —.— BI. .B BU" i' i» '■*——.' . |' — * • ■JJ*. ' —M V "i
,r /
iu x"! ■j"_ .' j
, ,,r, , , /
i.Y
//
.i ^. .u"j*i.* *
, P,rTTrr
..""..""
T
*I _ JI^'
— B ,, ^J''—rr BVji%" i i'«__""V jj^'..hh__ ^^u"___ ""..
,, rr , ,,,rr r

r r
.

•I J./'uVL /L**IiI***JJ4X"V"'^i^*V"^ _ \ * ".i"..**YiiY".Y._YVi^I.. *. .

"^.^Y.Y**.JBuYJ*B ^~ J V.t"' * ^ ^ ,

r , ' A V ^ —
■"JI**" J""""*Ji—".H" " V ^^JI"j" |j. t','.^ "J• J jl*. •«■\. Aft • . "Jj . vV:.YftI ^'..*'^\.U^ •^.VV*[Link] .J^ .. •%^_. •7*U^ u ^ "t t"
.^v.
B .
._ j. . * i j _—.Vj—.:<~.—>
w
B ••—v. .^/ "j i.^N^*'" v. .[Link]::B^ . .VY»^uVB^ _^—J.:<~.-*

— ^ r , TW ,
r , ,
,r r
^r r TT T
•^. ■..^\^^\■*•aVB\.^aaB..V\.*^x• ;_^.^./^A.\. ^y.Y^u^ju*-
^.jV*J..•YJ•.. v_..\ |•^^J^^^ . V l _ • ^ • ^ l ' . • _ . ■ ^ ^ . ■ * ^ ^
.

T*tnlrvrPvV..——.[Link]......P*—*—....BaB...........M..... —.***B*a...*.1iI1ii. .
* . . * . Ip—Pp* l■>•***b. ft*ai*BlliwtTT*rf.y.*. *T*pr. —Ii ~— — tTr—.. Ii BTTarin* •*• tP*p*v.* •MB*jn*r...*i*x.*.*iI *'• ——*Tr *•■. *■ *■•T*rT..

***—*■»—M* Il «. . **—rv*ly1"*—«**—i ***"!1*1******P*| * I|* **1Il—»M .—""Bt**!*BP*|*ap*|*fr*"*y^—* — —** ■* ■ ——**i**f**|**fb|f i**|— PP1*1*IT iI *a*n*BT*I*B**l**l»» *^ JP* * IB**l*TP ^—^**■■. ■*■**■*■ JP«B***I»**!■***P»**U***I*.—T— JPII M***— ■.■******»J———Mml.»i..11
m r ,, m r rm , ,, r , , T, — r
ViA*"*V — ^ . * ,w**m,, , mr *Vi r _ ^~i . '"""" . ""*JBrn __ '*"rnI I _^""iV" .n""T*—,r—.,** Y\J" _ *r— '*~ —r1 . ^rr, T ~B'r,TT,, T B _ ~I . ,' Uu", , '.""Hi . "»"J*\."I XV
VlA*"*[Link]'n8B_'—^^l.'^*.. .BBtB."'"J""m.r'.'[Link]"~^^l'"'—il'.,B'.^"^—*^^.~—J."JB".^j•_BJ**"«U""|*U*«"Bu'^■uuu'i■ru'^.u'■".•""^J.b.■.*.*/«*'h~"l"j**JX^iHIaJ».'>.'J_"'.. .""'',*.^^.**B'^B"BUu'»Bl..""*lBi^.."**.r»i.'^'•^..""•*%""•"j'*"■«%""•"x*—rit^^^Xtjr"._"**I"ix"XlJ_r"j**~

,
■^».*^.V..V.'^—A—.AB.. . 'i—P—"^i—"^iPIju^r.—^*.—U1.^*ri,a*/B.r—i,a*/f.^ .V^r^
,

•" » . "i.—."i. j. j ' ,* —■ •jj. .r.—ji*.r ^A x.* * *^Jii—Jt*.* » ^*> B'_"i*h—j*. .**i_ n >u*j

1 B 1 B. _ . *1 J.I _ .P .* . \ ^ j . ^ ^ a B — j . r ^ ^ B ' _ « — x . ^ * u " a « u .

p i t l i_ **i M Hit u u i | . .w
*—*Pl* * Mw l I* —.*. I IiI** *!***p*l*>*%.._.... *B*.*B*——*M#**W_ II _—■!»**■—***!——■■—TV*—_ BB—***B—_ —.**. UU ** | | *MM—*—M** *JPJPI*JPI UU II III *BB*B***MM .a. B——MMB***w..—1i***!.. ——M"* BMIUBfcw...**1—i***!..**w*Vttv.*ta
II
P T V . H f T 1* B VU PV » M .
* BMP r*.. *[Link] M«* *"B*B. . .B —.'"**... . nPrB TPTf lI111.|. .
*r1*1i .** xX .. * rP 1 r* rP ra TPnI — PJBPrjar—* * rP ■**..... '*_ n1 Pr U*B I.P* *B • * rP. . ' XITTT "
.. ■* *
,
,, m m m m 1

uiJfl|B^r*i*,!>—™'™'"»*—PiPT—.T—y**MBM|'—MBM|»"*I"","^™I'I'P'*VJ"''|**»*11**.1|*—l*B*I***«,"8*P**I*Jfi||P*|—.T*1|PJ*B*PIP**l|i**i******i|i*.."BT*****B*PT*M—i—'VP*i** tj^^|H|"iXI.—xyttj^ll|tx*tv^^^ww—t^X^B^XVa***".lB*lxyB*B'*»ii'i'—'i*****'.—U'l'^^TPtj^'i*vvvjijTWfg'jll————**T*I**I*||PT—■vwjujTWfg'jMaa'v^''jrvrII**B*——**■a'v^''jrvr■*li**nv^Vfgjfrxrffl*.—.**111111nv^Vfgjfrxrffl111''tI'M—IIII'fimj^''t'fumuj^''—BB **#. ■—"**

r^^
" Td * _B *fr,* ^*" f ljlj.'.
uI l_B BB T"B'B*VB.B*fp,*rI—^ ~^*"I—.,fM! \. '*tB"."».*AUuV
..**u ,*,* pp |' Vrl<BM<**VBa\™I*BB"
**B **B "jPT*"
jin UL
I^J".i^A^L"jrx*ia^^*f
"jrx fc a^^*r *^
~x
p.*ij p
. J
B
. *^^
V I , I , ,, ,

,
TTT TT , , , r r r
.^**.__r^*»r..Y*l*^..Y. .«.«^BhVB^iVV".'^BhVB"iV»*aaYyu*.*^i*"^i''.V... '"...'*x\ p \-
\.»V*A*iV™AYBV^».^.^A.<i*»|. *uBui^ri^ri*^> H^".^"^** i ^
M rr p r ,, ,, ,,pp ,W, r ,M, ,, mm W nn rr , , a,l m8 r ,l n ,n,m , p , r pr p , r rT r
JS~^~t\K\^t^L'V«"iV*"l*i*i*B*^l"^li*^^iiilW*^i*i*i|V■i*i***..prr—Tf
..*."i *-i*Ja''T'ii*.'.'J^u\'utBi —*m"".S~'Jm'J'Ji|t^i*ni*il*"* .Yj"'B'Vp..*i*'•..ilWt*il\i^^**yii*BB,../j"*", j—*a'j'Ja• '¥j|*l%iii**H*i****li**1'.'"^lJY~^J'i'tt^.•J■'.YYt.t—
t B
,

B
,

—^\'^\^"•Y'.^**lt*■~BV*Yt*iTi*f*'^"^uV*'tu» t ..iii*^~'•_—-JA^AiW"T*T^.^i'~'^i*I*****.."*j'~j"**T■"'*"*"*'iit.'iiVj..~~*"~*.±T..*.Vtt".ru-'M~MLj.i.~B"

•t,*i^al1Plap i P_T, ,j,*TilTJ l .Vj»^HjxT«ii | . i.. \\ « r AA y^rii'' iiY "" .. V|i ^T V . Y . ^.■. ^^ i^. YY B. •"'".' Pt I_ P .*B*— ./ v... i n i '■■"i ul,, * zyxw
1
V r1
l P a _rTPTjj
n j pr r - ^ - ^ . l j [Link] r. r
J J.. x T « | J . < .. \ « r *TT,*J *1 **Tn/yty1 1*T ^V
- ^ ^j,*r*/fi- rB*nA*r f j f f ^JS 1f| rrv:,-! — . ■*..»ii<
■ .' .

*||■,t»■.*. ■' ^—
" Il rj
'* I
*UUJ iM PJ. , B —/.■. , *I .
'■' ii
I '■!!„•Ulu* m,s

T ,
—,
"h^•■~*I .^""«*mI*m""**r*"TJiB'—B^,^^MP^H*TJ..*B**\«AVavHV4v.,^I
—— 11I—r r TTr ,rrr , , , , ,
, rrr ,,, n, n
^^V*^"t^p^Vi*a"r%B BAA1l^.^. Bjk"*arM/i"*jP,*iaTx«u^^'tri^B u
Y4^
*^*~ .*\.' ^*
j / T ir it.p*[Link].*trj/ *j.* P TPT,T T,j TjT|r..r
.*.*U .. V . ^^
.p..r jjy^
_ ^jr .Yj 'M 'i'r
l i
.*^**ar
Tmi»inl* T

M^*,*i B l fc u 11 *T,TiT.p pp*^j*B u

I U I IBX*l* ^***'r. *tl.* J*X*ll.r'*T U*^*j'TT '^*".>''.* l'j'B*'B**< t AYj''**''' r
IMII BJ.*l ' '' ^ . *J*X*l l. IJ *J l.'^^.''. "l'j'B* B" l *AYj''**'''''u.*l^**'' .Y
i• Jj7B T*pyry__mm_aAi^rBB_**.,Bj~/^T.*.Tt*i.*r_r1,*j*uB *j*y^B ^, n *H"^i\' r
u r
j\'*^ ^
iJ */ B *rT*,* _ mj**A*i ^BV . "a~ . 'H' . "uVB"j"H
iJ */B *rT*,*_mJ A ^TB _*,Bj~.*^T.B ^ yB*j^*n^* p 1 * . T l_ p y . * i B B X P / j * ^
r n n ,
, nl r l ,r 1 m 1 1 T1 ,, , r rT r,—— r
i™~ij"rB'
V•B*jV,*iViYi*i._Y'*i '.B*fr'.*."*n__iY*^i»*jj*,*T*j','nu* 1nn^^j*/^1u*MiYjVV*xn^V^"B^\'t "BB**ri^'p't.i.i"'^'t*^n,*[Link]**1 B*—rnpiJ—_•S'j .r*j"~Vi,i^"!"*~*u"i"B'VV.j™ VB ^. V'i"V . xY"i"Vfc . ".' ji'u'*~V*" *" V

V iYi .Y r . 'n " a j Y ^ " i x . * a u ,

iY V^"B^\ , p n . ^ ' B r * lr — i "j". 1
r,— r ,
T, B ,, . r i " — « . Y " i " r .T
■'L'UI'n ,

• ••I"*!*""**. ■ * •*■ •••»•

,•t ,
r ,
If"*! * I ■**'••*1— —* T
• t*™" — '!"■ ••
,
•**•**n,r,rB",—r
T 1*T, — **•**!*■TT
,
■**■!*r—,l•*'l."**l.— n
I "!■ . "l •""*•, " .""B a"™—.* rnl™l * *l—■*

_ i _ j * n " r tur_n jp'i r — • lx _ *j*u"l~ *x^"*.i*.' " ' "/^*l.*' .u'j'*~".Vj'T\^J. . Jl. ".l'*r' '. u.V"*~4.'. .\'j. ' j " \ i Va~B u'Vm"VVjr^"*i"j" *-~m'."B. \ .'.A''m'V

, _ , i uu u
_ u _*jij i i— ll xx _ *j*j * _ 11 ll —pP*A.i,— x ^ " * . ''* " •.'.'j' " j i ' « « " "i'' u ^ " . . h i.'j" u" Bi" i ••.•j"*
\ , ' \ i *Ti * ^ .r_ * Y V—" j "r ^ _ V , r , l r r r 11nl l , , p 1 , Br . " ^ j \ * ^ \ " ",
l

B,* Y . ^ ,

\• »jrjtrrV -"r, -.t—ji—t"-«-i au~rB.'—t "JIi -.jl-iBmi..—" ~"^V*j' .aVa'"[Link]%*.~ Vnt™WiW»^HA»'%j..VL^^\n^VA V V .".*iY".jBY.». j V i " \ . i" iV.'" ' .

a ja • . *,*. . • J * B — I T . . . x.n .*a* . a . 1 . I. T . 1 . T J*.* .•■• i . * . a ■•. *x* a 1 B m Jfc —vfc * . * * »—r.v B J I .

a*xi *. *,*. . * XB—r*pi.. . . . x . i iI .*B** . B.I . I. *rp .p* . r . jr.* .*.* i .y*.«B ..**.. *x* BIIBm
, * ^ J . I j * P « ^ l _ V _ * . l j P X , * _>j l / j | . , X T T y j i*
jj .. —v%in. .** * • —p.v tpxi | jV.*l, ,P^*> 1 j^

,
r r , ,
r ,
,p lr r ,,

i\i*__*'^ ..*i..ij*p.j*""W«.i..V¥ .V 'ii»Yi'. i»x..""j*VYV"*^^."ri^^y. *jV . ^" . uY~•A"jV . «^ %B i'^^^Y"i i" ^^ . ~i B . »a ". . * "r
v ,
,v r , , 1 1 T,,T ,,
i\"»aj"i\"%j"'*YiiY4.\***^iYj"n**Vi\rr*iBB**VBtr"«..r"ro«B.^i"*"YW*^i'^"ij' ^i^Yi..VA Btu*Y.^^
1
I*l|i*t»!■——**BIUar*r*1*1I—* .|Iy!■*.—i—IPp M■ M'■' *iV.V*'*/M—^ *M''iI UIIIml MI. P .* XPJTM«JPfc*Ufc*..*[Link]..%Ban*l...BTPUU'l II ——I*.I**.—*.——.*—..X*B—V*—..—** *.**!..PIn — VPI**PT.M.*AI.** B*I—l —"Ii. *t.*T***P»vrr*
I* ||—"Q i—r—• | i . r r — —. '. . ' g . *'i u I *. r . ~r—r y I T I T I *. w i 1 I ^ J — — r^" . u . —"— —."t , rt a » ^ T — VI* *^^
Vvrt^••l•ItY"\..*iYi^""l^•V'*rrI*\V.*.".V"'. ^.."V~.~V.v.V."^.■^>«".V.^•..V.■..•.•i*\"Iu./"^'i*l■i•l•I'_l\.■■^^'|i■»l».[Link].^•u•J^jl*•Jj•^j^•^^"a" Y ^ ^

r TT T ,— — ,
, ,
^ r T Tp ,
J

^ rr p rpp ^ ,rr, ,r r , ,
, rr, , m,m nnnvvv,w rrn r r rr r 1, r r
•i ^" li V'*^^^B^BB^rBrBB■'•iii__V.[Link]•iiBiB»BJBj"j"j"" ""l'_"iji"*.".'.■.""»»"""B.a"Ba..""* "~o~jtj/T/^^'...^^j_jamaVVYiii''tt*JJAA'. .H.''JJ. . ''^j~\\V^V——~ViiV.~jjAiY x uui .''~S~aaM'B ~^Ji\V\"'. '^i"'i "\'i''~.'"~tMuuS '
,
r, n ,nr r ,, rr, , r a , r, — rr ,a r
,r r r
, , B
n , ,
l, u
A'*'*~.A'**~H"."\\V»".\\V~.tJ."*JtJ..\'^'~"J...\'^'~l."'^V^\^t"il.x'^V^\^/hJ^~l'^*j^J^~l'^™aJu^V.."i*.•aJLUMV~l'tt'^\^t^.J~l'p^f aJLUMV. ~l'tt' "■' ' h*• Ii "i'*i'iI_.'Y*[Link]..Y*%VJ^*...'iI.. *IVuY*B'aj"t.'t—'Mjj~'J".*~n.^..I.'~'u'^TV"Ajt.B"^*"t*BB..».»."".''~
, rr r, , ,r r pr , r, rT r — lr , , ,
r , ,,, l ,, r — —— ,
1_ ,
,, ,,, , ^, , T 1 1 ,

*•""j•jJ"l"lVl"*~"j.j"tBrmj/^m..H".\'l..V"".j.V'M"jM.j~*~B.*B,"'.^, ^''*.^*Y'l. '."J*~'l'.l'*i'r 'lL.'I' "LL|B"*"%.'\*B.«WI*B*..'^^.*^*X*\I^l*'*VVlUj^lVYjll'YYj"j'' B^"BjB" 4'Bn» ' B"'^ lftcB"*.B" .BV.'VlH»'l*U»lIIJ»JJX'iaJi■j'j•"'~*lX^ X• . ■* J.'"jl' " B u V B " L V l .
rr , , ,

AV.•AVil■*W..v."i*nu\*B«/.'"•»BBVi'V.^•«B"./Y^.V.Y•Wii**nu\W»/.Y*^*...'•••*.«ViU.t.■*««Y.Y"'J._*.Y^VYWVl•^vrl«YW^Y^H Wri«^.s^*.•BB^ Ww.."•".VVfcVl•^•c. V i * ^ ^

^ i j P ^ a , , ^ * , — i — I | * I I, *.I *B . I |^ [Link]*I PP T ^ta w j'^Viju.' i.V 1j" *'i l, "W»\UiiV.".
*•^«I•.VjiT«B^rIBa"B,«*,*1aB."^,fB*BI!"B1,*«^—,»1—lr"HiB*|* tt*lluI*,*A".lf'c.■*B'"■I ."'"ta^l•B|.•| .■P"j*"l BfBc*" T"^ V . ' ^ ' ^

r r r n B rr , r

u i * B ,M— -UV i fUu W ^I u B ^ \ * ^ ' B~B^ i M ~ h m ™ S t \ ' i i^B±'^V± S ~'^f'/ '*. i

•*rp«*i .y r ,
r
,
, ^
I"^»'«culBBB44^"**lj'l**r**..t.**itBtIB. It*T,.TB. B .I'PP'*^'A*i|*^'"«1h'A^^l,l,Bl|IJJj"j^fr*j*jj**TTXfr ;xv
,

"mYP. r
*fB*X^"P^"."."_j^*

•**TT*\l. i"^." TT~^'l^.*i"mY*\.TIV *^ViV^

_ , — ^l r tr r ,M r
Pj*Pn'p j " V* ti"l*v^.\ . B*X« _ i"BiHh_tiV . *i —■ ™.' . *lmVV_m*l * J^^*\."* J^ * ij"^^. ^ ll . |*^ljW . *™ * . V"l^ _ lYl Ll'" . l '
r,

xl _ *l
r

..^« A.*»^»*«_Bnj. .
, ,
r JjJ**in/n". V/11,j,*V BB.n / i# B*Tirj'^n*X j%'a .^Hi1jj'*iY*B'^**,*"BB*^"*^n >*" ,'"pA r B'n j Jj. v'~*x™B", n. m * fjc»p *'"j"^'jf'cB B* .^f P\ i j | J*^* j'"." ^

r r r r ,1 ,1 wTr l,n , d ,
n , , r , fn T w
■tp~ 4 jj,*.B.B^*x,*^ix U |—/j*p/,ajj*rT B ^.pr t ^,*t»,/Jr t^

T rT
" ^ j~ JJ a/ Vi
l ,
.\" ..'^^VWi AA .* Wx w«—~
, ,

.j• "Jr ^ . Y' i" J jJ j 1

"'Bwi^*^ *^ '* ' i T
|j "|'i^*^^ r . ' j ^ ' ^
■PT—jJj ,*.B.B^*X,*A_lx *u *—/J*T/,aJj*rTB B^ _prt^ ,* t»,V
.•BV.*BV.V\.¥V"»*^\»'V *"n .*l* «*W^*m»a_BB"^. ^»B^B ,, , ,
 ,
^ , ,
,

V _« B
,V *p lWl ^V.* r
r TT ll B"BV J^^
.•BV.¥V"»* V "n *i*^
,„,,•*•* BiI .*• .*• BTB*nrTB*M*J..MiI.i .,.* ,i1i .
* *—.— ,
*....*■•■••....*■•■[Link].l1ii.T...*p*[Link].*iB*BI*Bi*B*..**.**[Link],i,[Link].n,v,i..Liy.'.Ti«,.r.'',T^*..v,.V.I'.i.J
, , ,
,
*J._ ii.. iu/,,l.'.i..

, , * . « B I .* .* B T B . J . I V l ....*.—..*. . .
.■•■ ■ I.I.I1.T. .*P. Br _*BI_._*.*1_XJ,l,l.n,V,1 m .. V.I.'_l.*_l. . u.V.
— ^ ^ * — ^ n / _ — | T ,r^ * n ^ * . T ^Ti . B . i / . T , * r T x . . * r i . T H i ^ —.^* . B_.—BVB\—Bji\VV^—JBTjiJ. *aju.^^j»\«*.\ TJ . _**.*'A^.*^B^AB1riY^ir»1BB. B
T T
—^ _B.—B—.B\V—.B .*Jfc 'j*\ J_W^^
t i mtF r F a t ma m l , , , r
BV*...B^, ~SjfJ?*iJ r& ^ *n?fTm ^ ^t*BwJ M * *FJ^'— T.^'.V—\Vu*VLVA."^ ^*'^VBB« Btt.V.*«i."*.* «BfcB*.i'*»B»" * **.**ji

vr>m>/uY..*«%tBBBA«.l—.—— .
Kazunori Miyata
Stone Wall Pattern

Described here is an algorithm for generating stone wall patterns. This

algorithm requires only a few parameters as input data. The output data are a bump
plane, which represents each stone's height data, and an attribute plane, which
represents each stone's attributes. The method is an enhancement of C. I. Yessios'
work [1].
The algorithm models a stone wall's joint pattern by a "node and link" model,
shown in Fig. 1. Each node has a position and link data. The links are restricted to
four directions: upper, lower, right, and left. Each enclosed space of the joint
pattern is equivalent to the space occupied by a stone in the wall.
The generation procedure has the following six steps. Figure 2 shows the flow
chart.

1. The basic joint pattern is generated by using the average size of a stone in the
wall and the variance of its size (Fig. 3).
2. The basic joint pattern is deformed by relocating its nodes. After node
relocation, each line segment is subdivided recursively, using the fractal [2]
method (Fig. 4).
3. The space occupied by each stone is found by using the link information of the
basic joint pattern. The stone space is a polygon formed by nodes and line
segments.
4. The texture of individual stones is generated by subdividing the stone prim itive
recursively [3]. For this, the fractal method and the roughness value of the stone
are used.
5. The stone texture is clipped by cut polygons, which are contracted polygons of
the stone spaces.
6. The height data and the attributes of the clipped stones are placed in the bump
plane and the attribute plane respectively, by the scanline method.

Bump data are used for the shading process, and attribute data are used to
change the color of each stone, its optical features, and so on. An example of a
generated stone wall pattern is shown in Fig. 5. Highly realistic images of walls,
pavements, and steps can be obtained by mapping the generated patterns (Fig. 6).
More details are given in [4].
32
 Stone Wall Patternzyxwvutsrqp33

R e f e r e n c es
1. C. I. Yessios, "Computer drafting of stones, wood, plant and ground materials",
Computers and Graphics 3, 2 (1979) 190-198.
2. B. B. Mandelbrot, Fractals Form, Chance, and Dimension (Freemann, 1977).
3. A. Fournier, D. Fussell and L. Carpenter, "Computer rendering of stochastic mod-
els", Commun. ACM 25, 6 (1982) 371-384.
4. K. Miyata, "A method of generating stone wall patterns", Computer Graphics 24,
4 (1990) 387-394. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFE

Figure 1. Node and link model.

Figure 2. Procedure flow.
34 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

F i g u re 3 .zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONExamplbasijoinpattern.
Figure 4. Example of a final joint pattern.
 35

Figure 5. Example of a generated stone wall pattern.

^c>

n
Figure 6. Edo castle.
Alan Peevers
Novel Representations, Pattern, Sound

Various techniques for music visualization, music transcription, melody storage,

and melody matching have been proposed in the past (see references). A few weeks
ago, I was playing with an audio data acquisition board in con junction with a music
synthesizer — acquiring about 7 seconds each of various voices. After I adjusted
the pitch so that the period was a multiple of 512 samples, I was able to display the
timevarying waveform a s a 5 1 2 x 5 1 2 x 8 greyscale image on my NTSC frame
buffer. Some of the patterns I looked at are really quite beautiful (see figure).
The image is taken from 256K 16bit samples at a 50 kHz sampling rate (a bit
over 5 seconds of sound). The source of the audio is a Yamaha (TM) DX100 FM
synthesizer programmed to modulate the timbre of the sound using a Low
Frequency triangle wave giving an overall "wawa" effect. I adjusted the pitch so
that an integral number of fundamental periods fit into 512 samples, which
correspond to one scanline in the image. So, successive scanlines are just successive
"slices" of the sound, giving an impression of the overall temporal evolution of the
sound. I wrote an assembly language program on the PC that takes the 16bit 2's
complement samples, converts them to offset binary, scales them, and truncates the
values to 8 bits, suitable for displaying on a greyscale display. There is only one
axis, time. Successive scanlines (the y axis) represent successive "slices" of the
sound. Each slice represents some multiple of the fundamental period.

References
1. R. Cogan, New Images of Musical Sound (Harvard University Press, 1984).
2. C. Roades, "Research in music and artificial intelligence", ACM Comput. Surv.
17 (1985) 163190.
3. D. Starr, "Computer chorus", Omni Magazine (May 1984) 41.
4. C. Pickover, "Representation of melody patterns using topographic spectral
distribution functions", Computer Music Journal 10, 3 (1986) 7278.
5. J. Pierce, The Science of Musical Sound (Scientific American Library, 1983).
36
37
Clifford A. Pickover
Haeckel's SeaLilies

Ernst Heinrich Haeckel (18341919) was a German biologist and philoso pher
interested in the beauty of natural forms. Throughout his career he made detailed
drawings of a range of organisms. He seemed particularly interested in deep sea and
microscopic life. Shown here are some of his drawings of various species of sea-
lilies (animals related to starfishes and seaurchins).

Reference
1. E. Haeckel, Art Forms in Nature (Dover, 1974).
38
39
Daniel Platt
Diffusion Limited Aggregation

Described here is a pattern produced by chaotic behavior arising from ag gregating

particles. "Aggregation" is a term used to describe growth arising from the agglomeration of
diffusing particles. In 1981, Witten and Sander de veloped a computer model for aggregation
starting with a single seed particle at the center of a space [1]. Their computer program
introduces a new particle which moves randomly until it approaches another particle and
sticks to it. The first particles attach to the seed, but soon a branched, fractal structure
evolves with a dimension of about 1.7. This process is called diffusion lim ited aggregation
(DLA). Since the introduction of the WittenSander model in 1981, considerable research has
been devoted to the properties of DLA [26]. In the diagram here, the additional contour lines
give extra information to scientists concerning the growth process.

References
1. T. Witten and L. Sander, "Diffusion limited aggregation: A kinetic critical phenomenon",
Phys. Rev. Lett. 47 (1981) 1400.
2. M. Batty "Fractals — geometry between dimensions", New Scientist (April 1985) 3140.

3. E. Stanley and P. Meakin, "Multifractal phenomena in physics and chemistry", Nature

335 (September 1988) 404409.
4. F. Family and D. Landau, eds., Kinetics of Aggregation and Gelation (North
Holland, 1984).
5. H. Stanley and N. Ostrowsky, eds., On Growth and Form (MartinusNijhof, 1985).
6. J. Feder, Fractals (Plenum Press, 1988).
40
41
Mels Sluyser and Erik L. L. Sonnhammer
RNA Structure Based
on Prime Number Sequence

This display represents the secondary structure of an RNA sequence based on

the first 2000 prime numbers.
The relationship between prime numbers has previously been investigated by a
large number of methods [1]. Here, we describe a new approach in which we
designate pairs of primes by the symbols A, G, C or U, depending on the property
of that prime pair as described below. These symbols were chosen because they are
commonly used to designate nucleotide bases in RNA. The computer program
FOLD has been designed to investigate the properties of ribonucleic acid sequences
[2, 3]; here we implement this program to analyze the series of primes.

To calculate the parity of prime numbers they were expressed in the binary
system. If the sum of the coefficients (Sc) is odd, the parity {Pa) of the prime is
1 2
assigned 1 (example: prime 7 = 1 x 2° + 1 x 2 + 1 x 2 ; 5C(7) = 3; Pa(7) = 1). If Sc
1
is even, the parity of the prime is assigned 0 (example: prime 3 = l x 2 ° + l x 2 ;
Sc(3) = 2; Pa{3) = 0). Successive prime numbers (P) were coupled pairwise: [2, 3],
[5, 7], [11, 13], [17, 19], [23, 29]. ... and expressed as parities. This yields Pa(P):
[1, 0], [0, 1], [1, 1], [0, 1], [0, 0] ... . Parity pairs [1, 1], [0, 0], [1, 0] and [0, 1] were
assigned the symbols A, U, C and G, respectively. The first fifty letters of the
sequence are thus:

CGAGUACCACAUCACACGCAAAGAAAAUGUCCCCUUCCGGAAAGGUACGGG

The sequence of nucleotides of an RNA molecule determines its secondary

structure. Computer methods have been used to minimize free energies for the
prediction of the secondary structure of biological RNA molecules. Zuker and
Stiegler [2] implemented a dynamic programming algorithm for this purpose. We
used Zuker's FOLD program (UWGCG) [3] with the free energy parameters
calculated by Freier et al. [4] to investigate the prime parity sequence described
above. The FOLD program calculates a secondary structure which exhibits a base-
paring structure where an energy minimum exists. The figure shows the optimal
folding of the RNA sequence representing the first 1000 prime pairs.
42
RNA Structure Based on Prime Number Sequence 43

R e f e r e n c es

1. P. Ribenboim, The Book of Prime Number Records, 2nd ed. (SpringerVerlag, 1989).

2. M. Zuker and P. Stiegler, "Optimal computer folding of large RNA sequences using thermodynamics
and auxiliary information", Nucl. Acids Res. 9 (1981) 133148.
3. J. Devereux, P. Haeberli and O. Smithies, "A comprehensive set of sequence anal ysis programs for
the VAX", Nucl. Acids Res. 12 (1984) 387395.
4. S. M. Freier et al., "Improved freeenergy parameters for predictions of RNA duplex stability", Proc.
Natl. Acad. Sci. U.S.A. 83 (1986) 93739377.

Duplex display of a pseudo RNA structure obtained from the first 1000 prime pairs. This SQUIGGLES
display was obtained by analysis with the FOLD program. A biological RNA molecule with this
sequence is predicted to have a free energy of —256.9 kcal/mole.
Thomas D. Schneider
Genetic Patterns as shown by
Sequence Logos

The two patterns shown here represent two different views of a control switch in
the genetic material of a virus called T7. Like many other viruses, T7 is made of
proteins that coat and protect its genetic material DNA (de oxyribonucleic acid).
The virus looks like a lunarlander landing on the moon. When a T7 particle contacts
the surface of the bacterium Escherichia coli, it sticks there. Once it has landed, it
injects its DNA into the cell, much like a hypodermic needle does. The DNA
contains instructions for taking over the bacterial cell. The cellular machinery
unwittingly copies these instructions into RNA. The RNA is then used to make
proteins that stop the normal cell mechanisms. Other instructions tell the cell to
make copies of the viral DNA and to make the proteins of the viral coat. These are
assembled to produce perhaps a hundred new T7 particles. The final instruction
causes the cell to burst open, just as in the movie "Alien".

T7's strategy for taking over the cell is to replace the bacteria's RNA making
machine with a new one. This new "RNA polymerase" binds to certain spots on the
viral DNA, and then makes the RNA molecules. Other machinery then translates
these RNAs into the proteins that make up the virus.
The entire genetic material of T7 consists of exactly 39,936 chemical "let ters"
named A, C, G and T. There are 17 places on this string of letters to which the RNA
polymerase binds. Their sequences are:

221111111111 ++++++
1098765432109876543210123456

1 ttattaatacaactcactataaggagag
2 aaatcaatacgactcactatagagggac
3 cggttaatacgactcactataggagaac
4 gaagtaatacgactcagtatagggacaa
5 ctggtaatacgactcactaaaggaggta
6 cgcttaatacgactcactaaaggagaca
44
 Genetic Patterns as shown by Sequence Logos 45

7 gaagtaatacgactcactattagggaag
8 taattaattgaactcactaaagggagac
9 gagacaatccgactcactaaagagagag
10 attctaatacgactcactaaaggagaca
11 aatactattcgactcactataggagata
12 aaattaatacgactcactatagggagat
13 aatttaatacgactcactatagggagac
14 aaattaatacgactcactatagggagac
15 aaattaatacgactcactatagggagaa
16 gaaataatacgactcactatagggagag
17 aaattaatacgactcactatagggagag

The coordinates of each position axe written vertically above the sequences. The
"sequence logo" pattern in Fig. 1 was created from these sequences.
The sequence logo consists of stacks of letters on top of one another. The height of
the stack is the consistency of preservation of the pattern, measured in bits. The
vertical bar is 2 bits high. The heights of individual letters are proportional to the
number of times they appear in the sequences given above. The error bars show the
expected variation of the stack heights. In Fig. 2, they are an underestimate of the
variation.
The area of the sequence logo in Fig. 1, which measures the total con servation
of these genetic patterns, is 35 bits. In contrast, the amount of conservation needed
for the job of finding the promoters in the cellular DNA is only about 17 bits. So
there is twice as much sequence pattern as there should be. To test this, many T7
promoter DNAs were made by chemical synthesis. The logo for the functional ones
is shown in Fig. 2. The area of this logo is 18 ± 2 bits, which is about the amount
needed. So the RNA polymerase does use only 17 to 18 bits of information, and the
extra pattern in the upper logo must have some other unknown function.

References
1. T. D. Schneider, G. D. Stormo, L. Gold and A. Ehrenfeucht, "Information content of
binding sites on nucleotide sequences", J. Mol. Biol. 188 (1986) 415431.
2. T. D. Schneider, "Information and entropy of patterns in genetic switches",
MaximumEntropy and Bayesian Methods in Science and Engineering, Vol. 2, eds. G.
J. Erickson and C. R. Smith (Kluwer Academic Publishers, 1988) pp. 147154.
46zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA ..Schneider

3. T . D . Schneider a n d G. D . S t o r m o , "Excess information a t b a c t e r i o p h a g e T 7 genomic

p r o m o t e r s d e t e c t e d by a r a n d o m cloning t e c h n i q u e " , Nucl. Acids Res. 1 7 (1989)
659 - 674 .
4. T . D . Schneider a n d R. M. S t e p h e n s , "Sequence logos: A new way t o display consensus sequences",
Nucl. Acids Res. 1 8 (1990) 6097 - 6100 . zyxwvutsrqponmlkjihgfe

F i g u re 1. Pattern at T7 RNA polymerase binding sites.

Figure 2. Pattern required by T7 RNA polymerase to function.
Stephen Schiller
Gaussian Fractions

This pattern shows the distribution of fractions of Gaussian integers whose

denominators are limited to a certain size. A Gaussian integer is a com plex number
whose real and imaginary parts are both integers. A fraction of Gaussian integers,
A/B, is simply A divided by B using the rules of complex arithmetic. The size of a
2 2
Gaussian integer referred to above is its complex absolute value, given by \/[x + y )
for the complex number x + ly. In the pattern presented here the size limit was 25.
The individual fractions were plotted by using the real part as an x coordinate and
the imaginary part as a y coordinate. The page corresponds to the complex plane as
follows: the origin of the complex plane is at the center of the large dot in the lower
left corner of the page. The number 1 + iO would be the center of the large dot in
the lower right corner and the number 0 + il would be the large dot in the upper left.
This pattern arose from my desire to know how well fractions of Gaussian
integers, or Gaussian fractions, can approximate arbitrary complex numbers. A
similar, but more familiar problem is that of approximating ordinary real numbers
with fractions of integers or rationals. In this problem, there may be practical limits
on the sizes of the integers in the approximating fraction (such as the size of a
computer word). And so, one might ask, if I can only use integers up to a certain
size, say N, how well can an arbitrary real number between 0.0 and 1.0 be
2
approximated? It turns out that although the average error is about l/N , there are
certain places on the real line where the error can be much worse, as large as 1/(2N),
for example. These areasofbadapproximation occur near fractions with small
denominators. For each fraction of the form a/b there is an interval around a/b of
length at least l/(bN) where no other fractions occur. Thus, there are relatively large
gaps around fractions with small denominators like 0/1, 1/2, 1/3 and 2/3.

The arithmetic of Gaussian integers and Gaussian fractions is very similar to that
of the usual integers and usual fractions and the same comments as above apply
except for the novelty of the space of Gaussian fractions being in two dimensions
instead of one. I was curious to know how the fractions
49
50 S. Schiller

with smaller denominators were distributed relative to the other fractions. To this
end I made the size of the dot representing the fraction a/b to be inversely
proportional to the size of b (1/(2 Size(b)) to be exact). This made the fractions with
smaller denominators stand out. You can see around each of these larger dots a halo
of white showing the dearth of neighbors next to these fractions. The size of the dot
was also chosen so that the dots would just barely touch in areas where
approximation would be good. The resulting pattern gives a good visual indication
of areas of good approximation and areas of bad approximation.

The pattern was generated by a program in the Postscript page description

language (included below). The program is fairly straightforward but some
comments are in order.
The algorithm for plotting Gaussian fractions is brute force. The procedure
/fracmap enumerates all possible denominators whose size is less than the size limit
passed in. The procedure /doDenom enumerates all possible numerators for a given
4
denominator. For a given size, TV, there are on the order of TV such fractions so
the time to run the program goes up quickly in TV. The pattern shown took about
an hour to generate.
Multiple fractions plotted by the program may fall in the same place, but the
one with the smallest denominator makes the largest circle (corresponding to the
reduced fraction at that position) which covers the smaller ones at the same spot.
One could design a more efficient algorithm that only plotted reduced fractions.

I was originally only interested in plotting those fractions with absolute value
less than one. (These would all lie in a disc of radius one centered at the origin.) The
program actually generates and plots fractions outside of that range. This is because
I am taking all pairs of Gaussian integers whose size is less than TV and forming
their fractions and many of those numbers have sizes greater than one. To cut down
on the time it took to plot the pattern, I put in some simple checks in the program to
eliminate most of the fractions outside of the unit disc. For that reason you can see
that the density of the dots falls off as one gets outside of the unit disc. When I first
plotted the pattern, I liked the visual effect of these stray dots, so I left them in the
program.

'/,! gaussian fraction coverage: [Link]

7. Passed on the stack is the maximum abs value of denominators '/, that are
mapped.
Gaussian Fractions 51

/fracmap
{ gsave
10 10 translate
72 8 mul dup scale
0 setlinewidth
1
setlinecap
/maxD exch def
/maxDsqrd maxD dup mul def
0 1 maxD {
/[Link] exch def
0 1 maxD neg {
/[Link] exch def
/denom b. re dup mul b. im dup mul
add def denom maxDsqrd gt {exit} if
denom 0 gt {doDenum} if
} for
} for
grestore
} def

/doDenum { 0
1 maxD {
/[Link] exch def
0 1 maxD neg {/[Link] exch def doFrac {exit} if} for
1 1 maxD {/[Link] exch def doFrac {exit} if} for
} for
} def

/doFrac {
/re a. re b. re mul a. im b. im mul add
def /im [Link] [Link] mul [Link] [Link] mul sub
def /size 0.5 denom sqrt maxD mul div def
/x re denom div
def /y im denom
div def Dot
x 0 It x 1 gt or y 0 It y 1 gt or or '/. return true if out of
range } def

/Dot {x y size 0 360 arc fill} def

25 fracmap
showpage
Ian O. Angell
Lattice Design 1

The lattice pattern here was produced by a computer and demonstrates various
types of repetition and symmetry. To draw a lattice pattern, the com puter takes an
elementary set of line segments and arcs, and manipulates them using a space group
(a sequence of reflections, rotations, and translations) into a tile. These tiles are then
stacked in a regular lattice in twodimensional space, thus initiating further
symmetries. Even random, nonsymmetric starting sets of lines and arcs produce
beautiful symmetrical patterns.

Reference
1. I. Angell, Computer Geometric Art (Dover, 1985).
12
53
54
Henry F. Fliegel and Douglas S. Robertson
Goldbach's Comet

Shown here is the number of prime pairs G(E) which can be found to sum to a
given even number E, plotted as a function of E. (For example, for E — 10, G(E) is
2, since 10 can be expressed by just two prime pairs — as 3 + 7 or as 5 + 5.) This
pattern is connected with the famous Goldbach Conjecture, named after the Russian
mathematician, Christian Goldbach, who speculated in a letter to Leonhard Euler
that every even number greater than 4 can be expressed in at least one way as the
sum of two odd primes. Yes — at least. J. J. Sylvester was apparently the first to
show that "irregularities" should appear in the function G(E) because, if E is
divisible by distinct prime factors pi, p2, p3, ..., then G(E) will be increased by the
multiplier

(p i I ) ( p 2 l ) ( p 3 ! ) ■ ■ ■
(pl 2)(p2 2)(p3 2)

over even numbers which have no such prime factors. That multiplier splits G(E)
into bands, corresponding to values of E divisible by 3, 5, 7, 3 * 5, 3 * 7, etc. What
we call G(E) was analyzed extensively by Hardy and Littleton [1]. Nevertheless, to
this day, there is no strict proof from basic axioms of what appears so clearly in the
graph shown here — that Goldbach was right.

References
1. G. H. Hardy and J. E. Littlewood, "Some problems of 'Partitio Numerorum', III:
On the expression of a number as a sum of primes", Ada Mathematica 44 (1922)
170.
2. H. F. Fliegel and D. S. Robertson, "Goldbach's Comet: The numbers related to
Goldbach's conjecture", J. Recreational Mathematics 21, 1 (1989) 17.
55
 en01 zyxwvu
OJ

UJ zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
m
i
i

in
e
3
a
L
'8
OJ

e
L
a.
4;
0

L
OJ
JD
6
3
Z
Georg Jagoda
JCurve (OLFRAC) (14 Iterations)

Described here is a pattern showing a curve that I call Jcurve. Increasing the
recursion depth by one looks like adding a second Jcurve to the first one, rotated at
an angle of 90°.
The pattern is created with OLFRAC by Ton Hospel from LISTSERV at
BLEKULL11.
Datafile contains:

axiom: F
number of rules: 1
1st rule: F +F—F+
angle: 1 4 \ TT * 1/4 = 45°
recursion depth: 14
57
Georg Jagoda
Carpet (13 Iterations)

Described here is a pattern showing 4 Jcurves.

The pattern is created with OLFRAC by Ton Hospel from LISTSERV at
BLEKUL11.
Datafile contains:

axiom: F++F++F++F
number of rules: 1
1st rule: F +FF+
angle: 14 \ TT * 1/4 = 45°
recursion depth: 13
58
59
 Georg Jagoda
3DCubes

Descr ib ed her e is a pattern showin g 3 cu bes . The three cu bes are the pattern
3D-CUBE. The 180 "+" perform s the rotation of 180 * 7r * 97 0 / 360 0 =
180 * 180 0 * 970 / 3600 = 180 * 48. 50 = 87300 = 90 0 + 24 * 360 0 • The digit s in t
he axi om select the color : 1 (blue), 2 (red ), 4 (gre en).
The pattern is cre ate d with OLFRAC by Ton Hosp el from LISTSERV at
BLEKULll.
Datafile contains:
axiom : +++ + +++ +++++++++ + +++++++++++++ ++++++++++++ \
+ + + + + + + + + + + + + + + + + + ++++++ + + + + + + +++++ + + + + + + + \
++++ ++ + ++++++++ +++++++++++++++ ++++++++ + ++ + \
+++++++++++ +++++++ +++++ ++ ++++++ + ++ + ++ + ++++ \
++ ++ +++++ +++4 c+ds l+dd l l c l dd 12 c
number of rule s : 20
1st rul e: c IA+ IB+ IC I- -
2 n d rul e : 5 sd sds
3rd rul e, 1 ldd ldd l
4th rul e, d e
5th ru le, e f
6th rul e , A FF V++ FFH I--F FV++ffh[ - -
7 t h rul e, B FW I -FF I+FW I - ffi +
8th r ule, C FFXI- FJ+ FFX I- f j +
9th ru le , H HFFH[ 1-- f f v+ + FFHI - - f f vl f f v ++f f h ] F FH
l Oth rule , h hffh [ 1-- f f v ++FFHI -- ffvl f f v ++ f f h] f f h
11th rule, V VFFVI++ffhl- -FFV++ffh lf fhl -- ffv ) FFV
12th rule, v v ffv [++ffh l - -FFV++ffh lffh l --ffv ] ffv
1 3th rule, I I FF I[+fw l -FF I+fw lfw l - f f i ]FFI
14th rule, i iffi[+fw l-FFI+fw lfwl - ffi ]ffi
15 th rule , W WFW [ l- f f i +F WI - ff il f f i +fw) FW
16th rul e , w wfw[ [ -ffi+FW I - ff ilffi+fw )fw
1 7th rule , J J FJ [ + f f x l - FJ + f f x l f f x l - f j ] FJ
1 8th rule , j j fj [+ff xl -FJ+ffx lffxl-fj ]f j
19th r ule , X XFFX[ l - f j +FF XI - f j I fj+ f f x ] F FX
20 t h rule , x xffxl l - f j+FFX I-fj I f j+ f f x] ff x
ang le, 97 360 \ » * 97 /360 : 48 .5°
re cu rsion depth: 6
sid es A B C
DRAW hor izont a l H I J
d on't draw horizon tal h i j
DRAW vertica l V WX
d o n' t draw vert ical v w x
+ +
/ C/ I
+-- - -- - + I
I I BI
I A I +
I 1/
+ +
liU
61
Georg Jagoda
The-End

Described here is a pattern showing 95 cub es building word s . The 95 cubes

look like the pattern 3D-CUBE. Three sid es are shown but her e you can see the six
sides of the cub es . The 180 "+" performs the rotation of 180 * tt * 970 3600 = 180 /

* 1800 * 970 /3600 = 180 * 48.50 = 87300 = 900 + 24 * 3600 •

The pattern is created with OLFRAC by Ton Hospel from LISTSERV at

BLEKULll.
Datafile contains:
+ + ++ ++++ + ++ +++ + + + + + + + + ++ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + ++ ++ +\
+ + + + + + ++ + + + + + ++ + + + + + + + ++ ++ +++ ++++ ++ ++++ + + + ++ ++++ + + ++ +++++ ++ \
+++++ + ++ + + + + + + + + + + + + ++ + + + + + + + + ++ +++ +++ + +++ + ++ + +++ +++ ++ ++ ++ +\
+ + [ * r e r *d*d* d*d *d*d *uuuu uur * r* r r e d *d*d e d e d e d * uuu r* r* r* r*d* \ d *d *uuuu*
u* u * r r e r e r* r* lddd* 1 * 1 * u*u*ddd*d*d* r * r * r* Idddddddd * \ r * r * r e Iddd * 1 * 1 * u *u
*ddd *d *d * r* r*r* rr *u*u * u * u * u * u * rd *ct* rd * rd* \
d * rd *u *u * u *u *u*u* r r e r* d *d *d*d* d* dl * r * r* r*ru * u * u * u * u*u l * 1 *

\ axi om
37 \ number o f ru les
A BC I GLD I AB C I \ cube (6 s ides)
A a \ dela yed front side a
B b \ delay ed r i g h t -h a n d side b
C c \ de layed upper side c
S SeeS eeS \ Ste p ( big)
\ s tep ( s ma l l)
f \ delayed f
U ee S \ U P / I\
zzT \ U p I
D I ee S I \ D o wn I / C /1
d loz T I \ down \ 1/ +------+ I
R ++ eeS-- \ Righ t I IB I
r ++ z zT-- \ right I A I +
L I + + e e S-- 1 \ Left I 1/
I 1++ z zT- - 1 \ I eft /
N I +e s - I \ N ear 1/
n I+zt- I \ n ear +-
G +e s - \ G 0 -+ ｾｨｯｲｩｺｯｮｴ｡ｬｾ
9 +z t - \ 9 0 / I
ｾ ｾｾｾｾｾｚｔ ｾ / ｬｶ･ｲｴｾ｣｡ｬｬ
z e \ don't draw DRAW
a I F FV + +F FH I- - F F V++ F F H- - \ front side a : h v V H
b +FW I -FFI+FWI - FFI \ right sid e b : i. w W I
c ++IFFX I - FJ+FFXI -F J I- \ upper side c : j x X J
H HF FH ( 1-- ff v + + FF H I - - f f v I ffv++f fh ] FF H \ DRAW ho ri zont a l
h h f f h[ 1-- f f v + + F F H I - - f f v I f f v ++ffh ] ff h \ ...dr a w ho rizon tal
V VFF V [ + +f f h 1- - F F V+ + f f h l f fh l - - f f v ] FFV \ DRAW vertic a l
v v f fv [ + + f fh j - - F F V+ + f f h lf f h J - - f fv lf f v \ <d r-aw v ertica l a
I IFFI [ + f w l -fFI+fwlfw l - ff i ] F F I \ DRAW h ori zont a l b
.i if fil +fwl-fFI+fwjf w l - ffi ]ffi \ d o n 't draw h or i zontal b
W WFW [ l -ffi+fW I - ffi lffi+fw I FW \ DRAW vertical b
wfw [j - ff i+fWI - ff i l ffi +fw lfw \ d on 't draw ve rt i cal b
J JFJ [+ffx !-fJ+ffx lffx l - fj] FJ \ DRAW horizo nt a l
J j fjl + f f x l - FJ + f f x l f f x j - f j ) f j \ d on't dr aw ho r iz o n t a l
X XFF X [ l - f j + F F X I - f j I f j + f f x ] F FX \ DRAW v ertica l
x xf fxll -fj+FFX [- fjlfj +ffxlffx \ d on't draw v er t ic a l
97 360 \ an g le = 11 * 9 7 / 3 6 0 = 48. 5°
4 \ r e c u r s i o n de pth
up downrigh tleftn ear go
/ I\ I / -+
I \ 1/ 1/ /1
+- /
62
63
Clifford A. Pickover
The Ikeda Attractor

A deep reservoir for striking images is the dynamical system. Dynamical

systems are models containing rules describing the way some quantity under goes a
change through time. For example, the motion of planets about the sun can be
modeled as a dynamical system in which the planets move according to Newton's
laws. Generally, the pictures presented in this section track the behavior of
mathematical expressions called differential equations. Think of a differential
equation as a machine that takes in values for all the variables and then generates
the new values at some later time. Just as one can track the path of a jet by the
smoke path it leaves behind, computer graphics provides a way to follow paths of
particles whose motion is determined by simple dif ferential equations. The
practical side of dynamical systems is that they can sometimes be used to describe
the behavior of realworld things such as plan etary motion, fluid flow, the diffusion
of drugs, the behavior of interindustry relationships, and the vibration of airplane
wings. The pseudocode below de scribes how to produce the Ikeda pattern. Simply
plot the position of variables j and k through the iteration. The variables scale, aroff,
and j/off simply posi tion and scale the image to fit on the graphics screen. The
Ikeda attractor has been described by K. Ikeda (see references).

cl = 0.4, c2 = 0.9, c3 = 6.0, rho = 1.0;

for (i = 0,x = 0.1, y = 0.1; i <= 3000; i + +) ( temp
= cl c3 / (1.0 + x * x + y * y); sintemp =
sin(temp);
cos_temp = cos(temp);
xt = rho + c2 * (x * costemp y * sin_temp); y = c2
* {x * [Link] + y * cos_temp);

j = x * scale + xoft; k
= y * scale + j/off;
)
61
The Ikeda Attractor 65

R e f e r e n c es

1. K. Ikeda, "Multiplevalued stationary state and its instability of the transmitted light by a ring cavity
system", Opt. Commun. 30 (1979) 257.
2. I. Stewart, "The nature of stability", Speculations in Science and Tech. 10, 4 (1987) 310324.

3. R. Abraham and C. Shaw, Dynamics — The Geometry of Behavior, Part 3: Global Behavior (Aerial
Press 1985).*
'Actually, the entire book collection of Aerial Press, including the Visual Math Series, is an educational
wonderland.
J. RangelMondragon and S. J. Abas
StarNLPl5m

Depicted here is a pattern obtained from a nonlinear transformation ap plied to a Penrose

tiling. Such tilings are generated using two different tile shapes — a "Kite" and a "Dart" and
through imposing certain matching rules on their edges [1]. The properties of Penrose tilings
are of wideranging in terest. They give rise to nonperiodic, selfsimilar patterns with fivefold
sym metry which have a host of fascinating mathematical properties [2]. They offer a
theoretical model for some newly discovered materials which have been called "Quasi-
crystals" [3, 4]. They also offer a new structure on which to base aesthetically pleasing
designs [5].

By decorating the tile pieces imaginatively, it is possible to generate a variety of designs

based on the structure of a Penrose pattern in exactly the way that has been done for
centuries with other tile shapes. The pattern shown here was obtained by decorating the tile
pieces suitably [5] and applying a non linear transformation to the whole pattern through
altering the value of the golden ratio phi which arises in a number of ways in Penrose tiles
[1]. This particular pattern is obtained through the choice phi=l instead of the usual value
phi=(l+sqrt(5))/2.

References
1. M. Gardner, "Extraordinary nonperiodic tiling that enriches the theory of tiles", Scientific
American (January 1977) 110121.
2. B. Grunbaum and W. H. Shepherd, Tilings and Patterns (Freeman & Co., 1987).
3. P. J. Steinhardt, "Quasicrystals, definition and structure", Physical Review 34, 2 (1986)
596616.
4. S. J. Abas, J. RangelMondragon, and M. W. Evans, "Quasicrystals and penrose tiles", J.
Molecular Liquids 39 (1988) 153169.
5. J. RangelMondragon and S. J. Abas, "Computer generation of penrose tilings",
Computer Graphics Forum 7 (1988) 2937.
06
67
Figure 1. A complex pattern with fivefold symmetry obtained through a nonlinear trans formation
applied to a Penrose tiling.
Ilene Astrahan
Whirlpools

The pattern depicted here is an example of manipulation of mathematically

derived material to produce a design. I used a commercial program ("Doug's Math
Aquarium" — for the Amiga) to produce a fractal pattern. This software allows the
user to change values in a basic equation. Fortunately, several equations are
supplied with the software, which is helpful for innumerates Uke me. It takes a bit
of experimentation to arrive at a result I like and can use. I can then save the file in
IFF format and load it into my paint program. I can then manipulate it and use it in
any manner. In this case, I printed a 32color image in black and white. I used a
region of the following equation:

Val: fa3(x,y,0)
FA: (al*al+a2*a2)>l?fa3(.5*(a2frac(al~2+a2~2)),.5*al+.5*a2,
a3+l):a3+al

This is how it appears in this particular program.

Hardware: Commodore Amiga Xerox 4020 inkjet printer.

Software: Deluxe Paint (Electronic Arts) Doug's Math Aquarium (Seven Seas
Software).

Reference
1. B. B. Mandelbrot, The Fractal Geometry of Nature (W. H. Freeman and Co.,
1977).
GS
Whirlpools OS
Ilene Astrahan
FernWolf (Filicinae Lupus)

The patterns depicted here are examples of fractal horticulture, showing life
forms grown from userdefined seeds. A type of fractal genetics, perhaps. I used the
public domain program called "Fractal Generator" written by Doug Houck. It
allows you to draw simple line segments to create a shape. These segments
continue to be replaced with small copies of the entire shape following the
principles of selfsimilarity. The Koch Snowflake is the best known example of this
type of shape.
The Principle of Sensitive Dependence on Initial Conditions is very much at
work here. Slight alterations in the initial drawing can produce radical alterations in
the final shape (not to mention the amount of time required to generate it). I have
also done some drawing by hand on the printout. The initial shape is shown below
— actual size — as traced from the monitor. The hardware used was a Commodore
Amiga, a Xerox 4020 inkjet printer, and a Pigma .01 drawing pen.

Figure 1 shows the original printout. Figure 2 shows it after I did some
additional artwork.

References
1. B. B. Mandelbrot, The Fractal Geometry of Nature (W. H. Freeman and Co.,
1977) p. 43.
2. W. A. KcWorter and A. Tazelaar, "Creating fractals", Byte Magazine (August
1981) 125.
7C
 71

Figure 1.
Figure 2.
John Robert Hendricks
Magic Tessaract

Described herein is a pattern showing a projection onto a piece of paper of a

fourdimensional magic hypercube of order three. It is merely one out of 58 possible
ones of this size, each of which may be shown in 384 different aspects due to
rotations and/or reflections. Larger magic hypercubes, both in size and dimension,
have also been constructed.
You will notice in the figure that each horizontal row of three numbers, such as
1, 80, and 42, sums 123. The vertical columns, such as 1, 54, and 68, sum 123.
Each oblique line of three numbers such as 1, 72, and 50, sums 123. A fourth linear
direction shown by 1, 78, and 44 sums 123.
Pairs situated symmetrically opposite the center number 41 sum 82. Thus, 1, 41,
and 81 sum 123; or 44, 41, and 38 sum 123, etc. The tessaract is bounded by cubes
which are sketched. The corners of the cubes define the corners of the tessaract.
"Opposite corners" define the fourdimensional diagonals.
The figure was first sketched in 1949. The pattern was eventually published in
Canada in 1962, later in the United States of America. Creation of the figure
dispelled the notion that such a pattern could not be made and it has taken many
years for this representation to become generally accepted.

References
1. J. R. Hendricks, "The five and sixdimensional magic hypercubes of Order 3",
Canadian Mathematical Bulletin 52 (May 1962) 171189.
2. J. R. Hendricks, "The magic tessaracts of Order 3 complete", Journal of Recre
ational Mathematics 22, 1 (1990) 1526.
72
 73

50 12 ■ 61
16 59 48
57 52 14
72 31 20
29 Y\ 27 / 67 4
22 65 / 36
80 42
78 37
44 73
64 35 24
33 19 71
26 69
75 43
79 41
39 77
54 13 56
11 63 49
47
76
74 45
40
46 / 17 60
' 15 / 55 53
62 51 10
68 —■/■ 30 25
V
34 / 23 66
21 70 32
Pablo Tamayo and Hyman Hartman
The Reversible GreenbergHastings
Cellular Automaton

The patterns shown in the figure are snapshots from the time evolu t ion of the
reversible Greenberg-Hastings cellula r automaton . The original (non-reversible)
Greenberg -Hastings cellul a r automaton was studied by Greenberg and co-workers [1]
as a dis crete model for reaction-diffusion in excitable media. Each cell ca n take on e of
three st ates : resting or quiescent (0), a ct ive (1 ), and refractory (2) . A cell remains in
the resting state until it is activated by an active neighbor. Onc e active , it will bec ome
refractory a t the next time st ep ind ependent of the neighborhood . Once refractory, it
will become resting a t the next time step independent of the neighborhood. The model
attempts to model diffusion in an ex cit a b le medium composed of dis crete chemical
oscillators. It produces rotating spirals reminiscent of the Belousov-Zhabotinsky
reaction [2]. In the two-dimensional von Neumann neighborhood, the rule can be
described as :

t+l _ F( ) _ {a LJ + 1 mod 3 , if a LJ -=I- 0;

a i ,j - t - A . 'f t - 0. (1)
n erg , 1 AI ,] ,

wh ere ｡ ｾ Ｌ ｪ is the state of the cell at site (i , j) at time t , and An e ig is a Boolean

activation fun ction which is equal to 1 if at least on e of the neighbors is active and
equal to 0 otherwise,

An e ig = [aLl ,j = 1] or ｛｡ｾＫｬＬｪ = 1] or [aLJ-l = 1] or [aLJ+l = 1] . (2)

This cellular automaton is made reversible by the Fredkin method which consists of
subtracting the value of the cell in the past ,
t+l
a·· =
F()t -
t - l mo d 3 .
a·· (3)
t ,) 1,,)

In this way the rule be comes reversible and displays a different behavior from the
original non-reversible rule . It has a rich phenomenology which includes soliton-like
structures and chemical turbulence [3] . Depending on the initial conditions it develops
into three differ ent regimes: (i) regular regim e with short recurrence times , (ii ) che
mical-t u r bu len t regime, and (iii) disorder ed random regime. The most interesting on e
is the che m ica l-t urb u lent regime , examples
14
75
Figure 1. Patterns from the time evolution of the reversible GreenbergHastings model in the two-
dimensional hexagonal lattice. Only active (black) and resting (white) cells are shown. The sequence
shown in lefttoright, toptobottom order correspond to times: t = 343, t = 500, t = 735, t = 1004, t = 2043,
t = 15064. The skewness is due to the fact that the hexagonal lattice was simulated using 6 neighbors in
the square lattice.
76 P. Tamayo & H. Hartman

of which are shown in the figure (for a hexagonal twodimensional lattice). Only active
(black) and resting (white) cells are shown. The initial condition consisted of one single
active cell in a resting background. These patterns were generated using a CAM6 Cellular
Automata simulator [4]. For more information on the reversible GreenbergHastings model,
see reference 5.

References
1. J. M. Greenberg and S. P. Hastings, Siam J. Appl. Math. 34 (1978) 515; J. M. Greenberg,
B. D. Hassard, and S. P. Hastings, Bull, of the Am. Math. Soc. 84
(1978) 1296. See also: A. Winfree, Physica 17D (1985) 109.
2. H. Meinhardt, Models of Biological Pattern Formation (Academic Press, 1982).
3. Y. Kuramoto, Chemical Oscillations, Waves and Turbulence (SpringerVerlag,
1984).
4. T. Toffoli and N. Margolus, Cellular Automata Machines (MIT Press, 1987).
5. P. Tamayo and H. Hartman, "Reversible cellular automata and chemical turbu lence",
Physica D 45 (1990) 293.
Ilene Astrahan
Cybernetic Rapids

The pattern depicted here is an example of manipulation of mathematically

derived material to produce a design. I used a commercial program ("Doug's Math
Aquarium" — for the Amiga) to produce a fractal pattern. This soft ware allows the
user to change values in a basic equation. Fortunately several equations are supplied
with the software, which is helpful for innumerates like me. It takes a bit of
experimentation to arrive at a result I like and can use. I can then save the file in IFF
format and load it into my paint program. I can then manipulate it and use it in any
manner. In this case, I printed a 32color image in black and white. I used a region of
the following equation:

Val: fa3(x,y,0)*10
FA: a3<1000?fb4(al*al,a2*a2,2*al*a2,a3+l)*(ala2)+l : 0 FB: al+a2<4?
fa3(ala2+x,a3+y,a4):0

This is how it appears in this particular program.

Hardware: Commodore Amiga Xerox 4020 inkjet printer.

Software: Deluxe Paint (Electronic Arts) Doug's Math Aquarium (Seven Seas
Software).

Reference
1. B. B. Mandelbrot, The Fractal Geometry of Nature (W. H. Freeman and Co.,
1977).
77
78

P.

J3
>>
O
C. William Henderson
Pinwheels

Described here is a pattern produced on an AMIGA 1000, a 512K computer, and

output to a Xerox 4020 color ink jet printer. The software is Doug's Math Aquarium
by Seven Seas Software, a program that permits the artist to create images based
solely on mathematical equations. The AMIGA graphics dump can simulate in print
nearly all of the computer's 4,096 color range, though only 32 can be displayed on
screen at one time. However, in the case of the image shown here, because of the
black and white limitations, the graphics dump was set to reproduce in a range of 16
levels of gray.
Since the images are generated purely by mathematics, appearance is de
termined by the artist's selection and manipulation of various formulas. No paint
system nor image touchup is involved. Prom the formulas the software calculates a
threedimensional surface by equating Z to the computed com binations of X and Y,
then projects the threedimensional surface onto the screen as a twodimensional
contour map (though 3D displays are possible). The colors represent the contour
intervals. To the degree that the contours exceed 32 (in this case 16), the colors
repeat themselves.
As an example, the simplest formula might be just X. Thus, for every
combination of X and Y, Z would be equal to X only. For the entire height of the
screen, Z would be at a distance from the screen of X. This would result in a
computed vertical plane going into the screen from the right. Its contours would be
vertical lines. However, if the colors grade from dark to light to dark, the illusion on
screen would be a series of vertical cylinders. Thus, judicious selection of color and
sequence can produce illusions of 3D basreliefs, the actual threedimensional surface
having little resemblance to the 2D projection.

The X — Y formula used to generate the Pinwheels image here involves simple
trigonometric functions. The equation that resulted in this unusual image is simply ...
sin(X + cos Y)/ cos(y + sin X).
79
80
Pinwheels
C. William Henderson
Circlefest

Described here is a pattern produced on an AMIGA 1000, a 512K computer, and

output to a Xerox 4020 color ink jet printer. The software is Doug's Math Aquarium
by Seven Seas Software, a program that permits the artist to create images based
solely on mathematical equations. The AMIGA graphics dump can simulate in print
nearly all of the computer's 4,096 color range, though only 32 can be displayed on
screen at one time. However, in the case of the image shown here, because of the
black and white limitations, the graphics dump was set to reproduce in a range of 16
levels of gray.
Since the images are generated purely by mathematics, appearance is de
termined by the artist's selection and manipulation of various formulas. No paint
system nor image touchup is involved. From the formulas the software calculates a
threedimensional surface by equating Z to the computed com binations of X and Y,
then projects the threedimensional surface onto the screen as a twodimensional
contour map (though 3D displays are possible). The colors represent the contour
intervals. To the degree that the contours exceed 32 (in this case 16), the colors
repeat themselves.
As an example, the simplest formula might be just X. Thus, for every
combination of X and Y, Z would be equal to X only. For the entire height of the
screen, Z would be at a distance from the screen of X. This would result in a
computed vertical plane going into the screen from the right. Its contours would be
vertical lines. However, if the colors grade from dark to light to dark, the illusion on
screen would be a series of vertical cylinders. Thus, judicious selection of color and
sequence can produce illusions of 3D basreliefs, the actual threedimensional surface
having little resemblance to the 2D projection.

The following formula for CIRCLEFEST involves an IFTHENELSE state ment

and MODULO (an integer remainder, i.e., 11 mod 3 = 2). Some of the smaller
circles are a moire effect, a special feature of the program.
If 35 x (X'2 + Y"2)".l mod 4 = 1 then {FA mod 8) + 16 else {FB mod 16) + 8,
where FA = Y X and FB = X'2 + Y'2.
81
82
Circlefest
C. William Henderson
XOR Size

Described here is a pattern produced on an AMIGA 1000, a 512K computer, and

output to a Xerox 4020 color ink jet printer. The software is Doug's Math Aquarium
by Seven Seas Software, a program that permits the artist to create images based
solely on mathematical equations. The AMIGA graphics dump can simulate in print
nearly all of the computer's 4,096 color range, though only 32 can be displayed on
screen at one time. However, in the case of the image shown here, because of the
black and white limitations, the graphics dump was set to reproduce in a range of 16
levels of gray.
Since the images are generated purely by mathematics, appearance is de
termined by the artist's selection and manipulation of various formulas. No paint
system nor image touchup is involved. From the formulas the software calculates a
threedimensional surface by equating Z to the computed com binations of X and Y,
then projects the threedimensional surface onto the screen as a twodimensional
contour map (though 3D displays are possible). The colors represent the contour
intervals. To the degree that the contours exceed 32 (in this case 16), the colors
repeat themselves.
As an example, the simplest formula might be just X. Thus, for every
combination of X and Y, Z would be equal to X only. For the entire height of the
screen, Z would be at a distance from the screen of X. This would result in a
computed vertical plane going into the screen from the right. Its contours would be
vertical lines. However, if the colors grade from dark to light to dark, the illusion on
screen would be a series of vertical cylinders. Thus, judicious selection of color and
sequence can produce illusions of 3D basreliefs, the actual threedimensional surface
having little resemblance to the 2D projection.

The formula for this image involves trigonometry, XOR (bitwise exclusive OR
function of integer values) and an IFTHENELSE statement.
Thus, if X > Y then FAx FB, else FC x FB.
Where FA = (2X XOR 1) + (Y/2 XOR 1) x (Y XOR X) FB =
sin(X + Y) + sin(X Y)
FC = (X/2 XOR 5) + (2Y XOR 5) x (Y XOR X)
S3
84
XOR Size
C. William Henderson
Trig Gem

Described here is a pattern produced on an AMIGA 1000, a 512K computer, and

output to a Xerox 4020 color ink jet printer. The software is Doug's Math Aquarium
by Seven Seas Software, a program that permits the artist to create images based
solely on mathematical equations. The AMIGA graphics dump can simulate in print
nearly all of the computer's 4,096 color range, though only 32 can be displayed on
screen at one time. However, in the case of the image shown here, because of the
black and white limitations, the graphics dump was set to reproduce in a range of 16
levels of gray.
Since the images are generated purely by mathematics, appearance is de
termined by the artist's selection and manipulation of various formulas. No paint
system nor image touchup is involved. Prom the formulas the software calculates a
threedimensional surface by equating Z to the computed com binations of X and Y,
then projects the threedimensional surface onto the screen as a twodimensional
contour map (though 3D displays are possible). The colors represent the contour
intervals. To the degree that the contours exceed 32 (in this case 16), the colors
repeat themselves.
As an example, the simplest formula might be just X. Thus, for every
combination of X and Y, Z would be equal to X only. For the entire height of the
screen, Z would be at a distance from the screen of X. This would result in a
computed vertical plane going into the screen from the right. Its contours would be
vertical lines. However, if the colors grade from dark to light to dark, the illusion on
screen would be a series of vertical cylinders. Thus, judicious selection of color and
sequence can produce illusions of 3D basreliefs, the actual threedimensional surface
having little resemblance to the 2D projection.

The X — Y formula used to generate the TRIG GEM image here involves simple
trigonometric functions. The equation that resulted in this exotic image is simply:
[sin(X + Y)/ sin(X Y)] + [sin(X Y)x(X + Y)]. The image field is offset from the
intersection of the X and Y axes.
8b
86
Trig Gem
Gary Ricard zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDC
Locked Links

Described here is a pattern generated by a program which draws "locked links".

The links are always regular polygons and they are "locked" together by weaving
their edges over and under each other. This is best illustrated in Fig. 1. Input to the
program which generates the locked links includes the total number of sides for the
polygon in the center of each figure. Even numbered inputs provide figures with two
separate intertwined links, each with half the number of sides originally requested.
Odd numbered inputs generate a single link with the input number of sides that
intertwines with itself, basically forming a knot. Figure 1 contains examples for
initial inputs of 3, 4, 5, 6, and 8 sides. Other inputs to the program include the length
of a side of the polygon in the center of the figure, thickness of a link, center of the
figure, and angle at which to begin drawing. The program is an intense exercise in
algebra and trigonometry but is actually quite compact when completed. (My
version is about 100 lines of code.)

Figure 2 is a tiling using a 4-sided locked link. Each locked link is placed at the
center of a gothic cross. Thus, the base geometry for the tiling is given below:

What makes Fig. 2 interesting is that it is a minor optical illusion. The figures
along the diagonals in the pattern seem to bend and curve as one attempts to follow
a diagonal with the eye. If you stare at a particular region
of the pattern, the diagonals on the periphery of your vision seem to curve to zyxwv
87
88 G. Ricard

the boundary of the figure. The effect is certainly caused by the positioning of the
locked links relative to one another. The mind wants to force the center of the
locked links to be on a 45degree line even if the straight diagonal must be bent to
do it!
Figure 1.
89
F i g u re 2.
Gary Ricard
Exponential Tunnel

Described here is a pattern generated by incrementing variable "s" by 1 from 0

to 750 in the following 2 pairs of parametric equations:

1) x = exp(.012s) cos(1.24s) y
= exp(.012s)sin(1.24s)

2) x = exp(.012s) cos(1.24s + .5236) y =

exp(.012s) sin(1.24s .5236)

The first pair of equations generates a design with 5 arms spiraling clockwise
from the center. The second pair generates the same design but with the 5 arms
spiraling counterclockwise from the center. (If you are examining the color figure
these are the blue and red spirals respectively.) The result, especially in color, looks
like an exponential tunnel with spiral arms descending into infinity. Alternatively,
some see the result as a 5pelated flower. This is actually the more common
interpretation when both spirals are the same color. Another interesting point is that
the spirals are generated using only straight lines.
The intertwined border surrounding the spirals is generated using the "locked
links" program as described in the article by that name.
A program which will generate the spirals on the VGA screen using Turbo
Pascal follows. Unfortunately the screen version of the design does not look nearly
as impressive as the plotter version.

PROGRAM scrspiral;
USES graph;
VAR m, b, phase: REAL; grdriver,
grmode: INTEGER;

FUNCTION x(s: INTEGER; phase: REAL): REAL;

BEGIN
x:=exp(0.012 * s) * cos(1.24 * s+phase)
END;
90
Exponential Tunnel 91

FUNCTION y(s, sign: INTEGER; phase: REAL): REAL;

BEGIN
y:=exp(0.012 * s) * sin(sign * (1.24 * s+phase)) END;

FUNCTION scale(z:REAL): INTEGER;

BEGIN
scale:=round(z * m+b)
END;

PROCEDURE spiral(phase: REAL; sign: INTEGER);

VAR s: INTEGER;
BEGIN
s:=0;
moveto(scale(i(s, phase)), scale(y(s, sign, phase)));
WHILE s<=750 DO
BEGIN
s:=s+l;
lineto(scale(x(5, phase)), scale(y(s, sign, phase)))
END
END;

BEGIN
m:=239.5;
b:=m;
initgraph(grdriver, grmode, 'C:\PASCAL\GRAPHICS');
setwritemode(xorput);
setbkcolor (lightgray);
setcolor (blue);
spiral(0.0,l); setcolor(red);
spiral(0.5235987756,l);
readln;

closegraph
END.
92 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Exponential Tunnel
Hiroshi Okumura
4
A Generalization of the Regular Tiling (4 )

Described here is a generalized pattern of (44 ), the regular tiling by congru ent squares.
Figure 1 shows a configuration consisting of two points X, Y and lines Xi and yi (i and j are
integers) passing through X and Y respectively, such that the quadrangle determined by Xi,
a^+i, yj, yj+i has an incircle for any i and j [1]. Figure 2 shows a special case where both X
and Y are points of infinity, and it also shows that Fig. 1 can be regarded as a generalized
pattern of regular tiling (44 ).

Our generalized configuration has the following properties:

(1) For any two circles, one of the centers of similitude (two intersections of internal
common tangents and external common tangents) lies on the line XY.
(2) If we denote the curvature (reciprocal of the radius) of the circle touching
Xi, xi+1, y^ yj+i by [i, j] then

[i, j] + [i + m, j + n] = [i + m, j] + [i, j + n],

[i, j][i + m + n, j — m + n] = [i + m, j m][i + n, , j + n] ,

provided we make the convention that we regard the circles below the line XY as having
negative radii.

(3) For any integers i, j , k, the quadrangle determined by Xi, Xi+k, yj, yj+k has an incircle
(Rigby [2], Theorem 4.5).

References
1. H. Okumura, "Configurations arising from the threecircle theorem", Mathematics
Magazine 63, 2 (1990) 116121.
2. J. F. Rigby, "Cycles and tangent rays, circles and tangent lines", Mathematics Magazine
64, 3 (1991) 155167.
93
94

Parallel lines intersect at a point of infinity.

Figure 1.
 95

Figure 2.
Hiroshi Okumura
A Pattern by Fujita Configurations

Described here is a pattern by Fujita Configurations. Let ABCD be a

parallelogram; E, a point on the segment CD; F and G, points on the segments CE
and DA, respectively, and let FG meet BE at H. Suppose the quadrangles ABHG,
BCFH, and DGHE have incircles (see Fig. 1). This figure seems to have been first
considered by one of the Wasan mathematicians S. Fujita for the case where ABCD
is a square (see Fig. 2) [1]. (Wasan refers to Japanese mathematics developed
independently of Western science between the 17th and 19th centuries.) And we
call this figure a Fujita configuration.
The four incircles in the Fujita configuration (including the incircle of the
triangle FEH) have an interesting property, that is, the sums of the oppo site radii
are equal [2, 3]. Especially for a Fujita configuration, in which its parallelogram is a
square, FEH is a rightangled triangle with the sides EF:FH:HE=[Link] and the ratio
of the radii of the four incircles is [Link] (see Fig. 3). Using this figure, we can
construct a pattern in the plane (see Fig. 4). In this pattern, we can superpose similar
ones. Two similar patterns which are double and four times the size are drawn on
the figure.

References
1. S. Fujita, Seiyo Sampo 3 (1781).
2. H. Okumura, "Four circles in a parallelogram", J. Recreational Mathematics 19,
3 (1987) 224226.
3. H. Okumura, "Fujita configurations", J. Recreational Mathematics 21, 1 (1989)
2934.
96
 97

Figure [Link]+=+.
Figure 2.
98

Figure 3.
Figure 4.
Kenneth J. Hooper
Star Trails

Described here is a pattern showing one example of the structures contained

within the Mandelbrot Set [1]. This area is sometimes ignored by traditional
computer graphics methods, which produce images from just out side the
Mandelbrot Set [2, 3].
Computergenerated images of the Mandelbrot Set are created by repeat edly
solving the following two equations:

x=x —y+a and y = 2xy + b.

Variables a and 6 typically range in value from —2 to 2. Initially, x and y are set to
zero but subsequently take on the values from the previous calculation. Each xy pair
defines the position of a point on a complex plane and every repetition of the
calculations yields a new position. If a starting location inside the Mandelbrot Set is
chosen, the xy points remain bounded within a small region, sometimes tracing a
path that repeats itself. Figures 1 and 2 were generated by counting the number of
repetitions it took to trace a path before it repeats.

An algorithm for counting the number of repetitions makes use of the ratio y/x.
Each ratio is then stored for later comparison with the first y/x value to determine
when the sequence repeats. The following pseudocode for an inner loop details this
method (the outer loops supply values for a, b, and pixel column and row):

x < 0, y < 0 for

i < 0 to 15
xx<x*xy*y + a yy<2*x*y
+b
tan(i) < yy / (xx + .000001) Note: Prevents divide by zero.
X< XX

y < yy
endloop
for j < 1 to 15
error < tan(0) tan(j)
99
100 K. J. Hooper

if( |error| < .05) then Note: 1. |error| = absolute value

setpixel(column, row) 2. The loop counter can be
exit loop used to define pixel
endloop shade of color.

Figure 1 shows a starting location near a = —0.42, b = 0.63, which is near the
boundary of the Mandelbrot Set. Self similar patterns are apparent as the stars seem
to trail away in the distance. Actually, the patterns cross over to outside the
Mandelbrot Set but the ratio y/x still repeats. The difference is that the points are no
longer bounded. Figure 2 is located near a = 0.44, b — 0.34, and shows similar
patterns crossing the boundary of the Mandelbrot Set.

References
1. C. A. Pickover, "Inside the Mandelbrot Set", Algorithm 1, 1 (1989) 912.
2. B. B. Mandelbrot, The Fractal Geometry of Nature (W. H. Freeman, 1983).
3. A. K. Dewdney, "Computer recreations", Scientific American (September, 1986).
 2
Figure 1. Star patterns within the Mandelbrot Set seem pulled toward the boundary. zyxwvutsrqponmlkjihgfed
 2

Figure 2. A familiar "flame" structure from traditional Mandelbrot Set images is instead
suggested from the star patterns. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
David Scruton
An Iteration Map

Shown here is an image derived from complex dynamics. In particular, the

3
figure represents a complex variable recursion produced by iterating z = z + /i forty
times. ^ is a complex constant. Diverging points whose magnitude is between 2 and
3 are plotted. The picture was computed on a microVAX and plotted on a Talaris
laser printer.
103
Mieczyslaw Szyszkowicz
A SelfSimilar Structure Generated by a
BrokenLinear Function

Consider the brokenlinear transformation with the parameter

f(x,r)=r(l2\x0.5\), 0<r<l,

which maps
/(*, r) : [0, 1] —» [0, 1] .

Figure 1 shows the behavior of the sequences {xn} and {yn} produced by this
transformation used iteratively with r = 1

xn+i = 1 2\xn 0.5|,

j/„+i = 1 2|j/n 0.5|.

The iterative process is realized with the stopping criterion (xn+i — xn) + (j/n+i —
2
yn) < T, where r is the assumed threshold. Starting with the initial point (XQ, J/O) ^
[0, 1] x [0, 1], the sequence of the points (xn, yn) is generated. If the stopping
criterion is satisfied, then the iterative process is stopped and the number of executed
iterations modulo 2 is mapped on the point (XQ, yo) Figure 1 represents a black (0)
and white (1) selfsimilar pattern constructed by the above algorithm.
104
 105

Figure 1. Self-similar stru ct ur e generated by the fun ction f(x , r) = r(l - [z - 0 .51) with
r=l .
Mieczyslaw Szyszkowicz
Roots of Algebraic Polynomials

Algebraic polynomials are functions of the form

M

f(z) = Y<akzh>
fc=0

where a^, k = 0, 1,..., m are fixed real or complex numbers. The number m > 1 is the
degree of the algebraic polynomials. The fundamental theorem of algebra states that
every algebraic equation f(z) = 0 has at least one root. More precisely, if a
polynomial vanishes nowhere in the complex plane, then it is identically constant.
An algebraic equation of degree m has exactly m roots, counting multiplicities. If
the real polynomial / (i.e., the coefficients <2fc, k = 0, 1,..., m are real numbers) has
a complex zero r = a + i/3, then it has also the complex root r = a — i(i. Here, r
denotes the conjugate of r, a and /3 are real numbers, and i = %/—T. In other
words, the complex zeros of a real algebraic polynomial always appear in conjugate
pairs. The zeros of an algebraic polynomial with complex coefficients may not
occur in conjugate pairs. As an example, the following algebraic polynomial is
considered

f(z) = z(z 0.25i)(z + 0.5)(z + i),

which has the zeros 0, i/4, —1/2, and —i.

Figure 1 shows the behavior of Newton's method used to localize the roots of
the equation f(z) = 0, where / is defined above. The initial values ZQ belong to the
quadrant [—1.5, 1.5] x [—1.5, 1.5]. The sequence {zn} is generated according to
Newton's rule

zn+1=zn—— n = 0, 1, 2,... ,
/ {Zn)

and if |zn+i — zn\ < r (r is a fixed threshold), then the iterative process is stopped and
the number of executed iterations M is mapped to the point ZQ. Figure 1 shows the
result of this map as a black and white pattern: (black = 0, white = 1), where 0 and 1
are obtained by the rule, M modulo 2.
106
107
Figure 1. Newton's method used to localize the complex algebraic polynomial.
Rastislav Telgdrsky
Mosaics

Described here are patterns derived from the chessboard, where black and white
squares become rectangles, and their uniform colors are replaced by other patterns.

Each mosaic is a repetition, in an M by N matrix, of a mosaic element. This

element is generated by a single vector, whose coordinates are increasing. I have
used arithmetic, geometric, Fibonacci, and also random generated se quences to
control the rate of growth, regularity, etc. The mosaic element has four axes of
symmetry. The drawing of each rectangle is repeated four times, except for the
middle cross, where the number of repetitions is two — one at the nodes of the
cross, and one at the middle of the cross.
The following fragment of C code describes a mosaic element which is repeated
M by N times:

for (i = 0; i < m — 1; i + +)
for (j = 0; j < m 1; j + +) { fc = ((i
+ j) %t) + l+w;
pat(sel, a + v[i], b + v\j], a +v[i + 1] 1, b + v\j + 1] 1, k); if (i! =
m2)
pat(sel, a + sm — v[i + 1] + 1, b + v\j], a + sm — v[i], b + v[j + 1] — 1, fc); if (j! = m
2)
pat(sel, a + v[i], b + sm. — v\j + 1] + 1, a + v[i + 1] — 1, b + sm — v\j], fc); if (i + j \
=2*m 4)
pat(sel, a + sm — v[i + 1] + 1, b + sm — v\j + 1] + 1, a + sm — v[i\, b + sm — v\j], fc);
}

where k is the color, sel is the selected pattern pat (such as drawing two lines between
some of the four points), v is the control vector of dimension m (for example,
representing an arithmetic sequence, like v[0] = 0, v[l] = 3, for (i = 2; i < m; i +
+)v[i] = v[i — 1] + i + 1, sm = v[m 2] + v[m 1] 1), and (a, b) are coordinates of
the left top of the mosaic element.
Further modifications are obtained by using a random increment in the
definition of the control vector. Different mosaic patterns that result from warping
onto curved surfaces, and carefully projected onto a view plane, for example, are
achieved by using Painters Algorithm. Finally, color mosaics, displayed on a high
resolution screen, are a real feast for the eyes.
108
 109

F i g u r e 1. Mosaic based on
arithmetic sequence with two uniform fillings.
F i g u r e 2 . Mosaic based on arithmetic sequence and two alternating patterns: two diagonal line
segments and a recursive box.
110 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

F i g u re 3 . Mosaic based on Fibonacci sequence and two alternating patterns: two diagonal line
segments and a recursive box.
Figure 4 . Mosaic based on exponential sequence with two uniform fillings.
 I l l zyxwvu

F i g u re 5. Mosaic based on exponential sequence and two alternating patterns: two diagonal line
segments and a recursive box.
Figure 6. Mosaic drawn on hat-like surface and projected onto a view
plane.
Mieczyslaw Szyszkowicz
Pattern of Euler's Formula

Much computer graphics art is created as a result of the illustration of the

behavior of the sequence {zn} through some kind of iterative process. Usually this
sequence is generated by the iteration of the form

zn+1 = $ ( z n ) , n = 0, 1, 2 , . . . ,

where ZQ is a given initial value. The iterative function <& defines the recurrence
relation. Often the behavior of the sequence {zn} is studied as a function of the initial
values ZQ. The same approach is used here. Consider the following initial value
problem defined as the system of ordinary differential equations

dx/dt = sin(y),
dy/dt = sin(x),

with the initial values XQ, yo and xo, yo G [—5, 5]. The problem is to find the
function x{t) and y(t). t is an independent variable. One of the simplest nu merical
methods able to solve this problem is Euler's method. In this situation, Euler's
formula generates the following recurrence processes

xn+i = xn + hsm(yn),
2/n+i = yn + hsm(xn),

where h is a constant called the step of the integration. xn ~ x(tn) and yn ~ y(tn),
where x(t) and y(t), are the exact solutions. Figure 1 displays the number of
executions of the above process for the given initial values. This number modulo 2
is associated with initial point (xo, yo), and consequently, a black and white pattern
is created. The above process was realized with the following stopping criterion: IF
2 2
(xn + i xn) + (yn+i yn) < r, THEN stop, where r is a constant number. Other
stopping criteria produce interesting pictures.
112
113
Figure 1. Euler's formula as a function of the initial values.
Mieczyslaw Szyszkowicz
SelfAccelerating Version of Newton's Method

Traub [1] presented the following selfaccelerating recursion to solve the

nonlinear equation f(z) = 0

2 n + l = Zn , 71 = 0, 1, 2, . . . ,
In

_ f\zn + (^nfn) ~ In
In p^ 7 i
^njn

C —
7fcl

where ZQ is given and fn = f{zn). This method was proposed for use in the complex
plane. Presented here is a pattern showing the behavior of the sequence {zn}
generated by the above iterative process. The sensitivity of {zn} to the initial values
ZQ is displayed. Figure 1 shows the number of executed iterations (modulo 2) by
A
alternating colors. The equation f(z) = z — 1 = 0 is solved in the complex domain.
This equation has four roots, which are 1, —1, % and —i, i = \/—T. Co = —l/f'(z0).
The stopping criterion of this selfaccelerating algorithm is as follows: IF |z n+i — zn\
6
< r THEN stop, where T is a constant number (r = 10 ). Initial values are between
—1.5 and 1.5 in the real and imaginary directions.

Reference
1. J. F. Traub, Iterative Methods for the Solution of Equations (Prentice Hall, 1964).
114
115
Figure 1. Selfaccelerating methods used to localize the roots.
Ken Shirriff
Voronoi Fractal

Described here is a pattern showing a fractal generated by recursive con

struction of Voronoi diagrams. This technique can be used to generate patterns
resembling roadmaps, leaf veins, butterfly wings, or abstract patterns.
A Voronoi fractal is constructed by first drawing the Voronoi diagram of a set
of points. (A Voronoi diagram [1] of a set of points divides the plane into regions;
each region is closer to one point than any other.) Then, using a larger set of points,
a smaller Voronoi diagram is drawn inside each of the original regions. This process
continues, recursively subdividing each region by drawing smaller Voronoi
diagrams. Voronoi fractals are discussed in more detail in [2].

The figure shows a Voronoi fractal generated by three levels of subdivision on

points in a circle, with the points distributed towards the boundary. The first level
uses 20 points, and the following levels use 400 and 8000 points.

References
1. F. P. Preparata and M. I. Shamos, Computational Geometry (SpringerVerlag,
1985).
2. K. Shirriff, "Generating fractals from Voronoi diagrams", Computers and
Graphics 17, 2 (1993) 165167.
116
117
Mieczyslaw Szyszkowicz
The Logistic Map in the Plane

Presented here is a pattern obtained by repeating a simple mathematical operation, such

as

zfc+i = rxk(l Xk), k = 0, 1, 2, . . . ,

where XQ is a given initial value and r is a parameter between 0 and 4. The behavior of the
sequence {xk} is very sensitive to the value of the parameter r (see [1]). The obtained
sequence may be attracted (converge) to one point, a group of points, or an infinite number
of points composed by a fractal structure called a strange attractor. Figure 1 was generated
by the above iteration executed for two variables x and y:

xk+i = rxk(l Xk),

yk+i = syk(l yk),
k = 0, 1, 2 , . . . ,

where r = s or r / s. The above process was realized with r = s = 3.7 for the points (XQ, yo),
where x$ and yo have values between 0 and 1. For each initial point the sequences {xk} and
{yk} were produced with the following stopping criterion:

IF abs(xk xk+i) <r OR abs(yk — yk+i) < T THEN stop.

Here, r is a constant value. Figure 1 was generated with r = 0.45. The number of executed
iterations modulo 2 is represented.

Reference
1. R. May, "Simple mathematical models with very complicated dynamics", Nature 261
(1976) 359467.
118
 119

(S2wo*afa£
Figure 1. Pattern generated by the logistic parabola /(x) = rx(l — x).
Mieczyslaw Szyszkowicz
Modified Logistic Map in the Plane

Presented here is a pattern resulting from the iteration, or repeated applica-tion, of a

simple mathematical op eration . In particular, the pattern shows the behavior of the
sequences {x n} and {Yn} generated by the following formulae

Xn+l = R,.. x n(l - x n),

Yn+! = R,..Yn(1- Yn) ,
R,.. = r + (x n + Yn)/2, n
= 0, 1, =, 2, ... , N.

If r E [0,3], and xo , Yo E [0, 1], then the elements of the sequences {x n} and {Yn} are
between 0 and 1 [1] . Figure 1 shows the pattern obtained by the above iterations with r
= 3 and 0 < xo , Yo < 1. For eac h start point (xo , Yo) the it erative process is executed
and aft er seven iterations (N = 6) , the value X7 + Y7 is used to determine the color of
the initial point (xo, Yo). In our situation, black (0) and white (1) was used by the rules
5(X7 + Y7) mod 2. The logistic parabola function is only an example function, others
are good as well. The pattern produced by reduction of the interval [0, 1] to [0, a]' where
o < a < 1 shows more details.

Reference
1. R. May , "Sim p le mathematical models with very complicated dynamics" , Nature
261 (1976) 359-467.
120
121
Figure 1. Modified logistic function in the plane.
Mieczyslaw Szyszkowicz
Newton's Method in the CPlane

Presented here is a pattern showing the behavior of the sequence {zn} as a

function of the variable c in the following nonlinear equation:

/ ( z) = 2 4 C = 0.

This sequence is generated by Newton's method which is used to localize a root

of the above equation. The function / is considered in the complex plane. Newton's
method is realized by the iteration

zn+1=zn±p\ n = 0,1,2,.... (1)

J \zn)

This equation is well known and described in [1] for fixed values of c (usually c
= 1), and ZQ is changed over the complex plane. In this situation, the Juha sets are
created (see [1]). We propose the realization of the iteration (1) with a fixed ZQ and
variable c. This is something similar to the approach used to create the Mandelbrot
set [1]. Denote by f2c a subset of the complex plane, and assume that fi c is the region
of the variability of c. The iteration (1) is executed with the stopping criterion (the
6
test \zn+i —zn\ < r, where r = 10 ), and if this criterion is satisfied, then the iteration
is stopped. The number of executed iterations is expressed by black and white
contours. Figure 1 is an example pattern for ZQ = 1 + 2. For the uniquely defined f2 c
there is infinite number of patterns. Each of these is specified by an initial z$. The
reader may wish to study the behavior of {zn} simultaneously with variable c and ZQ
values.

Reference
1. H.O. Peitgen and D. Saupe, eds., The Science of Fractal Images (SpringerVerlag,
1988).
122
 123

Figure 1. Newton's method applied to the equation f(z) = z4 — c = 0, with the initial value 1 + i, and c
from the region [—0.5, 20.5] x [15.0, 15.0].
Mieczyslaw Szyszkowicz
Iterations with a Limited Number of
Executions

Iterative processes are used very often to create computer graphics arts. The classic
example is the iteration realized in the complex plane

Zn+l = zl+C, 71 = 0, 1, 2, ... ,

where zo = 0 and c is varied over the complex plane. This iteration is the basis for the
construction of the Mandelbrot set [1, 3]. In the above iterative process, the number of
executions is controlled by the value of the norm z, i.e., \z\ = y/x2 + y2 where z = x + iy.
These numbers are often displayed as color patterns. Another approach is to realize the
iterative process for a fixed number of times and to show the obtained value for the
sequence. As an example, consider the iteration generated by the logistic map [2, 4]

xk+1 = rxk(l xk), fc = 0, 1, 2, { i fc } c [ 0 , I ] ,

where io is a given initial value. The behavior of the sequence {xk} is very sensitive to the
value of the parameter r,r S [0, 4]. Figure 1 shows the values XK +VK, represented by black
and white, and obtained after K = 6 iterations

xk+1 = rxk{\ Xk),

2/fc+i = syk(l j/fe),

where r = s or r ^ s. Figure 1 was obtained with r = s = 3.9 and with the initial values xo, yo,
where XQ and yo have values between 0 and 1. The point with the coordinates (xo, y0) was
colored by scaled value i 6 + y6.

References
1. A. Dewdney, Computer Recreation (Scientific American, 1985).
2. B.L. Hao, Chaos (World Scientific, 1984).
3. B. Mandelbrot, The Fractal Geometry of Nature (W. H. Freeman and Co., 1983).
4. R. May, "Simple mathematical models with very complicated dynamics", Nature 261
(1976) 359467.
121
125
Figure 1. Pattern generated by six iterations of the logistic map.
Stephen D. Casey
The Starry Night — Iterates of Tan(z)

"The Starry Night" is a "mathematical Van Gogh", giving us a glimpse of a

nebular cluster which has been generated by mathematical, as opposed to natural,
forces. The study of this pattern falls under the iteration theory of Fatou and Julia,
which examines discrete dynamical systems on the Riemann Sphere C with a
dynamic given by a nonlinear holomorphic or meromorphic mapping from Cl C C to
C (see [13]). This particular pattern was generated by the meromorphic function

f(z) = A • tan(z)

in A or parameter space in a neighborhood of the point %/2/2 + i(\/2)/2. At each

pixel position (x, y) = x + iy = A, the orbit of the initial value 7r/4 was calculated,
i.e., zn+i = A • tan(z„), ZQ = 7r/4, which generated the sequence zi = A, z2 = A •
tan(A), z$ = A • tan(A • tan(A)),... . This sequence was generated until either the
absolute value of zn was greater than or equal to 100 (which guarantees that zn
diverges to oo) or the number of iterates exceeded 256. In the picture, the black
regions correspond to points A for which |z256| < 100, while the brighter regions
correspond to divergent zn. The brighter the tone, the fewer the number of iterates
needed for zn to escape. Note that there is a hot spot in the center of each "star".

The mathematical interpretation of the pattern tells us many things. First, we

should note that the pattern is actually a bifurcation diagram for the family of
dynamical systems f(z) = A ■ tan(z). From the picture, we can see that g{z) = tan(z) is
not structurally stable (as denned in [3]) because a slight variation in A can produce a
wide variation in the behavior of the iterates. For fixed A, the function f(z) has an
essential singularity at oo. Therefore, by the Great Picard Theorem, in every
neighborhood of oo, f(z) often takes on every complex number infinitely. By an
argument similar to that in [3], it can be shown that those values of A for which zn
diverges correspond to dynamical systems A ■ tan(z) which have a Julia set equal to
the entire complex plane. It is also of interest to note that the distribution of the
divergent points forms a fractal pattern, which is seen by observing that the larger
"stars" generate smaller ones in a quasiselfsimilar pattern.
126
 127 zyxwv

F i g u re 1. The Starry Night — Iterates of Tan(z) . A bifurcation diagram for the family of dynamical
systemszyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBAf(z)=■tan(z).
128 S. D. Casey

The pattern was developed in the spring of 1986. The generating algorithm was a
variation of the "escapetime algorithm" discussed in [4], modified so as to handle
iterates of the tangent. The author was influenced by the work of [3] on iterates of A •
2
e . However, this study of iterates of f(z) = A • tan(z) differs in many respects from
2 3 2 2 2
[3]. Since |e | = e * and e is entire, iterates of A ■ e go to infinity with the growth
of 3tzn. However, tan(z) has simple poles at odd multiples of 7r/2 on the real axis, and
is bounded by | coth(y)| off the axis. Thus, an iterate diverges by first being drawn
into a sufficiently small neighborhood of some pole. The "stars" in the pattern
represent how iterates of A • tan(z) spread the original poles throughout the complex
plane.
Calculations for the pattern were performed on a Digital VAX 11/750, with
code written in VAX11 FORTRAN (DEC). The pattern was then produced on a
Raster Technologies Model ONE/80, with the hardcopy produced by taking a
photograph of the screen. The equipment was made available by the signal
processing branch of Harry Diamond Labs. The author wishes to thank the branch
for access to the equipment and Rob Miller of the Labs for his assistance with the
Raster Technologies equipment.

References
1. P. Blanchard, "Complex analytic dynamics", Bull. Amer. Math. Soc. (N.S.) 11, 1
(1986) 85141.
2. S. D. Casey, "Formulating fractals", Computer Language 4, 4 (1987) 2840.
3. R. L. Devaney, "Julia sets and bifurcation diagrams for exponential maps", Bull.
Amer. Math. Soc. (N.S.) 11, 1 (1986) 167171.
4. A. K. Dewdney, "Computer recreations", Scientific American 253, 2 (1985) 16-
24.
Stephen D. Casey
From Asymmetry to Symmetry —
2
Iterates of Exp(z) and Exp(^ )

The two patterns in "From Asymmetry to Symmetry" are a study in con trast.
Figure 1 is a window into an asymmetric world in which one has no sense of
orientation, whereas Fig. 2 is an extremely complex pattern exhibit ing symmetry
with respect to both the x and y axes. The study of these patterns falls under the
iteration theory of Fatou and Julia, which examines discrete dynamical systems on
the Riemann Sphere C with a dynamic given by a nonlinear holomorphic or
meromorphic mapping from fi C C to C (see [14]). Figure 1 was generated by the
function

h{z) = \e\
while Fig. 2 was generated by the function

f2(z) = \e*2,

in A or parameter space. At each pixel position (x, y) = x + iy = A, the orbit of the

initial value 0 was calculated, i.e., for j = 1, 2, zn+\ = fj(zn), z0 = 0. This procedure was
repeated until either 5ft(zn) was greater than or equal to 100, which guarantees that zn
2 z 2
diverges to oo, or the number of iterates exceeded 256. (Since |e | = e^ and e is
z
entire, iterates of A ■ e go to infinity with the growth of %l{zn).) In the picture, the
black regions correspond to points A for which 9?(z256) < 100, while the bands in the
other regions correspond to the number of iterates (modulo 2) needed for 5R(^n) >
100.
Devaney [4] gives us a precise mathematical interpretation of the patterns. They
are bifurcation diagrams for the families of dynamical systems generated by fi{z) = A •
2
exp(z) and f2(z) = A ■ exp(2 ). From the pictures, we can see that neither exp(z) nor
2
exp(z ) is structurally stable (as defined in [4]), because
a slight variation in A can produce a wide variation in the behavior of the iterates.
Devaney also shows that those values of A for which zn diverges correspond to
dynamical systems with dynamics /i, f2 having a Julia set equal to the entire
complex plane. It is also of interest to note that the distribution of the divergent
points forms a fractal pattern, which is seen by observing that the larger "bends"
generate smaller ones in a quasiselfsimilar pattern.
129
130

Figure 1. A bifurcation diagram for the family f\(z) = X ■ ez

Figure 2. A bifurcation diagram for the family /2(z) = A ■ ez.
 Prom Asymmetry to Symmetry ... 131

The pattern was developed in the spring of 1986. The generating algorithm was
a variation of the "escapetime algorithm" discussed in [5], modified so as to handle
iterates of exponentials. The author was reproducing the experi ments of Devaney
2
with iterates of A•e . The symmetry in Fig. 2 was expected. After producing Fig. 2,
3
the author produced a "threefold" symmetric pattern using Aexp(z ). (See [2, 3] for
additional patterns, pseudocode for generating the patterns, and further discussion.)
Calculations for the pattern were performed on a Digital VAX 11/750, with code
written in VAX11 FORTRAN (DEC). The pattern was then produced on a Raster
Technologies Model ONE/80, with the hardcopy produced by taking a photograph
of the screen. The equipment was made available by the signal processing branch of
Harry Diamond Labs. The author wishes to thank the branch for access to the
equipment, Rob Miller of the Labs for his assistance with the Raster Technologies
equipment, and Joseph Comick for his assistance in the preparation of the prints. He
also wishes to thank Professors Alan Brownstein and Richard Kreminski for helpful
mathematical discussions.

References
1. P. Blanchard, "Complex analytic dynamics", Bull. Amer. Math. Soc. (N.S.) 11, 1
(1986) 85141.
2. S. D. Casey, "Formulating fractals", Computer Language 4, 4 (1987) 2840.
3. S. D. Casey, "Fractal images: procedure and theory", HDL Technical Report
HDL TR2119 (1987) p. 42.
4. R. L. Devaney, "Julia sets and bifurcation diagrams for exponential maps", Bull.
Amer. Math. Soc. (N.S.) 11, 1 (1986) 167171.
5. A. K. Dewdney, "Computer recreations", Scientific American 253, 2 (1985) 16-
24.
Stephen D. Casey and Nicholas F. Reingold
A Variation on a Curve of Mandelbrot

The pattern described here was generated by an efficient recursive algo rithm
developed by the authors which produces approximations of selfsimilar fractal sets
(see [3]). Such sets are constructed by a repeated scaling, trans lation, reflection,
and/or rotation of a fixed pattern or set of patterns. The procedure is a "pattern
rewriting system" in which a given geometric pattern is drawn repeatedly after
suitable mappings. The pattern used to generate the fractal set by the rewriting
system will be called a seed. The base is the initial configuration. This particular
pattern was produced by using the "monkey's tree" as a seed with the Gosper curve
a
as a base. The procedure was iterated 4 times. Using their algorithm, the authors
duplicated the selfsimilar fractal sets in Mandelbrot [4], which constitute
approximately 45% of the graphics plates in the book.

The curve C produced by repeating this particular generating process in finitely

is often a selfsimilar fractal. It can be mapped onto itself by appropri ate similarity
transforms, and thus is "selfsimilar". And, if Af(r) equals the minimal number of
closed line segments of length r needed to cover C, then C has the noninteger
fractal dimension D = linv^o wfi/tT ~ 18687, which, by definition, makes it a
fractal (see [24]).
The basic idea of the pattern rewriting system is to simply replace each segment
of the pattern at each level of recursion with an appropriately scaled and rotated
copy of the seed. The routine is a cousin of the string rewriting or L systems
introduced by A. Lindenmayer and developed by P. Prusinkiewicz (see [5]). which
are themselves cousins of LOGO, the language developed at MIT under the
direction of S. Papert (see [1]). However, it differs from these systems in some
important ways. The generating schemes for patterns are designs with some builtin
information on how to orient patterns in later levels of iteration. In Lsystems and in
LOGO, a pattern is generated by "teaching the turtle" where to move based on some
fundamental set of instructions. In pattern rewriting, if you can draw a piecewise
linear pattern on a piece of paper, identifying its vertices and its orientation scheme,
then you can produce

a
See [4], plates 31 and 146 for the former, and plates 46, 47, and 70 for the latter.
132
 133

Figure 1. Pattern generated by using the "monkey's tree" as a seed with the Gosper curve as a base.
134 S. D. Casey & N. F. Reingold

it on the computer. The system is also efficient, in that the only overhead between levels of
recursion is a single boolean variable. In Lsystems, the higher the level of iteration, the
longer the string produced. Pattern rewriting gave the authors an interactive Fractal
Geometry "text", with which they created and explored approximations to classic fractal sets
(e.g., Cantor sets, Peano curves), and reproduced many of the patterns in [1] and [4]. (See [2,
3] for further details and additional patterns.)

The authors would like to thank Larry Crone of American University for mathematical
and programming advice, and our student Nora Wade, who cat alogued all of the selfsimilar
fractal sets in [1, 4]. Calculations were performed on an AST Premium/386, with the
hardcopy produced on a QMSPS 410.

References
1. H. Abelson and A. diSessa, Turtle Geometry: The Computer as a Medium for Exploring
Mathematics (The MIT Press, 1980).
2. S. D. Casey, "Analysis of fractal and ParetoLevy sets: theory and application",
Proceedings of EFTF (1990) 205211.
3. S. D. Casey and N. F. Reingold, "Selfsimilar fractal sets: theory and procedure",
IEEE Computer Graphics and Applications 14, 3 (1994) 7382.
4. B. B. Mandelbrot, The Fractal Geometry of Nature (W. H. Freeman, 1983).
5. H.O. Peitgen and D. Saupe, eds. The Science of Fractal Images (SpringerVerlag, 1988).
Stephen D. Casey and Nicholas F. Reingold
An Asymmetric Sierpinski Carpet

The pattern described here was generated by an efficient recursive algo rithm
developed by the authors which produces approximations of selfsimilar fractal sets
(see [3]). Such sets are constructed by a repeated scaling, trans lation, reflection,
and/or rotation of a fixed pattern or set of patterns. The procedure is a "pattern
rewriting system" in which a given geometric pattern is drawn repeatedly after
suitable mappings. The pattern used to generate the fractal set by the rewriting
system will be called a seed. The base is the initial configuration. The seed for this
particular pattern consisted of three components. The base was a square drawn
counterclockwise. The procedure was iterated 5 times to produce the figure. Using
their algorithm, the authors duplicated the selfsimilar fractal patterns in
Mandelbrot's The Fractal Geom etry of Nature [4], which constitute approximately
45% of the graphics plates in the book. This particular pattern is an original
configuration.
The set Sc produced by repeating this particular generating process in finitely is
often a selfsimilar fractal. It can be mapped onto itself by appro priate similarity
transforms, and thus is "selfsimilar". And, if M(r) equals the minimal number of
closed disks of radius r needed to cover Sc, then Sc has the noninteger fractal
=
dimension D = limr_o °og(i/r) 1^1(3) ~ 14649, which, by definition, makes it a
fractal (see [24]).
The basic idea of the authors' pattern rewriting system is to simply re place each
segment of the pattern at each level of iteration with an appro priately scaled and
rotated copy of the seed. The routine is a cousin of the string rewriting or Lsystems
introduced by A. Lindenmayer and developed by P. Prusinkiewicz (see [5]), which
are themselves cousins of LOGO, the language developed at MIT under the
direction of S. Papert (see [1]). However, it dif fers from these systems in some
important ways. The generating schemes for patterns are designs with some builtin
information on how to orient patterns in later levels of iteration. In Lsystems and in
LOGO, a pattern is generated by "teaching the turtle" where to move based on some
fundamental set of in structions. In pattern rewriting, if you can draw a piecewise
linear pattern on a piece of paper, identifying its vertices and its orientation scheme,
then you can produce it on the computer. The system is also efficient, in that the
only
135
136
Figure [Link].
 An Asymmetric Sierpinski Carpet 137

overhead between levels of recursion is a single boolean variable. In Lsystems, the higher the
level of iteration, the longer the string produced.
The routine gave the authors an interactive Fractal Geometry "text", with which they
created and explored approximations to classic fractal sets (e.g., Cantor sets, Peano curves),
and reproduced many of the patterns in [1] and [4] (see [2, 3] for further details and
additional patterns). Motivation for devel oping the routine was provided by the figures in [1,
4] and the simplicity and elegance of LOGO.

The authors would like to thank Larry Crone of Amercian University for mathematical
and programming advice, and our student Nora Wade, who cat alogued all of the selfsimilar
fractal sets in [1, 4]. Calculations were performed on an AST Premium/386, with the
hardcopy produced on a QMSPS 410.

R e f e r e n c es
1. H. Abelson and A. diSessa, Turtle Geometry: The Computer as a Medium for Exploring
Mathematics (The MIT Press, 1980).
2. S. D. Casey, "Analysis of fractal and ParetoLevy sets: theory and application",
Proceedings of EFTF (1990) 205211.
3. S. D. Casey and N. F. Reingold, "Selfsimilar fractal sets: theory and procedure",
IEEE Computer Graphics and Applications 14, 3 (1994) 7382.
4. B. B. Mandelbrot, The Fractal Geometry of Nature (W. H. Freeman, 1983).
5. H.O. Peitgen and D. Saupe, eds., The Science of Fractal Images (SpringerVerlag, 1988).
Stephen D. Casey
The "Computer Bug" as Artist — Opus 3

The pattern described here was generated by an efficient recursive algorithm

developed jointly by the author and N. F. Reingold. The algorithm is explained
in[l].
The procedure used to develop the pattern duplicated the selfsimilar frac tals
sets in Mandelbrot's The Fractal Geometry of Nature [2], which constitute
approximately 45% of the graphics plates in his book. This particular pattern is the
result of a mistake by the author, and it reminded the author of graphics plates 246
and 293 in Mandelbrot, which are entitled The Computer "Bug" as Artist, Opus 1
(2), respectively. It is of interest to note that this "computer bug" (or, more
appropriately, "programmer bug") actually produced (some thing which looks like)
a bug.
Calculations for the pattern were performed on an AST Premium/386, with the
hardcopy produced on a QMSPS 410.

References
1. S. D. Casey and N. F. Reingold, "Selfsimilar fractal sets: theory and procedure",
IEEE Computer Graphics and Applications 14, 3 (1994) 7382.
2. B. B. Mandelbrot, The Fractal Geometry of Nature (W. H. Freeman, 1983).
L3S
 139

bO

3
9
Edward M. Richards
The Last Word in Fractals

Described here is a pattern which grows fonts or words or fractal shapes by

random walk techniques as described in the next TOPSYTURVY pattern. This
pattern however shows that the seeded points do not have to be regular geometric
figures. To generate this pattern, the word "fractals" was written using a drawing
package. For effect, the word was also underscored with a wiggly line. The word
"fractals" and its underscore were then used as seeded points. Each random walk
was started by choosing a column (0, 1,..., 639) and a row (0, 1,..., 479) at random.
The search would then begin in any one of the four major directions, the distance of
one pixel at a time, looking for seeded points. When a seeded point was found the
searching pixel stuck to it and another starting point was generated and the search
repeated. To prevent overcrowding as more and more pixels aggregated on the
screen, the program only allowed a starting point to occur if it was more than 9
pixels away from any other pixel on the screen. If overcrowding was detected, a
new potential starting was generated. Using this overcrowding protection prevents
the random starting points from occurring too close to the points which are already
aggregated. This helps to maintain the fractal quality of the pattern. Without the
overcrowding DENSITY function in the program, the growth could have lumpy
and solid spots. Seeded points can be made with different colors and an ABSORB
color function can cause searching pixels to take on the color of the seeded pixel.
Thus, clusters of different colors may be grown on the screen. With seeded pixels
broadly outlining anything from fonts to countries, interesting patterns emerge.

I have to thank C. Pickover for the idea for this pattern. The approach shown
here, while growing its own font, is different from his approach given in Algorithm
(July/August 1990). In the article "Growing your own fonts", Pickover actually
grows the font from the skeleton form of the letter outwards. In his approach, a
point on the letter is selected at random and then grown outwards. The pattern
shown here sort of sneaks in from the outside and sticks to whatever it finds. It was
after I read his article that I thought of applying my FRACTAL AGGREGATION
program to a similar task.
140
 The Last Word in Fractals 141

References
1. J. Feder, Fractals (Plenum Press, 1988).
2. C. A. Pickover, Computers, Pattern, Chaos, and Beauty — Graphics From an Unseen
World (St. Martin's Press, 1990).
3. C. A. Pickover, "Personal programs: growing your own fonts", Algorithm 1, 5 (1990) 11-
12.
Edward M. Richards
TopsyTurvy Fractal Growth

Described here is a pattern which was grown using a program developed by the
author. The program generates fractal shapes by random walks of "sticky" diffusing
particles as described by Pickover [2].
In this TopsyTurvy growth, a rectangle is drawn along the perimeter of the
screen using conventional graphics and constitutes the seeded points. The starting
point for the search starts anywhere on the screen as if it were dropped from the
third dimension. To prevent overcrowding, a DENSITY function is used which
rejects a starting point if it is within a certain distance, e.g., 9 pixels, of any pixel
already on the screen.
For this pattern, the random starting points are generated by random angles and
random radii. Each time a pair is generated, these polar coordinates are turned into
XY coordinates. The neighborhood of the potential starting point is then searched for
a distance of 9 pixels in all directions and the point is selected if there are no
neighbors. If there is a neighbor, a new potential starting point is calculated. The
probability function used to determine the radius for this pattern was the uniform
distribution used to select a column (e.g., 0, 1,..., 639 for a 640column screen). The
XY coordinates were tested to see if they fell within the range of the screen. If not,
they were folded back into it. Obviously, other probability functions can be used to
determine the radius.

The pattern shown here is technically not a diffusionlimited aggregation (DLA)

such as those shown in Feder [1] since the density of the growth would have
gradually increased to a solid as new random starting points were selected. This
overcrowding was prevented by the density function.

References
1. J. Feder, Fractals (Plenum Press, 1988).
2. C. A. Pickover, Computers, Pattern, Chaos and Beauty (St. Martin's Press, 1990).
142
143
e
William K. Mason
Symmetrized DotPatterns and
Cellular Automata

Clifford Pickover has experimented with symmetrized dotpatterns, which

resemble snowflakes [1, 2]. Generally speaking, symmetrized dotpatterns are
constructed by placing randomly positioned dots on a plane and then mirror
reflecting the dots to produce a pattern with sixfold symmetry. One difference
between these symmetrized dotpatterns and actual snowflakes is that the lat ter are
connected while the former consist of discreet dots. I have tried two algorithms for
expanding symmetrized dotpatterns so that they become con nected. The resulting
figures are often pleasing to the eye and have some resemblance to actual
snowflakes.
Both algorithms use the symmetrized dotpattern as a starting point for a cellular
automaton [3]. Since a snowflake has six arms, we use sixdot neigh borhoods for the
automatons. If a point has coordinates (x, y) its six neighbors have coordinates (x + 1, y
+ 1), (x + 1, y — 1), (x — 1, y + 1), (x — 1, y — 1), (x + 2, y) and (x — 2, y). In the
first algorithm, a point is "turned on" at time
t if exactly one of its 6 neighbors is "on" at time t — 1. Once a point is on it is never
turned off. In the second algorithm, a point is "on" at time t and only if an odd
number of its six neighbors are "on" at time t — 1.
At time zero, the only points that are "on" are the points in the sym metrized
dotpattern. We keep cycling the automaton until we get a connected figure. We
stop when the figure is most pleasing to the eye. See Figs. 1, 2, and 3 for examples.

References
1. C. Pickover, "Snowflakes from sound", in Computers, Pattern, Chaos and
Beauty (St. Martin's Press, 1990) Chapter 4.
2. C. Pickover, "Unseen worlds", Algorithm 2, 2 (1991).
3. A. K. Dewdney, "Wallpaper for the mind", The Armchair Universe (W. H.
Freeman & Co., 1988).
4. I. Peterson, "Written in the sky", in Islands of Truth (W. H. Freeman & Co.,
1990).
144
 145

Figure 1. Original Pattern — 500 points.

Figure 2. Type 1 Automaton — 3 cycles.

■: t J
^
Figure 3. Type 2 Automaton — 2 cycles.
146 W. K. Mason

Snowflake Algorithm.

VARIABLES sflake, nxtflake are 2dimensional arrays, 100x100 for

example sum
atype is 1 for type 1 algorithm, 2 for type 2 algorithm

Load symmetrized dot_pattern into array sflake(x.y) so that

sflake(x,y)=l if point (x,y) is a dot in the pattern
sflake(x,y)=0 if point (x,y) is not in the pattern

UNTIL key pressed

for x = 1 to 100
for y = 1 to 100
sum = sflake(x+l,y+l)+
sflake(x+l,yl)+
sflake(xl,y+l)+
sflake(xl,yl)+
sflake(x+2,y) +
sflake(x2,y)
IF atype = 1 THEN
IF sum = 1 THEN
nxtflake(x,y)=l
ELSE
nxtflake(x,y)=0
END IF
[end for atype = 1]
IF atype = 2 THEN
IF sum is odd THEN
nxtflake(x,y)=l
ELSE
nxtflake(x,y)=0
END IF
tend for atype =
2] [end for y]
[end for x]
for x = 1 to 100
for y = 1 to 100
IF atype = 1 THEN
IF nxtflake(x,y)=l THEN
sflake(x,y)=l IF atype = 2 THEN
sflake(x,y)=
nxtflake(x,y) [end for y]
[end for x]
Plot all points (x,y) for which sflake(x,y)=l
Erase all points (x,y) for which sflake(x,y)=0
Ian D. Entwistle
"Carpet of Chaos":
Mapping the Bilateral Symmetric
z
z + c in the Complex z and c Planes

The patterns Figs. 1 and 2 are respectively the Mandelbrot and modified Julia
z
sets for the function z — z + c.
n
For iteration of polynomials z + c where n is an integer, the "Mandelbrot set"
(M) consists of those complex numbers z such that the sequence 2 = 0, Zi = z, z2 =
2 2 2
z , z3 = (z + z) ,... , never satisfies \z\ > 2. A number of algorithms for carrying out
this iterative process have been widely used to generate the M set mappings of many
functions which were not polynomials
[1]. Complex functions which are analytic and continuous would be expected to
z
generate Julia sets on iteration. Those mapped from z + c are rather
unattractive. The M set for a function can also be defined as the collection of points
for which the Julia set of fc is connected. Julia sets derived from
points c close to the origin would then be expected to have outlines close to the
z
shape of the lemniscate controlled by the escape test. In the case of z + c, very few
points in the plane are bounded for values of c close to 0 + Oi. The Julia sets derived
from such points are therefore not of interest. A reflection perhaps of the less
attractive properties of the M set. Alternative escape tests
[2]do however yield mappings from the iteration, which are visually exiting [3]
such as Fig. 2. Figure 1 was first obtained by the standard Level Set Method
z
[1] applied to z + c using lines to plot the divergent set points only where adjacent
points differ in divergence rate. Selection of the point c (0.09 + Oi) allowed the
speed of mapping to be halved as the pattern is symmetrical. Even using a language
such as Fortran, which supports complex variables, the iteration process is much
slower than that for simple polynomial iteration. The Pseudo code utilized is listed.
This utilizes the "square" escape test previously discussed [2] and an additional test
to control the point plotting. The data for producing both figures is appended.
147
148

Figure 1.
149
Figure 2.
150 I . D . Entwistle

Appendix
nx , ny, xmin , xmax, ymin , ymax, itermax, c, es, ss
Fig. l. 2400, 2400 -2.0 2 .2 -2.0 2.0 25 0 .09 + Oi 30 35
Fig. 2. 1920 1840 -3.1 1.1 -2.2 2.2 100

Pseudo code for Julia set

Variables: Integer sx, sy, iter Real rz , i z , x, y
Complex c, z
DO sy =1 to nx
DO sx =1 to ny
y=ymin+sy*(ymax-ymin)jny
x=xmin+sx*(xmax-xmin)jnx z =
cplx(x, y)
DO iter=1 to itermax
z=zz + c
r z = rea l(z ):iz=imag( z)

References
1. H.-O. Peitgen and D. Saupe, The Science of Fractal Images (Springer-Verlag, 1986).

2. 1. D. Entwistle, "Julia set art and fractals in the complex plane" , Computers and Graphics 13,
3 (1989) 389--392.
3. 1. D . Entwistle, " ' E nt r a p ped Lepidoptera': An alternative mapping of a Julia Set
for the function z - Z4 - Z + c", in The Pattern Book: Fractals, Art, and Nature, ed . C . A.
Pickover (World Scientific, 1995).
4. C. A. Pickover, Computers, Pattern, Chaos , and Beauty - Graphics From an Unseen
World (St. Martis Press, 1990) .
Ian D. Entwistle
"Spidermorphs": Mappings from Recursion of the
z
Function z ^ z + z* — z + c

The two patterns Figs. 1 and 2 result from studying the Julia set mappings of the
z 4
dynamic f(z) = z + z — z + c and identifying areas of the maps which have
disconnected fractal geometry resembling biological forms.
Two types of computergenerated geometries have introduced the term
"Biomorph". Firstly, in the brilliant discourse on evolution, "The Blind Watch
maker" , the spidery line graphical images generated by recursion of lines were used
to represent evolution of biological forms [1]. At the same time, a col lection of
curious fractal patterns resembling low order biological forms result ing from
iteration of algebraic equations in the complex plane were classed as "Biomorphs"
[2]. For the purpose of describing Figs. 1 and 2, the term "Biomorph" encompasses
organismic morphologies created by small changes to traditional Julia set
calculations [3]. Of particular interest for such pattern generation are the escape
radius tests utilized, as well as additional tests for determining which points to plot
in the complex plane [4]. Among the many reported examples [2, 3] of biomorphs,
z
those derived from iteration of dynamics containing the bilateral symmetric z
appear to most closely resemble known biological shapes [3]. The patterns were
generated using tests previously de scribed [4, 5] and are outlined in the Pseudo
code listed. This utilizes the magnitude of the side of a square lemniscate rather than
the radius of the escape circle. In addition, a limit is put on the points to be plotted
by fixing the maximum reached by zreal or zimag. Figure 1 is symmetrical as its
origin point in the c plane of the function lies on the imaginary axis about which the
c plane mapping is symmetrical. Figure 2 can be observed if the window is extended
to —5, 5, —5,5. Data for generation of both patterns is appended.

Parameters

nx, ny, xmin, imax, l/min, ymax, itermax, c, es, ss

Fig. 1.. 2400, 3000 4 . 1 4.1 4.1 4.1 25 0.78 + Oi 20 25
Fig. 2. 2400 2400 1.31 0.17 1.35 0.13 25 0.1 + i0 20 25
151
152
Figure 1.
 153

Figure 2
154 I. D. Entwistle

Pseudo code
Variables:Integer SX, sy, iter Real TZ, i z, x, y Complex c, z DO sy =1 to
nx
DO sx =1 to ny y=ymin+sy*(ymax-
ymin) / n y x=xmin+sx*( xmax-
xmin) /nx z = cplx (x, y)

DO iter=1 to itermax z
=z z + Z 4 - Z + c
T z = real (z ) :iz= im ag (z )
2
if T z 2 > es or i z > es exit loop
END
2
if T Z < ss or iz2 < ss then print
END:END

References
1. R. Dawkins, The Blind Watchmaker (Longmans, 1986).
2. C. A . Pickover, "Biomorphs: Computer displays of biological forms generated from
mathematical feedback loops" , Compo Graphic Forum 5 (1986) 313-316.
3. C. A. Pickover, Computers , Pattern, Chaos, and Beauty - Graphics From an Uns een World
(St. Martis Press, 1990) .
4. 1. D. Entwistle, "Julia set art and fractals in the complex plane", Computers and
Graphics 13, 3 (1989) 389-392.
5 . 1. D. Entwistle, " 'Entrapped Lepidoptera': An alternative mapping of a Julia Set for the fun
ction z - Z4 - Z + c" , The Pattern Book: Fractals, Art, and Nature, ed. C. A. Pickover (World
Scientific, 1994).
Ian D. Entwistle
"Fractal Turtle" and "Elephant Star": Multiple
5
Decomposition Mappings of f(z) : z —» z + c
in the Complex Plane

5
The patterns illustrate a Julia set derived by iteration of the function z —* z + c
in the complex plane using a nonstandard escape test and additional tests utilizing the
values of z and \z\ at escape to control the point plotting.
Mappings in the complex plane derived from iteration of a wide range of al
gebraic functions have been explored by computer graphical methods in order to
study their fascinating and often beautiful fractal geometry [1]. Of special interest is
the way in which the iterative mathematics produce "chaos" patterns close to the
divergent and bounded point boundaries. For higher polynomial functions such as z
5
—» z + c the mappings are not visually very appealing since the area of chaos is
small and the "quasi" circle outlines which illustrate the different divergence rates
are close to the bounded set outline. The use of alternate divergence tests, especially
those that make direct use of the values of z or \z\, have been investigated [2]. They
frequently enhance the mappings of particular functions. For generation of the
patterns illustrated in Figs. 1 and 2, a number of such nonstandard tests were used.
Binary decomposition has been described in detail [3]. In brief, a complex number
can be considered to represent a vector in the complex plane which has a direction
from the ori gin as well as magnitude. If the angle of this direction is determined for
each point and the point plotted according to whether this angle is > or < 180° then a
pattern of alternate black and white will result. Multiple decomposi tion will be
affected if the 360° angle range is divided into numerous smaller ranges and more
color control is applied to these. Effective use of this method has been described for
illustrating alternative mappings of the Mandelbrot set [4]. A modified version of
the algorithm Binary Decomposition Method for Julia Sets (BDM/J) [5] was utilized
to generate Figs. 1 and 2. For computer languages which do not support complex
5
variables, algebraic substitution [6] z = x + iy and c = a + ib into z + c results in the
two equations (1) and

2
(2) which are iterated using the escape test (zimag) > 10. The angle data is utilized
by calculating the value (in radians) using equation (3). Color values
15:>
156
Figure 1.
 157

Figure 2.
158 I. D. Entwistle

of black and white are then related to the integer angle number associated with each
divergent point. Note that the ratio zimag/zreal uses the data generated in the iteration loop.
To obtain the pattern displayed by the bounded points, the minimum value of \z\ reached at
maximum iteration is converted to an absolute integer using equation (4). Values of 0 and 1
are then assigned to these integers for printing. The data for generating the patterns is
appended. Note that the "turtle" shape of Fig. 1 was achieved by elongating the imaginary
axis.

A p p e n d ix
Equations (1) zreal = x5 — 10*x3*y2 + 5*x4*y + creal
(2) zimag = y5 W*y3'x2 + 5*x*y4 + cimag
(3) Integer angle = INT(ABS(ARCTAN(zimag/2real))) MOD 40
(4) Integer \z\ = INT(Min|z|*120)
Data Pixels zmin, imax, J/min, t/max, Iterations, creal, cimag
Fig. 1. 2400*3300 1.5 1.5 1.5 1.5 50 0.4 0.68
Fig. 2. 2400*2400 1.5 1.5 1.5 1.5 50 0.55 0.6

References
1. C. A. Pickover, Computers, Pattern, Chaos, and Beauty — Graphics From an Unseen
World (St. Martins Press, 1990).
2. H.O. Peitgen and D. Richter, The Beauty of Fractals (SpringerVerlag, 1986).
3. I. D. Entwistle, "Julia set art and fractals in the complex plane", Computers and Graphics
13, 3 (1989) 389392.
4. I. D. Entwistle, "Julia sets: Alternative mapping of polynomial Julia sets", Fractal Report
15 (1991) 26.
5. [2], pp. 4044, 6476.
6. J. D. Jones, "Three unconventional representations of the Mandelbrot set", Com
puters and Graphics 14, 1 (1990) 127, 129.
7. H.O. Peitgen and D. Saupe, The Science of Fractal Images (SpringerVerlag, 1988).

8. F. J. Flanagan, Complex Variables: Harmonic and Analytical Functions (Dover

Publications, 1972).
Ian D. Entwistle
4
"Floral Table": A Mapping of the Function z —> z
— z + c in the Complex z Plane

The circular pattern in Fig. 1 illustrates a mapping of the polynomial fc(z) = c4 — 2 + c in

the complex z plane for which the divergent points in the z plane iteration of z + c were
identified by the iteration value for which \z\ < 2. This definition of the escape radius
implemented by Mandelbrot [1] determines that the curve for iteration value k = 1 is a circle.
Alternative tests for divergence of points in similar mappings have been used [2]. All the
points in the bounded set were iterated to a limit of 150.

Aesthetically pleasing patterns are generated by the mappings of linear polynomials of

the type zn + c in the complex plane and their properties have been widely studied. Many
algorithms for achieving variations in mappings have been reported [2]. Although the
perimeters of the bounded sets for these mappings appear to increase in complexity with
increase in n, the higher rates of divergence note for points close to the bounded
circumference lessens the aesthetic appeal of the resulting images. This is also due to the
tendency as n increases for the geometry of the outline to resemble more of a cirlce. One
alternative approach to obtaining more appealing images from mappings of functions in the
complex plane which has not been widely studied does produce attractive patterns [3]. In the
present example, addition of —z to the function z4 + c does produce a function which is
readily mapped in both 2 and c planes and, as the pattern Fig. 1 illustrates, has a more
attractive "Mandelbrot" set. Other examples of mappings produced by function sum mation
appear elsewhere in this book. The algorithm described in the pseudo code below for Figs. 1
and 2 is a modified version of the Level Set Method [4]. In order to better differentiate the
areas of the pattern mapped by divergent points close to the bounded areas in black and
white, an alternative strategy for plotting the points was adopted. Bounded set points and
those which di verge at rates differing from neighboring points are plotted. In addition, the
circular black area was mapped by printing all the points which diverged af ter only three
iterations. Figure 2, a magnification of the sea horse shapes in Fig. 1, illustrates clearly the
fractal nature of the patterns. The input data is appended.
159
160

Figure 1.
 161

Figure 2.
162 I. D. Entwistle

A p p e n d ix
nx = ny, xmin, xmax, t/min, (/max, itermax
Fig. 1. 2400 1.2 1.2 1.2 1.2 150
Fig. 2. 2400 .686 .665 .3313 .356 150

Pseudo code for modified Level Set Method

Arguments: nx, ny, xvain, raiax, ymin, ymax, intermax Variables:
real x, y, cs integer: ay, sx, iter array: cc(nx)
BEGIN FOR sy=l To ny
cy=ymm + sy' (ym&x ymm)/ny
FOR sx=\ TO nx
c.r=xmin + si*(imax imin)/n.r x=0:y-
0
FOR iter=lTO itermax
;x=x — 0 x j/ +J/ — x + c.r
:y=4'x3'y 4*x*y3 j/ + cy x = rx
: y=zy : cs=x2 + y1
IF cs > 4 THEN exit LOOP to subroutine
END LOOP:PRINT
SUBROUTINE
cc(.sx)=iter
IF cc(sx) cc{sx 1) <> 0 THEN PRINT:END ROUTINE IF iter=3
THEN PRINT:END ROUTINE

References

1. P. Fisher and W. Smith, eds., Chaos, Fractals, and Dynamics (Marcel Dekker, 1985).

2. I. D. Entwistle, "Methods of displaying the behaviour of the mapping z —> c2 + c",

Computers and Graphics 13, 4 (1989) 549551.
3. C. A. Pickover, Computers, Pattern, Chaos, and Beauty — Graphics From an Unseen
World (St. Martins Press, 1989).
4. M. Barnsley, Fractals Everywhere (Academic Press, 1988).
5. H.O. Peitgen and D. Saupe, The Science of Fractal Images (SpringerVerlag, 1988).
Ian D. Entwistle
"Entrapped Lepidoptera": An Alternative
Mapping of a Julia Set for the Function z —> zi
—z+c

4
The pattern is an alternative mapping of a Julia set for the function z —* z — z
+ c which utilizes an escape test |zreai| or |zjm ag | <sqrt 20 to determine the divergent
points in the complex plane.
2
Julia sets for the dynamic f(z) = z + c have been widely studied and many
beautiful images resulting from a boundedness test \zn\ > 2 have been published [1].
Fewer studies have been made of the patterns which emerge from iteration of other
dynamic functions when alternative tests or additional tests are applied. Of particular
interest are mappings of transcendental functions [2] and also studies with alternative
tests on iteration of higher polynomials and mixed algebraic functions [3]. Others
have given rise to some remarkable "biomorph" images [4]. In the pattern illustrated,
divergence is assessed by testing the value of either the real or imaginary values of z.
The lemniscate for iteration value k = 1 is thus squareshaped rather than circular. Both
4
the real and imaginary values of the function z —> z — z + c rise rapidly on iteration
and very few points require more than 20 iterations to reach the escape value. Since
4
this function is a simple sum of z and —z, the standard Level Set Method algorithm
was used to define a programme listing. A modification was used to control the
coloring of the map. This involves the extra subroutine test indicated in the pseudo
code listed. Thus, only points in the complex plane which have values of z at escape
between 20 and 25 are printed. The pattern is symmetrical about the imaginary axis
and so only half the points of the map need to be calculated. The input data required to
generate the pattern is appended. The fractal geometry of the pattern is exemplified by
the appearance of the "gothic butterfly biomorph" shapes at all scales.

Appendix
nx = ny, xmin, imax, J/min, ymax, es, ss, itermax, ex, cy 2400 1.52 1.52 1.52
1.52 20 25 25 .78 0
163
164
Figure 1
 "Entrapped Lepidoptera" : An Alternative Mapping of . . . 165

Pseudo code for modified Level Set Method

Arguments: n x , ny , xmin, xmax , ymin, ymax , itermax, cX,cy, cs ,ss
Variables: real x , y integer SX, sy , iter
BEGIN FOR sy=l TO ny
FOR sx=l TO nx
y=ymin + sy*(ymax-ymin)/ny
x=xmin + sx*( xmax-xmin) /nx
FOR iter=l TO itermax
4 2*y2
z x = x - 6*X + y4 - X + ex z
3
y = 4* x * y - 4*y3 - Y + cy
x =zx :y= zy
2
IF x > es or y2 > e s THEN exit LOOP to SUBROUTINE
END LOOP
SUBROUTINE
2
IF x < ss or y 2 < s s THEN PRINT
END ROUTINE

References
1. H.-O. Peitgen and P . H. Richter , Th e Beauty of Fractals (Springer-Verlag, 1986) .
2. R. T . Stevens, Fractal Programming in C (M&T Books , 1989) .
3. C . A. Pickover , "Chaotic behaviour of the transcendental mapping (z ->
Cosh(z)+/L) , The Visual Computer 4 (1988) 243-246.
4. 1. D . Entwistle , "Julia set, art and fractals in the complex plane" , Computers and Graphics
13, 3 (1989) 389-392.
5. C. A. Pickover, "Biornorphs: Computer displays of biological forms gen erated from
mathematical feedback loops" , Comp o Graph ics Forum 5 (1986) 313-316. 6. M.
Barnsley, Fractals Everywhere (Academic Press , 1988).
S. Dean Calahan and Jim Flanagan
SelfMapping of Mandelbrot Sets by
Preiteration

2
The images described here show how pictures of Mandelbrot sets of fc(z) = z +
3
c and fc(z) = z + c change shape as the points constituting them undergo additional
individual iterations. The canonical Mset computation colors a pixel based on the
escape conditions of the associated point in the complex plane. "Selfmapping" by
preiteration colors the pixel based on the escape conditions of forward points in the
orbit of the point under consideration. That is, for a point c € C, iterate ck times, the
result being c^ = /*(0). Color c based on the escape conditions of c^.

fe
Let Mn = {c £ C: /c (0) » oo as k —> oo}. Mi(= M0) is just the Mandelbrot set for
the function under study (the first two sets in this sequence are the same when, as
customary, the initial z point is considered to be 0+0i). M2 maps the escape
conditions of the first iterates of points in M\ back to the original point. M3 maps the
escape conditions of the second iterates of each point in MQ back to the original
point. Alternately, M3 maps the escape conditions of the first iterate of each point in
Mi back to the point in M2. Thus, the sequence of maps Mo, Mi, M%,... , suggest the
description that each map results from "folding" or "kneading" the points of the
previous map by one iterative step.

The Figures
2 3
Figures 2 and 3 show sequences of preiteration maps for z + c and z + c.
Figure 3 depicts the topmost bud of Fig. 1 : M20 at higher magnification (corners at
—0.4 + 1.14i, 0.14+0.61i). Figure 4 enlarges the entire map shown in Fig. 1 : Mw.

Commentary

An apparent feature of these sequences is that as the number of preitera tions

grows larger, the maps look more and more like M 0. At each step, groups of buds
vanish, replaced by "ghosts" of the original, apparently a kind of dust suggesting
the outline of the original bud in M0 (note especially Fig. 1: M20
166
 167

F i g u r e 1.

F i g u r e 2.
Figure 3. F
igure 4.
168 S. D. Calahan & J. Flanagan

and M30, and Fig. 3). As the process described essentially throws away points of
the original set which escape quickly, this is perhaps not surprising.
Suggestions for further study of this phenomenon immediately suggest
themselves:

• Relating the "ghostbuddutsts" to external arguments generated by binary (or n-

ary) decompositions.
• Animation of these sequences, especially for higher degrees of preiteration. The
discrete nature of the step between maps would probably yield dissat isfying
animations for a low degree of preiteration. Perhaps in the future, the concept of
"partial iteration" will acquire meaning, or another model for this process will be
discovered, that affords a smooth transition between frames.

• Study of the "ghosting" effect. The disappearance of parts of the set indi cates
that the forward orbits of some of the points inside a Mandelbrot set
contain points outside the set. How else might these points be different?
• Preiterated Julia sets. Such sets should be interesting themselves, as well as
helping to illuminate the "ghosting" effect.
• Comparison with vector field maps. This mapping technique emulates philo
sophically the study of vector fields, measuring a kind of "flow" in which points
are carried to other points by the process under consideration. Fur ther
investigations might profitably compare and contrast pictures of the vector fields
of iterative steps with pictures of the kind shown here.

Acknowledgments

A SPARCstation 2 running software under development by coauthor Flanagan

produced these images.
S. Dean Calahan
Unrolling the Mandelbrot Set

These images graph some measurements on components of a binary decom

position of the Mandelbrot set, M. For excellent renderings of binary decom
positions, see [1] p. 74.
Figure 1, a schematic diagram of the binary decomposition, emphasizes its
regular tree structure — a collection of fork bifurcations. The horizontal boundaries
between black and white cells correspond to segments of boundaries between level
sets of M. The vertical boundaries divide differences in the escape conditions of
groups of points in a level set. For points in white cells, the orbit escapes above the
real line; for points in black cells, the orbit escapes below the real line.

Within a level set the number of cells n is a function of the escape time e of
e
points in the level set: n = 2 . Two kinds of boundaries divide the cells within a
level set, in an alternating fashion: the outer "tine" of a fork, crossing the level set
and forking yet again, and the middle tine of a fork, crossing an infinite number of
level sets and approaching the tip of a tendril. At a bifurcation, the outer tines (the
"handles" of new forks) are connected to the left and right crossbars. The middle
tine is always connected to the right crossbar, separated from the left crossbar by a
gap. In a picture of M, few of the tines or crossbars are straight (actually only the
outermost ones): indeed the curving of these forks exposes some of the dynamics of
M.

The Figures

Figures 24 show graphs of measurements performed on a binary decompo sition

of the level set with escape time 7, which consists of 128 cells. Only 64 cells are
depicted due to the mirror symmetry of M. The figures are centered on an axis as
they look prettier that way, and the nature of the patterns seems more apparent.
Thus, the measurements discussed below refer to the upper half of level set 7. In
each case, the leftmost component of the graph repre sents the tine crossing level set
7 at the leftmost tip of M (almost touching the leftmost tip of M), the rightmost
component represents the tine crossing the level set at the rightmost portion of the
set (entering the main cardioid), and the longest measurement corresponds to the
portion of the level set in the
169
170

Figure 1. Sch em atic . F igure 2 . Fork handles .

F igure 3 . Left branch es . Figure 4 . R igh t b r anch es .

F igure 5 . Level set 7 .
Unrolling the Mandelbrot Set 171

midst of the Seahorse Valley.

Figure 2 depicts measurements performed on the outer tines of the forks crossing the
level set. Odd numbered bars (counting from the left) correspond to the lefthand tine, even
numbered bars correspond to the righthand tine. Figure 3 depicts measurements on the left-
hand crossbars of the bifurcations, on the boundary between level sets 6 and 7. Figure 4
depicts measurements of the righthand crossbars of the bifurcations, also on the boundary
between level sets 6 and 7. Figure 5 is a plot of the entire level set with an escape time of 7,
and escape radius of 4. The measurements reflect the relative curve lengths, accurate to
within a few percent. Graphs of the components of other level sets exhibit similar
characteristics: tapering off to the left, almost constant towards the right, and multiply
pinched and noisy in between.

C o m m e n t a ry

The magnitude of the difference between the longest and shortest compo nents of
bifurcations in the same level set is impressive. A visual inspection of pictures of binary
decompositions suggests that the variance is smaller at level sets with escape times less than
7; increasing as escape time increases. A search for regularities in the patterns of local
maxima and minima in these measure ments should prove fruitful. Indeed, over some
intervals in these measurements, local minima occur every fourth measurement. However,
this relation does not appear to be true globally, nor does it seem to hold for local maxima.

Acknowledgments

The data for Figs. 24 was generated by "BudOriented Zooming" software under
development in the author's alleged spare time.

Reference
1. H.O. Peitgen and P. H. Richter, The Beauty of Fractals (SpringerVerlag, 1986).
S. Dean Calahan
Parameter Space of Mandelbrot Sets from
Variation of Coefficients

The images described here show how Mandelbrot sets of the form z =
4 3 2
cz +bz +az change with changes in the ratios a : b,b : c, and a : c. Assuming an
appropriate escape radius, preserving the coefficient ratios preserves the shape of
the set: the magnification required to resolve the main body to a particular size
varies inversely with the length of the coefficient vector.
The Cartesian coordinates of vectors defining a spherical triangular grid on the
unit sphere in the positive octant afford the parameters of the present maps. Figure
1 presents a broad overview of this parameter space; Fig. 2 magnifies the corner of
image 11, the canonical Mandelbrot set.
The image parameters emphasize only the main body of each set, normaliz ing
the height of each to approximately 225 pixels (.75 inch at 300 dpi). These maps
ignore the seemingly disconnected islands emerging some distance from the main
bodies of many of the sets. Attempting to encompass them reduces the image size
unacceptably. Escapetime or potential mappings enhance the islands' appearance,
but in many cases they appear to be isolated regions of high potential rather than
attracting set material.

Figure 1

The images numbered 1, 11, and 15 represent Mandelbrot sets for the
4 1 3
polynomials z + k, z + k, and z + k respectively. Along the lines between them
only two powers of z contribute; all three powers of z contribute to the interior
images. Notice that the corner at image 11 apparently depends more strongly on
small parameter variations than do the other corners.

Figure 2

This map expands the corner at image 11, sharing with it images 7, 11, and 12.
The extraordinary length of image 28 prevents the alignment of adjacent images to
the triangular grid, dramatically emphasizing the apparent higher variability in this
region.
172
 173

Figurigure 1.
174 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Figure 2.
Parameter Space of Mandelbrot Sets from Variation of Coefficients 175

C o m m e n t a ry

This mapping technique affords a large universe of possible explorations: Animation of

the images by small changes in parameters between frames; explo ration of the Mandelbrot
sets at greater magnification; an orderly examination of their Julia sets; mapping other powers
of z; incorporating more powers of z into similar parameter maps; allowing complex
coefficients and exponents.
Any reader presently engaged in writing a fractal zooming program might consider
whether to include a parameter mapping interface that generates fig ures of this kind. Other
shapes also lend themselves as candidates for param eter maps. For example, the program that
generated these images iterates polynomials up to degree five, so to accommodate a fourth
power of z, add another vertex, forming a square with three edges identical to those here, a
new fourth edge, and interior images from the contributions of four powers of z.

Acknowledgments

A Macintosh computer running MandleZot 3.0 (by Dave Platt) produced these images.
Factors of Baloney Jim Flanagan, Bob Hagen, and Ed Osenbaugh invaluably critiqued and
inspired the ideas behind this pattern. Thank you all.
Nicolas Chourot and Vedder Wright
Lizards

Described here is a pattern showing an approximate copy of a famous M. C.

Escher drawing of lizards [1], using the pattern and tileediting soft ware Mosedit®.
The program itself was jointly created by N. Chourot and J. Baracs of Montreal.
Although the drawing could be constructed by hand as Escher did, its creation using
the computer serves as a litmus test of the flexibility and effortlessness of the
graphical interface of Mosedit®.
Every periodic planefilling pattern or tiling must belong to one of 17 pla nar
symmetry groups, which consist of certain combinations of basic maneuvers called
symmetry operations. These operations are only four in number; trans lation (a
rigid sliding motion), rotation, reflection, and glide (a combination of reflection and
translation, like footprints in the sand). To construct his fa mous periodic drawings,
M. C. Escher had to learn the fundamental laws of planar symmetry. The Montreal
team had to learn similar things, combined with tiling theory, in order to develop
the software.
The lizard pattern was initially created from a regular hexagonal cell whose
alternate vertices contain three distinct centers of threefold rotational symme try. A
center of nfold rotational symmetry in the plane is a pivot point around which the
entire pattern (or portion of that pattern) can be rotated to coin cide with itself n
times in one complete revolution. The threefold centers of the lizard pattern are
located at the creature's left cheek, at the right rear knee and at the inside of the left
rear paw.

Reference
1. C. H. MacGillavry, Fantasy and Symmetry: The Periodic Drawings of M. C. Es
cher (Harry Abrams, 1976) pp. 7677.
176
177
Koji Miyazaki and Manabu Shiozaki
FourDimensional Space Flowers

Described here are flowerlike patterns derived from orthogonal projection into 3space of
typical 4dimensional regular polytopes (120 and 600cell) and semiregular polytopes which
are derived from regular truncation around ver tices of the 120 and 600cell polytopes.
These semiregular polytopes have two kinds of regular and/or Archimedean typed
semiregular polyhedral cells, a fixed number of which fit together around every vertex in
fixed order and any two cells have a face in common.
Each of the top row, from left to right, shows a portion of the cell, vertex, face, and edge-
center projection of the 120cell polytope, and each of the bottom row, from left to right, a
portion of the vertex, cell, edge, and face center projection of the 600cell polytope. A vertex-
center projection means a 3dimensional solid model having a vertex at the bodycenter. Each
of the edgecenter, facecenter, and cellcenter projections is similarly defined.
The others are semiregular polytopes or their portions born between the abovementioned
projections of the 120 and 600cell polytopes.
They have the following geometrical characteristics:

1. The outermost vertices lie on concentric spheres, rather than on a single sphere.

2. Their interiors are filled with a small kind of distinct cells, juxtaposed face toface.

3. The volume of a cell is reduced when it occurs in the outer layers of the polytope, and
may even be reduced to zero in the outermost layer.
4. Each vertex has simple cartesian coordinates and the chord factors can easily be
determined.
5. They have many planar polygonal sections determined by the choice of certain vertices,
edges, and faces.
6. They can be stacked in periodic and in aperiodic arrays, meeting one another across
planar polygonal boundaries, or sharing certain cells.
7. Various complicated designs can be continuously and systematically derived according to
changes in the direction of the projection in 4space.
178
179
Istvan Lengyel, Irving R. Epstein,
and Helena Rubinstein
The Dolphin Head Bifurcation Diagram

Described her e is a pattern showing a sequence of period doubling bifurca-tions

toward chaotic motion originating from a stable steady state destabilized by diffusion.

One normally thinks of diffusion as acting to equalize concentration differ-ences in

space. However , as Turing showed nearly four decades ago in a remark-able paper
entitled "The Chemical Basis for Morphogenesis" [1], diffusion can have the opposite
effect. Two reactors or cells in which the same chemical re-action occurs under the
same conditions are coupled through a semipermeable membrane. Without coupling
each reactor has the same unique stable steady state. Because of the mass exchange this
stable steady state may be unstable in the coupled system depending on the dynamics of
the reaction, the ratio of diffusion coefficients, and the coupling strength. The general
conditions for this type of instability are given in [2] .

A model that shows diffusion-induced instability is the Degn-Harrison model,

which describes the temporal behavior of a specific bacterial culture [3]. The coupling
strength C can be expressed as a function of the volume V of the reactors, and the
surface area A and thickness l of the membrane: C = A/(V l). The diffusion coefficients of
the reacting species must differ from one another; in particular, the components that
activate the reaction should diffuse less rapidly than the inhibiting species. The kinetic
equations in this 2-variable coupled model system are:

dXI XIYI
-d = b- Xl - 2 + D xC(X2 - xd
t 1+ qX I
dX2 X2Y2
-d = b- X2 - 2 + DxC(XI - X2)
t 1+ qX 2
dYI XIYI
yC(Y2
-----;}1 = a - 1 + qxi + D - YI)
dY2 X2Y2
-dt = a- 1 2 + DyC(YI - Y2)
+ qX 2
180
F i g u re [Link]: (a,zyxwvutsrqponmlkjihgfedcbaZYXWb,q,D,D)
5 3
(8.951, 11.0, 0.5, 1 0 " , 10~ ) . zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA 2
182 /. Lengyel, I. R. Epstein & H. Rubinstein

where t is the time, x%, yi, x%, yo are the reactants in the two reactors, and Dx, Dy are the
corresponding diffusion coefficients. Without coupling the steady state of the system is a;liSS
= x2,ss = ba, yiiSS = y2,ss = a[lq{ba)]/{ba).
We have chosen parameters at which this steady state is stable and y diffuses faster than x.

To calculate the figure, we take the coupling strength as the bifurcation parameter. If c is
too small or too large, the same steady state is established in both reactors. For small c, the
reactors behave essentially independently, while for very large c they act as if we had only a
single reactor of volume 2V. Between these two limits the mass exchange destabilizes the
steady state and new stable states appear in which the concentrations in the two reactors are
different. For clarity we plot only one branch of the solutions. At points A and B oscillations
appear and, after some period doubling and inverse period doubling sequences, choas is seen.
The apparent symmetry is striking at first sight (especially in view of the logarithmic scale
for c), but closer observation reveals that on the lefthand side there is an extra period
doubling bifurcation.

Diffusioninduced instability requires relatively high coupling strength and a sizeable

difference between the diffusivities of the reactants, conditions which probably cannot be
attained under the standard experimental conditions ap plied in inorganic chemistry. In
biological systems, however, the high sur face/volume ratio and thin membrane walls of cells
and the large differences between diffusion coefficients resulting from the wide range of
molecular sizes and strengths of interactions with membranes make it far easier to realize the
necessary conditions for diffusioninduced instability. It seems likely that nature has already
discovered and made use of the dynamical possibilities of diffusioninduced instability to
generate some of the wide variety of spatio temporal behavior found in living systems.

References
1. A. M. Turing, "The chemical basis for morphogenesis", Phil. Trans. Roy. Soc. B237
(1952) 3772.
2. I. Lengyel and I. R. Epstein, "Diffusion induced instability in chemically reacting
systems: Steadystate multiplicity, oscillation, and chaos", Chaos 1 (1991) 6976.
3. V. Fairen and M. G. Velarde, "Dissipative structures in a nonlinear reaction dif fusion
model with inhibition forward", Prog. Theor. Phys. 61 (1978) 801814.
Bob Brill
Embellished Lissajous Figures

Described here is an openended class of mathematicallygenerated line drawings based on

Lissajous figures, but overlaid and sometimes entirely dis guised with a variety of
embellishments. Lissajous figures constitute a family of curves well known to scientists and
engineers. They can be generated by the following C program:

for (t = 0.0; !closure(); t += Tstep) {

x = A'amplitude * sin(A'frequency * t + Aphase); y = >r
amplitude * sin(Yfrequency * t + Yphase); plot (x, y)\

To generate one of these curves, the user supplies values for the param eters (capitalized
variables above). Lissajous figures are well behaved, being continuous at every point, forming
elegant, sweeping curves that reveal their derivation from the sine function, closing
seamlessly upon themselves, and fit ting neatly into any desired rectangle (whose sides are 2 *
Aamplitude by 2 * ^amplitude). The closure function returns true when the current coordi
nates and heading are the same as the starting coordinates and heading.

Although such plots are handsome, they are much more interesting when embellished, as
shown in the figures. Instead of plotting each point, as above, I have drawn a pair of arcs,
each of which begins at the calculated point. Each arch is determined by three user supplied
parameters that specify the angle offset from the current heading, the degree of curvature
(e.g., 360 is a circle, 90 a quarter circle, 1 a straight line), and the length of the straight line
segments composing the arc. One arc is drawn counterclockwise and is offset to the left of the
current heading, while the other is drawn clockwise and is offset to the right.

Many interesting figures can be drawn by varying the Lissajous parameters and the
embellishment parameters, as described, but many other variations are also possible by
altering the Lissajous equations or by changing the nature of the embellishment.
183
184 B. Brill

One fruitful idea was to change the heading before drawing the arcs. As each
point is calculated, the heading is determined to lie in the direction of the undrawn
line connecting the previously calculated point to the newly cal culated point. Thus
the embellishment follows the path of the curve. We can make the embellishment
rotate in various ways by changing the heading after each recalculation as follows:
heading = 360 — heading or heading = 360 * | sin(t) | — heading or heading = 360 * |
sin(i)| or even heading = |sin(£)|. This by no means exhausts the possibilities. The
embellishment parameters can also take | sin(t)| or 1 — | sin(t)| as multipliers, causing
the off set angles, curvature or line lengths to vary periodically between 0 and their
specified values. Any of these modifications may be in effect separately or in
combination.

Why stop at drawing two arcs? I have also drawn polygons, parameter ized for a
number of sides and side lengths. These also have been modified to cause the
polygons to expand and contract in size. As for changing the Lissajous equations,
there are many possibilities. I like the one where after the calculation of x, I add the
line:

x * = sin(t).

When Yfrequency is 1 and Yphase is 0, this causes the value of x to be 0

whenever y is 0. The effect is like throwing a belt over the design and cinching it up
tight across the waistline (i.e., the x axis).
Here is another such modification whose effect I invite you to discover.
Initialize y to 0 outside the loop, then change the equations as follows:

x = Samplitude * sin(Xfrequency * t + Xphase) — y * sin(t);

y = Yamplitude * sin(Yfrequency * t + Yphase) + x * sin(t);

Space does not permit a complete description of everything I have tried. The
interested reader is encouraged to explore new variations.
185
i 86
I
187
2
(Landscape) zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
189
190
191
192
193

c
a
►3
S. J. Abas
Islamic Pattern

Unlike the arts of other cultures, Islamic art sets out deliberately to shun
anthropomorphic forms and concepts. It was led to concentrate on the explo ration
of symmetry in geometrical patterns, an enterprise which resulted in an
extraordinarily large, complex and elegant collection of patterns [1, 2, 3, 4]. Apart
from their aesthetic merit, these patterns offer a treasure house of symmetry which
make them of great interest to a large number of scientists and educators [5, 6].

The most sophisticated Islamic patterns rely on the use of concealed polyg onal
grids and the pattern depicted here is of this construction. It is rather an unusual
example, for it manages to combine sevenrayed stars with squares and octagons. It
is also a deceptive one, in that at a cursory glance it seems to contain trilobed
symmetric regions (such as the two on the opposite sides of the center point) which
on careful examination turn out not to be so.
We give below an algorithm for its construction which is simpler than the one
published previously by Hankin [7].

1. Draw a grid of heptagons as shown in Fig. 1(a). This gives rise to a series of
small squares si.
2. Refer to Fig. 1(a) again. Centered on each of the squares si, draw a circle cl
circumscribing si and a square s2 of side d, where d is the smallest distance
between two of the nodes on the grid. From the nodes surrounding the squares,
draw lines 11 and 12 to the vertices of the heptagon which lie on the edge
forming the squares si.
3. Refer to Fig. 1(b). Use cl to draw the octagon shown then discard cl. Use
11 and 12 to cut off s2. Replace the two lines with the circle c2. From a point on
the circumference of c2, draw line 13 to a vertex of si. The intersection of 13,
with line 14 joining two of the nodes, defines the radius of the circle c3.

4. Figure 1(c) shows how the pattern emerges by symmetrically performing the
above steps in the region surrounding one of the squares si.
5. Figure 1(d) shows the same when the same operations are continued on a larger
region of the grid.
194
 195

(a) (b)

(c) (d)

Figure 1. An unusual Islamic pattern containing sevenrayed stars, octagons and squares.
196 S. J. Abas

References
1. J. Burgoin, Arabic Geometrical Pattern and Design (FirminDidot, 1879 and Dover,
1973).
2. K. Critchlow, Islamic Patterns: An Analytical and Cosmological Approach
(Thames and Hudson, 1976).
3. I. ElSaid and A. Parman, Geometrical Concepts in Islamic Art (World of Islam
Festival Publ. Co., 1976).
4. D. Wade, Pattern in Islamic Art (Cassell & Collier Macmillan, 1976).
5. J. Niman and J. Norman, "Mathematics and Islamic art", Amer. Math. Monthly 85 (1978).

6. E. Makovicky and M. Makovicky, "Arabic geometrical patterns — a treasury for

crystallographic teaching", Jahrbook fur Mineralogie Monatshefte 2 (1977).
7. E. H. Hankin, "Some difficult saracenic designs", Math. Gazette 18 (1934) 165168.
Paul D. Bourke
Swirl

The following is a means of drawing computergenerated swirling tendrils. The

technique uses the following to generate a series of x, y points given any initial point
xo, j/o and four constants an, 0,12, 021, 022

xn+i = sin(an2/n) cos(ai2zn)

yn+1 = sin(a2ix„) cos(a22t/„)

To create an image, each point Xi, yi, after ignoring the first 10 terms in the
series say, is drawn as an "infinitely small" point on the page. The resulting image
shows all the possible coordinates the series can generate, that is, the attractor of a
chaotic system. Any initial point XQ, J/O (except for some rare special cases) gives
the same set of coordinates and hence, the same image but in a different order.

The examples provided were created using the following parameters, the images
are numbered left to right, top to bottom.

an ai2 <*21 0.22

1 2.70 0.08 0.86 2.20

2 2.24 +0.43 0.65 2.43
3 +2.00 1.00 1.00 2.00
4 +0.44 1.22 +2.50 1.50

It should be noted that the vast majority of values for an, ai2, 021, to 022 do not
yield interesting images, in such cases the attractor consists of only a few isolated
areas.

Acknowledgments

Attributed to Peter de Jong, Leden, Holland by A. K. Kewdney in Com puter

Recreations, Scientific American.
Software called FRACTAL to generate these images and many other fractal and
chaotic systems is available from the author for the Macintosh II family of
computers.
197
198
Swirl
Craig Cassin
Circlegraph 7.11

Described here is a pattern showing a combination of many circles. There are

fourteen rows of circles which emanate from the center like petals in a flower. The
twenty circles in each row are different in radius, line thickness, and position. This
pattern was produced by a program, Circlegraph, written by the author and printed
on an HP laserjet at 300 dots per inch.
scries: 7.11
roW3_oF_c[rcles: 14
ciicles_in_niw: 20
inner_c(icle_line_widllL 0 J
ouler_circ]c_line_widili: 2
eirclc_spicinE: 70
inneT_t:ilclt_tadiusi 1.8
auler_circle_Tndiu5: L
199
Craig Cassin zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFED
F l ow 2.8

Described here is a pattern showing a computer simulation of a flowing fluid.

Thirteen points were selected, then the computer began to plot lines from left to
right. The lines tend toward the closest point which was still to the right. Once that
point is passed, the lines tend toward the next point. Moire patterns arise because
the lines, which are close together, have ragged edges. This pattern was produced
from a program, Flow, written by the author
and printed on an HP laserjet at 300 dots per inch. zyxwvutsrqponmlkjihgfedcbaZY
200
John D. Corbit
Botanical Biomorph Arising from the
3
Iteration of z —* z + \i

The pattern described here is a biomorph image of the type first described by
Pickover [1, 2, 3]. The pattern arises from the iteration of a very simple expression,
3
namely, z —► z +fi, where // and z are complex numbers. The image was rendered
with the aid of a Julia set algorithm and the special convergence criterion required to
reveal biomorphic forms [1]. According to this criterion, z is taken to be convergent
and a point is plotted if either the real or the imaginary part of z is small after many
iterations.
3
Figure 1 shows a map resulting from the iteration of z —► z + /j, in a region of
the complex plane centered at real.z = 0.685426, imaginary.,? = —0.175747, and with
real and imaginary side lengths being [Link] = imagi [Link] = 0.04. The real and
imaginary parts of n were 0.5 and 0.0, respec tively, and the number of iterations was
100.
Qualitatively, the pattern has a plantlike character. Overall, there is a central
upheaval of frondlike foliage flanked by two botanic spirals of infinite depth. The
leafy fronds are seen by perceiving the white areas as figure and the black areas as
ground. Alternatively, spiky hairlike forms are prominent when one perceives black
as figure and white as ground.
The computations were performed on a Macintosh II using Fractal 2.4 soft ware
written by Paul Bourke of Auckland University, Auckland, New Zealand. Four 450
x 450 pixel quadrants of the image were computed separately and then assembled in
a graphics package giving a final image resolution of 900 x 900 pixels.

References
1. C. A. Pickover, "Biomorphs: Computer displays of biological forms generated from
mathematical feedback loops", Computer Graphics Forum 5 (1986) 313316.
2. A. K. Dewdney, "Computer recreations", Scientific American (July 1989) 110-
113.
3. C. A. Pickover, Computers, Pattern, Chaos, and Beauty — Graphics From an
Unseen World (St. Martin's Press, 1990).
201
202
Figure 1. A zoom on a region of the map for Z —> Z3 + fi. Note frondlike forms in the central area and
the two spirals at the sides. See text for more information.
W. H. Cozad
Transient Microstructure

Described here is a pattern formed by transient values generated when the

iterative logistic formula NEWX = K * OLDX * (1OLDX) is repeatedly mapped
over certain very small K intervals. The K interval for Transient Microstructure was
3.699253.7019. This interval is within the domain of the chaotic attractors, but I
have found no clear relationship between chaos and the microstructure. The pattern
varies with the value initially assigned to OLDX; here, it was 0.3.

The standard algorithm for the logistic map displays only attractor val ues.
Screen output begins after 100 or 200 iterations when, in most cases, the production
of transients has ceased and the process has fallen into the attractors' embrace.
Mapping the transients' structure requires modifying the algorithm to display the
output from each iteration. Interestingly, this modi fied algorithm will generate a
macrostructure of transients if a larger K interval is used. Image A, below, shows
the standard logistic map for the K interval 14, while image B includes an overlay of
the transient macrostructure. The form of the macrostructure, like that of the
microstructure, depends on the value initially assigned to OLDX.

The standard algorithm also must be modified by greatly reducing the to tal
number of iterations. The transient microstructure's fragile filagrees are obscured if
that number significantly exceeds the 25 iterations used for Tran sient
Microstructure. I have also excluded the results of the first 13 of these iterations as
they form relatively straight lines cutting horizontally across the filagrees.

Mapping transients to produce striking and significant patterns has been

developed by C. A. Pickover and is discussed in several works by him. See,
Computers, Pattern, Chaos, and Beauty — Graphics From an Unseen World
(St. Martin's Press, 1990); "Personal programs: Close encounters with strange
attractors", Algorithm 1, 2 (1990); and "Graphics, bifurcation, order and chaos",
Computer Graphics Forum 6, 1 (1987).
203
204

Transient microstructure
Douglas A. Engel
Engel Curves

The patterns described here show how Engel curves are generated and some
interesting computergenerated Engel curve circuits. These curves were published in
the February 1983 American Mathematical Monthly. In Fig. 1, a curve is generated
by first creating a grid of m rows of n points, m and n coprime and only one of m
and n even. Figure 1 has a 2 x 3 rectangular grid of 6 points, the grid then being
divided into 3 pairs of points, each pair connected by a single line to form a net of
valence 1. This pattern is then repeated 6 times in 3 x 2 array. The same array is then
replicated, turned 90 degrees and superposed over itself to form a net of valence 2
composed of 1 or more closed curves. Patterns that form a hamilton cycle, a single
closed curve using every point of the net like this one are not very common as the m
x n pattern increases. Several investigators showed that only 42 of them exist for 3 x
4 nets, none symmetrical. In what follows, the m x n net is called the efactor and the
final pattern the eproduct.

Engel curve sequences can now be described as shown in Fig. 2. The infinite
sequences of efactors shown here produce only hamilton cycles. Two distinct curve
networks can be achieved by multiplying an assymetrical efactor with itself in 2
ways. If one allows the efactors to be turned over then 2 different assymetrical
efactors can be multiplied to get 4 different eproducts. Several investigators have
attempted to describe these curves and elucidate their prop erties in terms of
sequences, knots, topology, number of curves produced, and so on.

Properties that can be investigated include the simulation of Brownian motion as

a random ecurve that is traced by a plotter, the circuit properties of intertwined
ecurves, 3 and higherdimensional Engel curve products where 3 or more edges meet
at each vertex, the generation of fractallike circuits as a sequence approaches
infinity, and many other properties too numerous to mention here.

The final figures shows a 28 x 9 inverse frame curve (it is called an in verse
frame curve if m is larger than n) surrounding a 10 x 11 perfect frame curve. This
shows how two efactors made according to the same rule produce wildly different
eproduct curves. The 10 x 11 frame curve is the first member
2U5
206 D. A. Engel
of an infinite family that repeats at 10 x (11 + lOn), n = 0, 1, 2, ... . The
method of generating the efactors of frame in curves is shown by the 4x5 pattern
Fig. 3. Only 3 other infinite families of perfect frame curves are known.
 207

Fig. 2.

Fig. 1. Fig. 3.

Fig. 4.

Figures 14 show the generation of Engel curves, curve sequences, and an inverse frame curve
surrounding a perfect frame curve.
Ian D. Entwistle
"Spinning Chaos": An Inverse Mapping of
the Mandelbrot Set

2
The pattern illustrates a mapping of the quadratic fc(z) = z + c in the complex
2
\/z plane. Iteration of z + c for the divergent points in the 1/z plane was controlled
by the standard escape test \z\ < 2. An alternative test was used to differentiate the
points in the bounded set.
In order to obtain the inverse Mandelbrot(M) set, all the points in the plane need
to be inverted. The inverted M set contains the those complex numbers such that the
2 2 2
sequence z = 0, z\ = z, z2 = z , zZ = (z + z) , zn,... , never satisfies \z\ > 2 when the
2
function z (n — 1) + 1/c is iterated. The position of an inverse point in the complex
plane is given by the relationship c * cinverse=l(l). Using complex algebra(2)
2 2 2
substitution by c = a + ib gives the relationship c * cinverse = a/(a + b ) — ib/(a +
2
b ). Replacement of a and b in the LSM algorithm by the new values of parts of c
then allows calculation of the inverse \z\. In order to create the artistic spinning
effect in the pattern, the value of the real part of \z\ corresponding to the minimum
value of the imaginary part of \z\ is used to control the print color. Note that the
divergent point color is controlled by the iteration value at escape. Those points in
the bounded set which are printed white because of the rounding down by the
integer function are further controlled by the coloring routine for the divergent
points. Adjusting the value of the Integer Factor can be used to further control the
variation in the pattern. Values of the data to produce the pattern are appended.
208
209
210 I. D. Entwistle

P s e u do Code for Modified Level Set Method

Arguments: ny, nx, xmin, xmax, ymin, ymax, itermax

Variables: real x, y, x2, y2, cs, ci integer:sy, sx, iter

BEGIN
FOR sy = 1 TO ny
cy = ymm\sy * (ymax.—ymin)/ny
FOR sx = 1 TO nx
ex = xmin+sa; * (xmax—xmm)/nx
2 2
bxy = l/(cx + cy ) : x 0 : y = 0 : xy = 0 : x2 = 0 : y2 = 0 : ci = 25 FOR iter=l TO
itermax
x = x2 — y2 + ex /bxy : y = xy + xy + cy/bxy
x2 = x * x : y2 = y * y : cs = x2 + y2 : xy = x * y IF y2 < ci
THEN ci=x2
IF cs > 4 THEN exit LOOP and SELECT COLOR
END LOOP
d=INT(ci * INTEGER FACTOR) TEST ci
AND PRINT
SELECT COLOR ON BASIS OF iter VALUE
END LOOP:END LOOP

Appendix

nx=1080, ny=2250, a;min=5,6, a;max=4.1, ?/min=3.6, ymax=1.8, itermax=120,

INTEGER FACTOR=100

References
1. R. Dixon, Mathographics (Basil Blackwell Ltd., 1987).
2. F. J. Flanigan, Complex Variables: Harmonic and Analytical Functions (Dover
Publications Inc., 1972).
3. H.O. Peitgen and D. Saupe, The Science of Fractal Images (SpringerVerlag, 1988).
Ian D. Entwistle
"Wings of Chaos": Mappings of the Function sinh(z)
+ c in the Complex Plane

The two figures illustrate patterns which are obtained when f(z):— sinh(z) + is
iterated in the z plane. They differ markedly from the full mapping of points which
diverge and of those which belong to the bounded set (the Mandelbrot set) for the
2
polynomial z + c derived patterns. Such patterns have been inten sively studied and
frequently generated [1]. The similar mappings for the tran scendental function
sinh(,2:) +c has received little attention. More recently, the strange behavior of
mappings of the transcendental hyperbolic cosine function has been studied in both
the z and c complex planes [2]. The dearth of reports on the z plane iterative
mappings of the hyperbolic sine function prompted a study of this mapping which
produces the two patterns, see Figs. 1 and 2. The Taylor expansion of sinh(z)
contains only terms of uneven powers of z and so under recursion it was thought
that the bounded set would not map to an outline shape similar to the topology of
2
the Mandelbrot set for z + c even if an approximate expansion was used.

The outlines of the patterns in Figs. 1 and 2 approximate to the topology of the
stable set for sinh(z) + c. Unlike the outline geometry of the stable set for cosh(z) + c,
it is not sensitive to changes in the escape radius value. The circlelike leminiscates
corresponding to increasing iteration values resemble more closely the shape of the
bounded set and so the patterns generated on magnification are not so pleasing as for
those of other functions on iteration in the complex plane. The behavior of the points
"inside" the bounded set for sinh(z) + c can, however, be mapped utilizing various
tests to give visually exciting patterns which are quite distinctive as illustrated by the
z z
two exam ples. The relationship sinh(z) + c = 0.5 * (e — e~ ) + c can be simplified
by substitution, with z = x + i * y using standard complex number algebra [3] to yield
x x
the two equations zreal = 0.5 * (e~ — e ) * cos(y) + creal
x x
and ,dmaginary=cimaginary—0.5 * (e + e~ ) * sin(y). These equations simplify
computation when used in the Level Set Method [4] if languages which sup port
2
complex variables are not available. The test for boundedness (|.z|) = 10 was used in
these mappings and all the divergent points were left unprinted. For Fig. 1, the
2
minimum value of (.zreal) was used to determine whether the
211
212 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA

I zyxwvut

I
I
!

I
I H
F i g u re [Link](z)+usinminimuvalue|z|)2-zyxwvutsrqponmlkjihgfedcbaZY
CO
214 /. D. Entwistle

point was printed black or white. The minimum value of {\z\)2 reached during iteration was
similarly used to control the printing of the stable set of points in Fig. 2. Of particular note
when computing the values of z is the symmetrical shape of the stable set. This allows the
mapping to be achieved with calculation for only one quarter of the total pixel count. To
allow classification of the minimum values on a size basis the real numbers were converted
to integer equivalents by an integer factor (IFF). In both mappings, groups of (zreal) 2 or (\z\)2
values were printed alternately in black and white. Data to allow generation of the patterns is
appended.

A p p e n d ix
IFF Max. Iteration Escape radius Minr, Maxr Mini, Maxi
Fig. 1. 80 200 10 1.6 1.6 3.2 3.2
Fig. 2. 50 200 10 1.55 1.55 3.2 3.2

References
1. H.O. Peitgen and D. Richter, The Beauty of Fractals (SpringerVerlag, 1986).
2. C. A. Pickover, "Chaotic behaviour of the transcendental mapping(z — cosh(z) + u)",
The Visual Computer 4 (1988) 243246.
3. C. A. Pickover, Computers, Pattern, Chaos, and Beauty — Graphics From an Unseen
World (St. Martins Press, 1990).
4. F. J. Flanagan, Complex Variables: Harmonic and Analytical Functions (Dover
Publications Inc., 1972).
5. H.O. Peitgen and D. Saupe, The Science of Fractal Images (SpringerVerlag, 1988).
Ian D. Entwistle
"Islands Among Chaos": Mappings of the
Transcendental Function cosh(z) + c
in the Complex Plane

The illustrated patterns result from the iteration of f(z) :—> cosh(z) + c for
complex z and c planes. Previous studies of this function at iteration have led to
important observations [1] about the morphology and behavior of the mapping. The
bounded set is a single unconnected cardoid when the con vergence test \z\ < 2 is
applied. For higher values of the escape radius this set is progressively distorted
2
until it only resembles the z + c bounded set for points in the plane < (—2, 0). The
shape of the leminiscates are then no longer circlelike and the mapping becomes
periodic (2iri). The more complex dynamics of the cosh(z) + c iteration are therefore
only realized by mapping with a large escape radius. Of particular interest in studies
of this mapping is the possibility that the main cardiod centered at —0.14, Oi is
connected to all the points in the bounded set as has been established [2]
n
mathematically for the polynomials z + c maps. In these mappings the morphology
of the central cardoid is retained by the island miniatures, thus giving a high degree
of self similarity to the fractal geometry of the maps. Greater variations in the
patterns mapped to divergent points and the geometrical shapes of the islands which
form the bounded set are observed for the function cosh(z) + c. Figures 14 illustrate
these variations. The central cardoid is the largest is land cardiod on the 0 axis and
2
has the geometric outline characteristic of the z Mandelbrot set. It appears to be
connected along the imaginary axis and along the spines emanating out to the
pattern edge or ultimately to the periodic boundary. The other miniature cardoids,
one of which is illustrated in Fig. 2 in the celllike parts of the pattern are
unconnected. This is similarly observed at the higher magnification of Fig. 2.
Figures 3 and 4 illustrate the quite different behavior of the islands of the bounded
set mapped in the area of the plane with positive imaginary coordinates belonging to
the region where the main cardoid is extended. Figure 3 illustrates the nonquadratic
behavior resulting in an unsymmetrical cardiod morphology. Connection to other
parts of the bounded set is possible but awaits mathematical proof. The behavior of
the
215
216 /. D. Entwistle

divergent points in this region of the map is also strikingly different, producing the
petallike pattern in Fig. 3. Even more distinct is the chaotic morphology of larger
islands in the bounded set. Figure 4 maps only the bounded set and shows the
changes in morphology of the bounded set with considerable loss in self similarity.
Where fractal geometry is evident, connection has not been es tablished. The
appearance of the miniature cardiod centered at 1.846, 1.60i in the lower left corner
of the pattern suggests that some quadratic behavior still persists. The relationship
2 2
cosh(z) = 0.5 * (e 4 e~ ) can be simplified to two equations for computation by the
x _ x
substitution of z = x + iy. These equations zreal = 0.5 * (e + e ) * cos(y),
x x
^imaginary = 0.5 * (e — e~ ) * sin(y) were iterated using published pseudo code [3]
for the Level Set Method. Taylor's expansion of cosh(:r) can be approximated to l +
2
(x )/2 in order to speedup the iteration. For Figs. 13, divergent points close to the
boundary of the bounded set were left uncolored in order to outline more clearly the
bounded set. In Fig. 4 only points in the bounded set were printed. Data to produce
Figs. 14 is appended and all the patterns were produced using a grid 900 x 1680
points.

jendix
Figure Max. iteration Escape radius Mini Maxi Minr Maxr
1 250 1000 0.002 0.002 3.4005 3.3985
2 300 1000 1.187 0.976 3.9 3.695
3 400 1000 1.325 1.475 0.704 0.854
4 200 1000 1.35 1.85 1.5 2.0

References
1. C. A. Pickover, "Chaotic behavior of the transcendental mapping (z —> cosh(z)
+ //)", The Visual Computer 4 (1988) 243246.
2. A. Douady and J. H. Hubbard, "Iteration des polynomes quadratique
complexes", CRAS Paris 294 (1982) 123126.
3. H.O. Peitgen and D. Saupe, The Science of Fractal Images (SpringerVerlag,
1988).
217
Figure 1. Island Mandelbrot set.
218

Figure 2. Isolated bounded set island.

 219

Figure 3. Unsymmetrical connected cardoid island.

220

>*£ I

F i g u re 4. Variations in morphology of bounded set islands.

Ian D. Entwistle
Alternative Methods of Displaying the
2
Behavior of the Mapping z —> z + c

2
The patterns illustrate alternative behavior of the mapping fc(z) = z + c in the
complex plane. Mandelbrot's study [1] of this function stimulated the publication of
numerous mappings showing their beautiful fractal patterns [2]. Popularising the
mapping algorithm [3] has resulted in a plethora of studies of the iterative properties
of other mathematical functions [4]. Alternative tests for divergence have been used
to produce many new pattern variations [5]. Commonly, the divergence of the
modulus of z was used to control the pattern colors but more recently, the minimum
size of real or imaginary values of z have been used to obtain novel mappings of
various functions [6]. Other fac tors can influence the output of a complex function
iteration. Both the choice of algorithm and computational language can affect a
variation in a mapping. In many studies, avoidance of the effects of computation
requires extended precision arithmetic and in others the use of scanning algorithms
such as the Limited Set Method [7] or Escape Time Algorithm [8] since precision
map pings need iteration for every point in the plane. For low power functions of z
being mapped over relatively large areas of the plane, algorithms such as the
Distance Estimator Method [9] can be used. Faster but less precise mappings can be
achieved using modified image processing algorithms such as Mariani's [9].
Implementation of the algorithm in the native CPU code is favored for fast iteration.
Where speed rather than accuracy is required, integer arith metic can be helpful
since it requires fewer CPU operations. Figures 1 and 2 were mapped using integer
arithmetic and illustrated an effect caused by slight miscoding. Since the
modification can be controlled, it represents an alterna tive mapping of the function
2
z + c. In the integer method, the floating point values are constantly adjusted to
integer values which are within the normal single precision limits for integers. This
process can produce rounding errors which result in the Figs. 1 and 2 differing from
standard mappings and they may not be similarly reproduced on other CPUs.
Utilizing the assembler code on the ARM2 32 bit RISC processor is particularly
advantageous for integer maths since this CPU can perform large arithmetic shifts in
one operation. Although careful control of these shifts was maintained, the escape
radius test
221
222 /. D. Entwistle

introduced an unforseen problem. In order to affect the divergence test the value of the
escape radius 4 was multiplied using a single arithmetic shift 2 27 before comparison with the
value of ^imaginary2 obtained directly from the integer iteration. This approach appears to
result in anomalous escape rates. When the assembler code is altered so that the integer value
of the imaginary part of \z\2 is divided by 227 , using a logical right shift before a comparison
with the escape value 4, the mapping resembles that produced using conven tional coding.
The petal shaping [6] of the divergentmapped curves in the patterns results from the use of
only the imaginary part of \z\2 and not from the inaccuracy of the arithmetic. Data used to
create the patterns is appended.

A p p e n d ix
Minr, Maxr Mini, Maxi Max. iterations Pixels x, y
Fig. 1. 2.0, 1.0 1.5, 1.5 1024 1280, 1960
Fig. 2. 1.786623, 1.785227 0.00095, 0.00095 1024 1280, 2600

References
1. B. B. Mandelbrot, The Fractal Geometry of Nature (W. H. Freeman, 1983). 2. H.O.
Peitgen and D. Richter, The Beauty of Fractals (SpringerVerlag, 1986).
3. A. K. Dewdney, "A computer microscope zooms in for a look at the most complex object
in mathematics", Scientific American 255, 8 (1985) 812.
4. C. A. Pickover, Computers, Pattern, Chaos, and Beauty — Graphics From an Unseen
World (St. Martin's Press, 1990).
5. I. D. Entwistle, "Julia set art and fractals in the complex plane", Computers and Graphics
13, 3 (1989) 389392.
6. I. D. Entwistle, "Methods of displaying the behaviour of the mapping z —> z2 +u",
Computers and Graphics 13, 4 (1989) 549551.
7. H.O. Peitgen and D. Saupe, The Science of Fractal Images (SpringerVerlag, 1988).

8. M. Barnsley, Fractals Everywhere (Academic Press, 1988).

9. R. Silver, "Mariani's algorithm", Amygdala 4 (1987) 35.
 223

Figure 1. Mapping of the full Mandelbrot(M) set using integer arithmetic.

224

i
73
A
a

A.
A.

o
<
v
tc

'S
ho
S3

4)
U
A
be
Ian D. Entwistle
"Serpents and Dragons": A Mapping
of f(z) —> sinh(2;) + c in the Complex Plane

The pattern shows the fractal features of the function z —* sinh(z) +c when it is
mapped in the complex plane by the iteration of z at a constant c.
2 z
For real values of z, the curve for f(z) = sinh(z) = 0.5* (e — e ) is a double
nonperiodic exponential curve without a turning point. If it is iterated in the c plane,
the resulting map is also nonperiodic unlike the similar cosh(z) + c map [1]. For zn =
f(zn— 1, c), n = 1, 2, 3, ... oo the lemniscates n = 1, 2, ...
are circlelike but morphologically unlike the familiar Mandelbrot set shapes of the
n
polynomials z + c or that of the bounded set of cosh(z) + c. The maps of sinh(z) +
n
c in the z plane are also quite dissimilar to the much studied Julia sets of z + c and
2n
have not been studied in much detail. The expansion of sinh(z) contains only z + 1
2
terms and so cannot under iteration become approximated to z as cosh(z) appears
to. The greatly different behavior on iteration may result from this difference. For
computation with languages which do not support complex variables sinh(z) can be
divided into real(a:) and imaginary(y) parts by substitution with z = x + iy using
complex algebra [3] for use in the Pseudo Code.

The pattern illustrates several aspects of sinh(z) +c behavior. The Julia set points
are mapped using the minimum value of \z\ reached during iteration to control the
printing color [4]. The divergent points which map the "serpents" have an escape
value for \z\ > 1000. An additional test was then applied to separate the points with
absolute values of real z or imaginary z < 100. The other divergent points were then
mapped with the opposite color. The pattern contains three distinct types of fractal
geometry, namely, the "serpents" formed from the divergent points, the "dragons"
mapped by the bounded points and the interior points of the "dragons". The periodic
nature of the mapping of sinh(z) + c in the complex z plane is not evident from the
pattern but mapping with a larger window reveals that the periodicity is 2pi and the
pattern is formed from one 2pi unit along the imaginary axis.

A modified version of the Level Set Method [5] was used to generate the pattern
from the appended data. In order to clarify the division between the bounded and
divergent points, divergent lemniscates n > 15 were not printed.
22!)
226
Figure 1. "Serpents and Dragons": A Julia set map of the function z —> sinh(z) + c.
 "Serpents and Dragons": A Mapping of ... 227

P s e u do Code: Modified Level Set Method

Arguments: nx,ny, xmin, xmax, ymin, j/max, itermax, ex, cy.
Variables: real x,y,cr, cs,ess,csi integer sy,sx.
BEGIN FOR sy = 1 TO ny FOR sx
= 1 TO nx
y = ymin+sj/ * (j/max—ymin)/ny :x = xmin+sx * (xmax—xmin)/rtx : csi = 25

FOR iter=l to itermax

x = 0.5 * (ex — e~x) * cos(y) + ex :y = —0.5 * (ex — e _ I ) * sin(y) + c cr = x * x\
ci = y * y : ess = cr + ci
IF ess < csi THEN csi = ess
IF ess > 1000 THEN EXIT LOOP TO TEST
END LOOP
csi=INT(csi * 110):REM ASSIGN COLOR VALUES TO csi
IF ABS(x) < 100 OR ABS(y) < 100 THEN ASSIGN COLOR VALUE
END LOOP:END LOOP

Appendix
ra/=2420, na;=1320, xmin=—1.5, imax=2 . 1, ymm=—2.1, ymax=l.6,
ar=1.1666, q/= . 8084, itermax=150.

References
1. C. A. Pickover, "Chaotic behavior of the transcendental mapping (z — cosh(z)+u", The
Visual Computer 4 (1988) 243246.
2. C. A. Pickover, Computers, Pattern, Chaos, and Beauty — Graphics From an Unseen
World (St. Martins Press, 1990).
3. F. J. Flanagan, Complex Variables: Harmonic and Analytical Functions (Dover
Publications Inc., 1972).
4. I. D. Entwistle, "Julia set art and fractals in the complex plane", Computers and Graphics
13, 3 (1989) 389392.
5. H.O. Peitgen and D. Saupe, The Science of Fractal Images (SpringerVerlag, 1988).
Ian D. Entwistle
"Star of Chaos": A Multiple Decomposition
7
Mapping of f(z) : z —* —z + c in the Complex
Plane

The pattern illustrates an elaborated Julia set fractal obtained from itera tion of
7
the function z + c in the complex plane.
2
Iterative complex plane maps of the function z + c have been widely ex plored
by graphical methods because of the fascinating fractals that can be generated [1].
The striking beauty of these maps and the relationship of their geometry to the
corresponding z plane mappings has prompted the study of the similar behavior of
higher polynomial functions. On iteration, such func tions generate z(n) values
approaching infinity within a few iterations (n) and very few of the points in the
complex plane are part of the "bounded set" if the value of c is in the main cardoid
of the corresponding Mandelbrot set. The resulting Julia sets fractals have outlines
which are quasicircles, i.e., they are homomorphic to a circle. The maps are not
2
visually so appealing as those obtained for z + c since there is no complex fractal
behavior close to the boundary between the divergent and bounded points. Most
points are mapped to only a few distinct quasicircles. Variations in the escape
values employed or control of the point mapping positions on the basis of the z(n)
p
value have been utilized to produce visually exciting maps for the functions z + c
(p = 3, 4, 5) [4]. Other variations related to the pattern have also been studied [5].
In order to generate the complex pattern evident in the figure, secondary data is
generated from the values of z(n) after completion of itera tion for each point. This
new value controls the printing color of the point and results in the transformation
to treelike patterns which radiate to the pat tern edges. This is achieved using a
modification of the Binary Decomposition Method for Julia Sets (BDM/J) [6]
derived from the mathematics of binary decomposition [7]. A simplified description
has been reported which allows multiple decomposition maps to be generated [8].
7
Application of BDM/J to the function z + c can be simplified for computation
7
using languages which do not support complex variables. Substitution of z + c by z
= x + i * y using standard complex algebra [9] leads to the two equations (1 and 2)
228
22
9