PROLOGUE

THE VIRTUAL UNREALITY MACHINE

I have a dream.
I am surrounded by-nothing. Not empty space, for there is no space to be empty. Not blackness, for there is
nothing to be black. Simply an absence, waiting to become a presence. I think commands: let there be space. But
what kind of space? I have a choice: three-dimensional space, multidimensional space, even curved space.
I choose.

Another command, and the space is filled with an all- pervading fluid, which swirls in waves and vortices,
here a placid swell, there a frothing, turbulent maelstrom.
I paint space blue, draw white streamlines in the fluid to bring out the flow patterns. I place a small red
sphere in the fluid. It hovers, unsupported, ignorant of the chaos around it, until I give the word. Then it slides off
along a streamline. I compress myself to one hundredth of my size and will myself onto the surface of the sphere, to
get a bird's-eye view of unfolding events. Every few seconds, I place a green marker in the flow to record the
sphere's passing. If I touch a marker, it blossoms like a time-lapse film of a desert cactus when the rains come-and
on every petal there are pictures, numbers, symbols. The sphere can also be made to blossom, and when it does,
those pictures, numbers, and symbols change as it moves.
Dissatisfied with the march of its symbols, I nudge the sphere onto a different streamline, fine-tuning its
position until I see the unmistakable traces of the singularity I am seeking. I snap my fingers, and the sphere
extrapolates itself into its own future and reports back what it finds. Promising ... Suddenly there is a whole cloud of
red spheres, all being carried along by the fluid, like a shoal of fish that quickly spreads, swirling, putting out
tendrils, flattening into sheets. Then more shoals of spheres join the game-gold, purple, brown, silver, pink .... I am
in danger of running out of colors. Multicolored sheets intersect in a complex geometric form. I freeze it, smooth it,
paint it in stripes. I banish the spheres with a gesture. I call up markers, inspect their unfolded petals, pull some off
and attach them to a translucent grid that has materialized like a landscape from thinning mist.

Yes!

I issue a new command. "Save. Title: A new chaotic phenomenon in the three-body problem. Date: today."
Space collapses back to nonexistent void. Then, the morning's research completed, I disengage from my
Virtual Unreality Machine and head off in search of lunch.
This particular dream is very nearly fact. We already have Virtual Reality systems that simulate events in "normal"
space. I call my dream Virtual Unreality because it simulates anything that can be created by the mathematician's
fertile imagination. Most of the bits and pieces of the Virtual Unreality Machine exist already. There is computer-
graphics software that can "fly" you through any chosen geometrical object, dynamical-systems software that can
track the evolving state of any chosen equation, symbolic-algebra software that can take the pain out of the most
horrendous calculations-and get them right. It is only a matter of time before mathematicians will be able to get
inside their own creations.
But wonderful though such technology may be, we do not need it to bring my dream to life. The dream is a
reality now, present inside every mathematician's head. This is what mathematical creation feels like when you're
doing it. I've resorted to a little poetic license: the objects that are found in the mathematician's world are generally
distinguished by symbolic labels or names rather than colors. But those labels are as vivid as colors to those who
inhabit that world. In fact, despite its colorful images, my dream is a pale shadow of the world of imagination that
every mathematician inhabits-a world in which curved space, or space with more than three dimensions, is not only
commonplace but inevitable. You probably find the images alien and strange, far removed from the algebraic
symbolism that the word "mathematics" conjures up. Mathematicians are forced to resort to written symbols and
pictures to describe their world-even to each other. But the symbols are no more that world than musical notation
is music.
Over the centuries, the collective minds of mathematicians have created their own universe. I don't know
where it is situated-I don't think that there is a "where" in any normal sense of the word-but I assure you that this
mathematical universe seems real enough when you're in it. And, not despite its peculiarities but because of them,
the mental universe of mathematics has provided human beings with many of their deepest insights into the world
around them.
I am going to take you sightseeing in that mathematical universe. I am going to try to equip you with a
mathematician's eyes. And by so doing, I shall do my best to change the way you view your own world.

1
 NATURES NUMBERS
CHAPTER 1
THE NATURAL ORDER

We live in a universe of patterns.

Every night the stars move in circles across the sky. The seasons cycle at yearly intervals. No two
snowflakes are ever exactly the same, but they all have sixfold symmetry. Tigers and zebras are covered in patterns
of stripes, leopards and hyenas are covered in patterns of spots. Intricate trains of waves march across the oceans,
very similar trains of sand dunes march across the desert. Colored arcs of light adorn the sky in the form of
rainbows, and a bright circular halo sometimes surrounds the moon on winter nights. Spherical drops of water fall
from clouds.
Human mind and culture have developed a formal system of thought for recognizing, classifying, and
exploiting patterns. We call it mathematics. By using mathematics to organize and systematize our ideas about
patterns, we have discovered a great secret: nature's patterns are not just there to be admired, they are vital clues to
the rules that govern natural processes. Four hundred years ago, the German astronomer Johannes Kepler wrote a
small book, The Six-Cornered Snowflake, as a New Year's gift to his sponsor. In it he argued that snowflakes must
be made by packing tiny identical units together. This was long before the theory that matter is made of atoms had
become generally accepted. Kepler performed no experiments; he just thought very hard about various bits and
pieces of common knowledge. His main evidence was the sixfold symmetry of snowflakes, which is a natural
consequence of regular packing. If you place a large number of identical coins on a table and try to pack them as
closely as possible, then you get a honeycomb arrangement, in which every coin-except those at the edges-is
surrounded by six others, arranged in a perfect hexagon.
The regular nightly motion of the stars is also a clue, this time to the fact that the Earth rotates. Waves and
dunes are clues to the rules that govern the flow of water, sand, and air. The tiger's stripes and the hyena's spots
attest to mathematical regularities in biological growth and form. Rainbows tell us about the scattering of light, and
indirectly confirm that raindrops are spheres. Lunar haloes are clues to the shape of ice crystals.
There is much beauty in nature's clues, and we can all recognize it without any mathematical training.
There is beauty, too, in the mathematical stories that start from the clues and deduce the underlying rules and
regularities, but it is a different kind of beauty, applying to ideas rather than things. Mathematics is to nature as
Sherlock Holmes is to evidence. When presented with a cigar butt, the great fictional detective could deduce the age,
profession, and financial state of its owner. His partner, Dr. Watson, who was not as sensitive to such matters, could
only look on in baffled admiration, until the master revealed his chain of impeccable logic. When presented with the
evidence of hexagonal snowflakes, mathematicians can deduce the atomic geometry of ice crystals. If you are a
Watson, it is just as baffling a trick, but I want to show you what it is like if you are a Sherlock Holmes.
Patterns possess utility as well as beauty. Once we have learned to recognize a background pattern,
exceptions suddenly stand out. The desert stands still, but the lion moves. Against the circling background of stars, a
small number of stars that move quite differently beg to be singled out for special attention. The Greeks called them
planetes, meaning "wanderer," a term retained in our word "planet." It took a lot longer to understand the patterns of
planetary motion than it did to work out why stars seem to move in nightly circles. One difficulty is that we are
inside the Solar System, moving along with it, and things that look simple from outside often look much more
complicated from inside. The planets were clues to the rules behind gravity and motion.
We are still learning to recognize new kinds of pattern. Only within the last thirty years has humanity
become explicitly aware of the two types of pattern now known as fractals and chaos. Fractals are geometric shapes
that repeat their structure on ever-finer scales, and I will say a little about them toward the end of this chapter; chaos
is a kind of apparent randomness whose origins are entirely deterministic, and I will say a lot about that in chapter 8.
Nature "knew about" these patterns billions of years ago, for clouds are fractal and weather is chaotic. It took
humanity a while to catch up.
The simplest mathematical objects are numbers, and the simplest of nature's patterns are numerical. The
phases of the moon make a complete cycle from new moon to full moon and back again every twenty-eight days.
The year is three hundred and sixty-five days long-roughly. People have two legs, cats have four, insects have six,
and spiders have eight. Starfish have five arms (or ten, eleven, even seventeen, depending on the species). Clover
normally has three leaves: the superstition that a four-leaf clover is lucky reflects a deep- seated belief that
exceptions to patterns are special. A very curious pattern indeed occurs in the petals of flowers. In nearly all flowers,
the number of petals is one of the numbers that occur in the strange sequence 3, 5, 8, 13, 21, 34, 55, 89. For instance,
lilies have three petals, buttercups have five, many delphiniums have eight, marigolds have thirteen, asters have
twenty-one, and most daisies have thirty-four, fifty-five, or eighty-nine. You don't find any other numbers anything
like as often. There is a definite pattern to those numbers, but one that takes a little digging out: each number is
obtained by adding the previous two numbers together. For example, 3 + 5 = 8, 5 + 8 = 13, and so on. The same
numbers can be found in the spiral patterns of seeds in the head of a sunflower. This particular pattern was noticed

2
many centuries ago and has been widely studied ever since, but a really satisfactory explanation was not given until
1993. It is to be found in chapter 9.
Numerology is the easiest-and consequently the most dangerous-method for finding patterns. It is easy
because anybody can do it, and dangerous for the same reason. The difficulty lies in distinguishing significant
numerical patterns from accidental ones. Here's a case in point. Kepler was fascinated with mathematical patterns in
nature, and he devoted much of his life to looking for them in the behavior of the planets. He devised a simple and
tidy theory for the existence of precisely six planets (in his time only Mercury, Venus, Earth, Mars, Jupiter, and
Saturn were known). He also discovered a very strange pattern relating the orbital period of a planet-the time it takes
to go once around the Sun-to its distance from the Sun. Recall that the square of a number is what you get when you
multiply it by itself: for example, the square of 4 is 4 x 4 = 16. Similarly, the cube is what you get when you
multiply it by itself twice: for example, the cube of 4 is 4 x 4 x 4 = 64. Kepler found that if you take the cube of the
distance of any planet from the Sun and divide it by the square of its orbital period, you always get the same
number. It was not an especially elegant number, but it was the same for all six planets.
Which of these numerological observations is the more significant? The verdict of posterity is that it is the
second one, the complicated and rather arbitrary calculation with squares and cubes. This numerical pattern was one
of the key steps toward Isaac Newton's theory of gravity, which has explained all sorts of puzzles about the motion
of stars and planets. In contrast, Kepler's neat, tidy theory for the number of planets has been buried without trace.
For a start, it must be wrong, because we now know of nine planets, not six. There could be even more, farther out
from the Sun, and small enough and faint enough to be undetectable. But more important, we no longer expect to
find a neat, tidy theory for the number of planets. We think that the Solar System condensed from a cloud of gas
surrounding the Sun, and the number of planets presumably depended on the amount of matter in the gas cloud, how
it was distributed, and how fast and in what directions it was moving. An equally plausible gas cloud could have
given us eight planets, or eleven; the number is accidental, depending on the initial conditions of the gas cloud,
rather than universal, reflecting a general law of nature.
The big problem with numerological pattern-seeking is that it generates millions of accidentals for each
universal. Nor is it always obvious which is which. For example, there are three stars, roughly equally spaced and in
a straight line, in the belt of the constellation Orion. Is that a clue to a significant law of nature? Here's a similar
question. 10, Europa, and Ganymede are three of Jupiter's larger satellites. They orbit the planet in, respectively,
1.77, 3.55, and 7.16 days. Each of these numbers is almost exactly twice the previous one. Is that a significant
pattern? Three stars in a row, in terms of position; three satellites "in a row" in terms of orbital period. Which
pattern, if either, is an important clue? I'll leave you to think about that for the moment and return to it in the next
chapter.
In addition to numerical patterns, there are geometric ones. In fact this book really ought to have been
called Nature's Numbers and Shapes. I have two excuses. First, the title sounds better without the "and shapes."
Second, mathematical shapes can always be reduced to numbers-which is how computers handle graphics. Each tiny
dot in the picture is stored and manipulated as a pair of numbers: how far the dot is along the screen from right to
left, and how far up it is from the bottom. These two numbers are called the coordinates of the dot. A general shape
is a collection of dots and can be represented as a list of pairs of numbers. However, it is often better to think of
shapes as shapes, because that makes use of our powerful and intuitive visual capabilities, whereas complicated lists
of numbers are best reserved for our weaker and more laborious symbolic abilities.
Until recently, the main shapes that appealed to mathematicians were very simple ones: triangles, squares,
pentagons, hexagons, circles, ellipses, spirals, cubes, spheres, cones, and so on. All of these shapes can be found in
nature, although some are far more common, or more evident, than others. The rainbow, for example, is a collection
of circles, one for each color. We don't normally see the entire circle, just an arc; but rainbows seen from the air can
be complete circles. You also see circles in the ripples on a pond, in the human eye, and on butterflies' wings.
Talking of ripples, the flow of fluids provides an inexhaustible supply of nature's patterns. There are waves
of many different kinds-surging toward a beach in parallel ranks, spreading in a V-shape behind a moving boat,
radiating outward from an underwater earthquake. Most waves are gregarious creatures, but some-such as the tidal
bore that sweeps up a river as the energy of the incoming tide becomes confined to a tight channel-are solitary.
There are swirling spiral whirlpools and tiny vortices. And there is the apparently structureless, random frothing of
turbulent flow, one of the great enigmas of mathematics and physics. There are similar patterns in the atmosphere,
too, the most dramatic being the vast spiral of a hurricane as seen by an orbiting astronaut.
There are also wave patterns on land. The most strikingly mathematical landscapes on Earth are to be found
in the great ergs, or sand oceans, of the Arabian and Sahara deserts. Even when the wind blows steadily in a fixed
direction, sand dunes form. The simplest pattern is that of transverse dunes, which-just like ocean waves-line up in
parallel straight rows at right angles to the prevailing wind direction. Sometimes the rows themselves become wavy,
in which case they are called barchanoid ridges; sometimes they break up into innumerable shield-shaped barchan
dunes. If the sand is slightly moist, and there is a little vegetation to bind it together, then you may find parabolic
dunes-shaped like a U, with the rounded end pointing in the direction of the wind. These sometimes occur in

3
clusters, and they resemble the teeth of a rake. If the wind direction is variable, other forms become possible. For
example, clusters of star-shaped dunes can form, each having several irregular arms radiating from a central peak.
They arrange themselves in a random pattern of spots.
Nature's love of stripes and spots extends into the animal kingdom, with tigers and leopards, zebras and
giraffes. The shapes and patterns of animals and plants are a happy hunting ground for the mathematically minded.
Why, for example, do so many shells form spirals? Why are starfish equipped with a symmetric set of arms? Why
do many viruses assume regular geometric shapes, the most striking being that of an icosahedron-a regular solid
formed from twenty equilateral triangles? Why are so many animals bilaterally symmetric? Why is that symmetry so
often imperfect, disappearing when you look at the detail, such as the position of the human heart or the differences
between the two hemispheres of the human brain? Why are most of us right-handed, but not all of us?
In addition to patterns of form, there are patterns of movement. In the human walk, the feet strike the
ground in a regular rhythm: left-right-Ieft-right-Ieft-right. When a four-legged creature-a horse, say-walks, there is a
more complex but equally rhythmic pattern. This prevalence of pattern in locomotion extends to the scuttling of
insects, the flight of birds, the pulsations of jellyfish, and the wavelike movements of fish, worms, and snakes. The
sidewinder, a desert snake, moves rather like a single coil of a helical spring, thrusting its body forward in a series of
S-shaped curves, in an attempt to minimize its contact with the hot sand. And tiny bacteria propel themselves along
using microscopic helical tails, which rotate rigidly, like a ship's screw.

Finally, there is another category of natural pattern-one that has captured human imagination only very
recently, but dramatically. This comprises patterns that we have only just learned to recognize-patterns that exist
where we thought everything was random and formless. For instance, think about the shape of a cloud. It is true that
meteorologists classify clouds into several different morphological groups-cirrus, stratus, cumulus, and so on-but
these are very general types of form, not recognizable geometric shapes of a conventional mathematical kind. You
do not see spherical clouds, or cubical clouds, or icosahedral clouds. Clouds are wispy, formless, fuzzy clumps. Yet
there is a very distinctive pattern to clouds, a kind of symmetry, which is closely related to the physics of cloud
formation. Basically, it is this: you can't tell what size a cloud is by looking at it. If you look at an elephant, you can
tell roughly how big it is: an elephant the size of a house would collapse under its own weight, and one the size of a
mouse would have legs that are uselessly thick. Clouds are not like this at all. A large cloud seen from far away and
a small cloud seen close up could equally plausibly have been the other way around. They will be different in shape,
of course, but not in any manner that systematically depends on size.

This "scale independence" of the shapes of clouds has been verified experimentally for cloud patches
whose sizes vary by a factor of a thousand. Cloud patches a kilometer across look just like cloud patches a thousand
kilometers across. Again, this pattern is a clue. Clouds form when water undergoes a "phase transition" from vapor
to liquid, and physicists have discovered that the same kind of scale invariance is associated with all phase
transitions. Indeed, this statistical self-similarity, as it is called, extends to many other natural forms. A Swedish
colleague who works on oil-field geology likes to show a slide of one of his friends standing up in a boat and
leaning nonchalantly against a shelf of rock that comes up to about his armpit. The photo is entirely convincing, and
it is clear that the boat must have been moored at the edge of a rocky gully about two meters deep. In fact, the rocky
shelf is the side of a distant fjord, some thousand meters high. The main problem for the photographer was to get
both the foreground figure and the distant landscape in convincing focus.

Nobody would try to play that kind of trick with an elephant.

However, you can play it with many of nature's shapes, including mountains, river networks, trees, and
very possible the way that matter is distributed throughout the entire uni- verse. In the term made famous by the
mathematician Benoit Mandelbrot, they are all fractals. A new science of irregularity-fractal geometry-has sprung
up within the last fifteen years. I'm not going to say much about fractals, but the dynamic process that causes them,
known as chaos, will be prominently featured.
Thanks to the development of new mathematical theories, these more elusive of nature's patterns are
beginning to reveal their secrets. Already we are seeing a practical impact as well as an intellectual one. Our
newfound understanding of nature's secret regularities is being used to steer artificial satellites to new destinations
with far less fuel than anybody had thought possible, to help avoid wear on the wheels of locomotives and other
rolling stock, to improve the effective- ness of heart pacemakers, to manage forests and fisheries, even to make more
efficient dishwashers. But most important of all, it is giving us a deeper vision of the universe in which we live, and
of our own place in it.

4
CHAPTER 2
WHAT MATHEMATICS IS FOR?

We’ve now established the uncontroversial idea that nature is full of patterns. But what do we want to do with them?
One thing we can do is sit back and admire them. Communing with nature does all of us good: it reminds us of what
we are. Painting pictures, sculpting sculptures, and writing poems are valid and important ways to express our
feelings about the world and about ourselves. The entrepreneur's instinct is to exploit the natural world. The
engineer's instinct is to change it. The scientist's instinct is to try to understand it-to work out what's really going on.
The mathematician's instinct is to structure that process of understanding by seeking generalities that cut across the
obvious subdivisions. There is a little of all these instincts in all of us, and there is both good and bad in each
instinct.
I want to show you what the mathematical instinct has done for human understanding, but first I want to
touch upon the role of mathematics in human culture. Before you buy something, you usually have a fairly clear idea
of what you want to do with it. If it is a freezer, then of course you want it to preserve food, but your thoughts go
well beyond that. How much food will you need to store? Where will the freezer have to fit? It is not always a
matter of utility; you may be thinking of buying a painting, You still ask yourself where you are going to put it, and
whether the aesthetic appeal is worth the asking price. It is the same with mathematics-and any other intellectual
worldview, be it scientific, political, or religious. Before you buy something, it is wise to decide what you want it
for.
So what do we want to get out of mathematics?

Each of nature's patterns is a puzzle, nearly always a deep one. Mathematics is brilliant at helping us to
solve puzzles. It is a more or less systematic way of digging out the rules and structures that lie behind some
observed pattern or regularity, and then using those rules and structures to explain what's going on. Indeed,
mathematics has developed alongside our understanding of nature, each reinforcing the other. I've mentioned
Kepler's analysis of snowflakes, but his most famous discovery is the shape of planetary orbits. By performing a
mathematical analysis of astronomical observations made by the contemporary Danish astronomer Tycho Brahe,
Kepler was eventually driven to the conclusion that planets move in ellipses. The ellipse is an oval curve that was
much studied by the ancient Greek geometers, but the ancient astronomers had preferred to use circles, or systems of
circles, to describe orbits, so Kepler's scheme was a radical one at that time.
People interpret new discoveries in terms of what is important to them. The message astronomers received
when they heard about Kepler's new idea was that neglected ideas from Greek geometry could help them solve the
puzzle of predicting planetary motion. It took very little imagination for them to see that Kepler had made a huge
step forward. All sorts of astronomical phenomena, such as eclipses, meteor showers, and comets, might yield to the
same kind of mathematics. The message to mathematicians was quite different. It was that ellipses are really
interesting curves. It took very little imagination for them to see that a general theory of curves would be even more
interesting. Mathematicians could take the geometric rules that lead to ellipses and modify them to see what other
kinds of curve resulted.
Similarly, when Isaac Newton made the epic discovery that the motion of an object is described by a
mathematical relation between the forces that act on the body and the acceleration it experiences, mathematicians
and physicists learned quite different lessons. However, before I can tell you what these lessons were I need to
explain about acceleration. Acceleration is a subtle concept: it is not a fundamental quantity, such as length or mass;
it is a rate of change. In fact, it is a "second order" rate of change-that is, a rate of change of a rate of change. The
velocity of a body-the speed with which it moves in a given direction-is just a rate of change: it is the rate at which
the body's distance from some chosen point changes. If a car moves at a steady speed of sixty miles per hour, its
distance from its starting point changes by sixty miles every hour. Acceleration is the rate of change of velocity. If
the car's velocity increases from sixty miles per hour to sixty-five miles per hour, it has accelerated by a definite
amount. That amount depends not only on the initial and final speeds, but on how quickly the change takes place. If
it takes an hour for the car to increase its speed by five miles per hour, the acceleration is very small; if it takes only
ten seconds, the acceleration is much greater.
I don't want to go into the measurement of accelerations. My point here is more general: that acceleration is
a rate of change of a rate of change. You can work out distances with a tape measure, but it is far harder to work out
a rate of change of a rate of change of distance. This is why it took humanity a long time, and the genius of a
Newton, to discover the law of motion. If the pattern had been an obvious feature of distances, we would have
pinned motion down a lot earlier in our history.
In order to handle questions about rates of change, New- ton-and independently the German mathematician
Gottfried Leibniz-invented a new branch of mathematics, the calculus. It changed the face of the Earth-literally and
metaphorically. But, again, the ideas sparked by this discovery were different for different people. The physicists
went off looking for other laws of nature that could explain natural phenomena in terms of rates of change. They

5
found them by the bucketful heat, sound, light, fluid dynamics, elasticity, electricity, mag- netism. The most esoteric
modern theories of fundamental particles still use the same general kind of mathematics, though the interpretation-
and to some extent the implicit worldview-is different. Be that as it may, the mathematicians found a totally
different set of questions to ask. First of all, they spent a long time grappling with what "rate of change" really
means. In order to work out the velocity of a moving object, you must measure where it is, find out where it moves
to a very short interval of time later, and divide the distance moved by the time elapsed. However, if the body is
accelerating, the result depends on the interval of time you use. Both the mathematicians and the physicists had the
same intuition about how to deal with this problem: the interval of time you use should be as small as possible.
Everything would be wonderful if you could just use an interval of zero, but unfortunately that won't work, because
both the distance traveled and the time elapsed will be zero, and a rate of change of DID is meaningless. The main
problem with nonzero intervals is that whichever one you choose, there is always a smaller one that you could use
instead to get a more accurate answer. What you would really like is to use the smallest possible nonzero interval of
time-but there is no such thing, because given any nonzero number, the number half that size is also nonzero.
Everything would work out fine if the interval could be made infinitely small-"infinitesimal." Unfortunately, there
are difficult logical paradoxes associated with the idea of an infinitesimal; in particular, if we restrict our- selves to
numbers in the usual sense of the word, there is no such thing. So for about two hundred years, humanity was in a
very curious position as regards the calculus. The physicists were using it, with great success, to understand nature
and to predict the way nature behaves; the mathematicians were worrying about what it really meant and how best to
set it up so that it worked as a sound mathematical theory; and the philosophers were arguing that it was all
nonsense. Every- thing got resolved eventually, but you can still find strong differences in attitude.

The story of calculus brings out two of the main things that mathematics is for: providing tools that let
scientists calculate what nature is doing and providing new questions for mathematicians to sort out to their own
satisfaction. These are the external and internal aspects of mathematics, often referred to as applied and pure
mathematics (I dislike both adjectives, and I dislike the implied separation even more). It might appear in this case
that the physicists set the agenda: if the methods of calculus seem to be working, what does it matter why they
work? You will hear the same sentiments expressed today by people who pride themselves on being pragmatists. I
have no difficulty with the proposition that in many respects they are right. Engineers designing a bridge are entitled
to use standard mathematical methods even if they don't know the detailed and often esoteric reasoning that justifies
these methods. But I, for one, would feel uncomfortable driving across that bridge if I was aware that nobody knew
what justified those methods. So, on a cultural level, it pays to have some people who worry about pragmatic
methods and try to find out what really makes them tick. And that's one of the jobs that mathematicians do. They
enjoy it, and the rest of humanity benefits from various kinds of spin-off, as we'll see.
In the short term, it made very little difference whether mathematicians were satisfied about the logical
soundness of the calculus. But in the long run the new ideas that mathematicians got by worrying about these
internal difficulties turned out to be very useful indeed to the outside world. In Newton's time, it was impossible to
predict just what those uses would be, but I think you could have predicted, even then, that uses would arise. One of
the strangest features of the relationship between mathematics and the "real world," but also one of the strongest, is
that good mathematics, what- ever its source, eventually turns out to be useful. There are all sorts of theories why
this should be so, ranging from the structure of the human mind to the idea that the universe is somehow built from
little bits of mathematics. My feeling is that the answer is probably quite simple: mathematics is the science of
patterns, and nature exploits just about every pattern that there is. I admit that I find it much harder to offer a
convincing reason for nature to behave in this manner. Maybe the question is back to front: maybe the point is that
creatures able to ask that kind of question can evolve only in a universe with that kind of structure.*
Whatever the reasons, mathematics definitely is a useful way to think about nature. What do we want it to
tell us about the patterns we observe? There are many answers. We want to understand how they happen; to
understand why they hap- pen, which is different; to organize the underlying patterns and regularities in the most
satisfying way; to predict how nature will behave; to control nature for our own ends; and to make practical use of
what we have learned about our world. Mathematics helps us to do all these things, and often it is indispensable.

For example, consider the spiral form of a snail shell. How the snail makes its shell is largely a matter of
chemistry and genetics. Without going into fine points, the snail's genes include recipes for making particular
chemicals and instructions for where they should go. Here mathematics lets us do the molecular bookkeeping that
makes sense of the different chemical reactions that go on; it describes the atomic structure of the molecules used in
shells, it describes the strength and rigidity of shell material as compared to the weakness and pliability of the snail's
body, and so on. Indeed, without mathematics we would never have convinced ourselves that matter really is made
from atoms or have worked out how the atoms are arranged. The discovery of genes-and later of the molecular
structure of DNA, the genetic material-relied heavily on the existence of mathematical clues. The monk Gregor
Mendel noticed tidy numerical relationships in how …

6
* This explanation. and others. are discussed in The Collapse of Chaos. By Jack Cohen and Ian Stewart (New York:
Viking. 1994).

… the proportions of plants with different characters, such as seed color, changed when the plants were crossbred.
This led to the basic idea of genetics-that within every organism is some cryptic combination of factors that
determines many features of its body plan, and that these factors are somehow shuffled and recombined when
passing from parents to offspring. Many different pieces of mathematics were involved in the discovery that DNA
has the celebrated double-helical structure. They were as simple as Chargaff's rules: the observation by the Austrian-
born biochemist Erwin Chargaff that the four bases of the DNA molecule occur in related proportions; and they are
as subtle as the laws of diffraction, which were used to deduce molecular structure from X-ray pictures of DNA
crystals.
The question of why snails have spiral shells has a very different character. It can be asked in several
contexts-in the short-term context of biological development, say, or the long- term context of evolution. The main
mathematical feature of the developmental story is the general shape of the spiral. Basically, the developmental
story is about the geometry of a creature that behaves in much the same way all the time but keeps getting bigger.
Imagine a tiny animal, with a tiny proto- shell attached to it. Then the animal starts to grow. It can grow most easily
in the direction along which the open rim of the shell points, because the shell gets in its way if it tries to grow in
any other direction. But, having grown a bit, it needs to extend its shell as well, for self-protection. So, of course, the
shell grows an extra ring of material around its rim. As this process continues, the animal is getting bigger, so the
size of the rim grows. The simplest result is a conical shell, such as you find on a limpet. But if the whole system
starts with a bit of a twist, as is quite likely, then the growing edge of the shell rotates slowly as well as expanding,
and it rotates in an off-centered manner. The result is a cone that twists in an ever-expanding spiral. We can use
mathematics to relate the resulting geometry to all the different variables-such as growth rate and eccentricity of
growth-that are involved.
If, instead, we seek an evolutionary explanation, then we might focus more on the strength of the shell,
which conveys an evolutionary advantage, and try to calculate whether a long thin cone is stronger or weaker than a
tightly coiled spiral. Or we might be more ambitious and develop mathematical models of the evolutionary process
itself, with its combination of random genetic change-that is, mutations-and natural selection.
A remarkable example of this kind of thinking is a computer simulation of the evolution of the eye by
Daniel Nilsson and Susanne Pelger, published in 1994. Recall that conventional evolutionary theory sees changes in
animal form as being the result of random mutations followed by subsequent selection of those individuals most able
to survive and reproduce their kind. When Charles Darwin announced this theory, one of the first objections raised
was that complex structures (like an eye) have to evolve fully formed or else they won't work properly (half an eye
is no use at all), but the chance that random mutation will produce a coherent set of complex changes is negligible.
Evolutionary theorists quickly responded that while half an eye may not be much use, a half- developed eye might
well be. One with a retina but no lens, say, will still collect light and thereby detect movement; and any way to
improve the detection of predators offers an evolutionary advantage to any creature that possesses it. What we have
here is a verbal objection to the theory countered by a verbal argument. But the recent computer analysis goes much
further.
It starts with a mathematical model of a flat region of cells, and permits various types of "mutation." Some
cells may become more sensitive to light, for example, and the shape of the region of cells may bend. The
mathematical model is set up as a computer program that makes tiny random changes of this kind, calculates how
good the resulting structure is at detecting light and resolving the patterns that it "sees," and selects any changes that
improve these abilities. During a simulation that corresponds to a period of about four hundred thousand years-the
blink of an eye, in evolutionary terms- the region of cells folds itself up into a deep, spherical cavity with a tiny iris
like opening and, most dramatically, a lens. Moreover, like the lenses in our own eyes, it is a lens whose refractive
index-the amount by which it bends light-varies from place to place. In fact, the pattern of variation of refractive
index that is produced in the computer simulation is very like our own. So here mathematics shows that eyes
definitely can evolve gradually and naturally, offering increased survival value at every stage. More than that:
Nilsson and Pelger's work demonstrates that given certain key biological faculties (such as cellular receptivity to
light, and cellular mobility), structures remarkably similar to eyes will form-all in line with Darwin's principle of
natural selection. The mathematical model provides a lot of extra detail that the verbal Darwinian argument can only
guess at, and gives us far greater confidence that the line of argument is correct.
I said that another function of mathematics is to organize the underlying patterns and regularities in the
most satisfying way. To illustrate this aspect, let me return to the question raised in the first chapter. Which-if either-
is significant:

7
Figure 1: Computer model of the evolution of an eye. Each step in the computation corresponds to about two
hundred years of biological evolution.

… the three-in-a-row pattern of stars in Orion's belt, or the three in-a-row patterns to the periods of revolution of
Jupiter's satellites? Orion first. Ancient human civilizations organized the stars in the sky in terms of pictures of
animals and mythic heroes. In these terms, the alignment of the three stars in Orion appears significant, for
otherwise the hero would have no belt from which to hang his sword. However, if we use three-dimensional
geometry as an organizing principle and place the three stars in their correct positions in the heavens, then we find
that they are at very different distances from the Earth. Their equispaced alignment is an accident, depending on the
position from which they are being viewed. Indeed, the very word "constellation" is a misnomer for an arbitrary
accident of viewpoint.
The numerical relation between the periods of revolution of 10, Europa, and Ganymede could also be an
accident of viewpoint. How can we be sure that "period of revolution" has any significant meaning for nature?
However, that numerical relation fits into a dynamical framework in a very significant manner indeed. It is an
example of a resonance, which is a relationship between periodically moving bodies in which their cycles are locked
together, so that they take up the same relative positions at regular intervals. This common cycle time is called the
period of the system. The individual bodies may have different-but related-periods. We can work out what this
relationship is. When a resonance occurs, all of the participating bodies must return to a standard reference position
after a whole number of cycles-but that number can be different for each. So there is some common period for the
system, and therefore each individual body has a period that is some whole-number divisor of the common period.
In this case, the common period is that of Ganymede, 7.16 days. The period of Europa is very close to half that of
Ganymede, and that of 10 is close to one-quarter. 10 revolves four times around Jupiter while Europa revolves twice
and Ganymede once, after which they are all back in exactly the same relative positions as before. This is called a
4:2:1 resonance.

The dynamics of the Solar System is full of resonances. The Moon's rotational period is (subject to small
wobbles caused by perturbations from other bodies) the same as its period of revolution around the Earth-a 1:1
resonance of its orbital and its rotational period. Therefore, we always see the same face of the Moon from the Earth,
never its "far side." Mercury rotates once every 58.65 days and revolves around the Sun every 87.97 days. Now, 2 x
87.97 = 175.94, and 3 x 58.65 = 175.95, so Mercury's rotational and orbital periods are in a 2:3 resonance. (In fact,
for a long time they were thought to be in 1:1 resonance, both being roughly 88 days, because of the difficulty of
observing a planet as close to the Sun as Mercury is. This gave rise to the belief that one side of Mercury is

8
 incredibly hot and the other incredibly cold, which turns out not to be true. A resonance, however, there is-and a
more interesting one than mere equality.)
In between Mars and Jupiter is the asteroid belt, a broad zone containing thousands of tiny bodies. They are not
uniformly distributed. At certain distances from the Sun we find asteroid "beltlets"; at other distances we find hardly
any. The explanation-in both cases-is resonance with Jupiter. The Hilda group of asteroids, one of the beltlets, is in
2:3 resonance with Jupiter. That is, it is at just the right distance so that all of the Hilda asteroids circle the Sun three
times for every two revolutions of Jupiter. The most noticeable gaps are at 2:1, 3:1, 4:1, 5:2, and 7:2 resonances.
You may be worried that resonances are being used to explain both clumps and gaps, The reason is that each
resonance has its own idiosyncratic dynamics; some cause clustering, others do the oppo- site. It all depends on the
precise numbers.
Another function of mathematics is prediction. By under- standing the motion of heavenly bodies,
astronomers could predict lunar and solar eclipses and the return of comets. They knew where to point their
telescopes to find asteroids that had passed behind the Sun, out of observation?-l contact. Because the tides are
controlled mainly by the position of the Sun and Moon relative to the Earth, they could predict tides many years
ahead. (The chief complicating factor in making such predictions is not astronomy: it is the shape of the continents
and the profile of the ocean depths, which can delay or advance a high tide. However, these stay pretty much the
same from one century to the next, so that once their effects have been understood it is a routine task to compensate
for them.) In contrast, it is much harder to predict the weather. We know just as much about the mathematics of
weather as we do about the mathematics of tides, but weather has an inherent unpredictability. Despite this,
meteorologists can make effective short-term predictions of weather patterns- say, three or four days in advance.
The unpredictability of the weather, however, has nothing at all to do with randomness- a topic we will take up in
chapter 8, when we discuss the concept of chaos.

The role of mathematics goes beyond mere prediction. Once you understand how a system works, you don't
have to remain a passive observer. You can attempt to control the system, to make it do what you want. It pays not
to be too ambitious: weather control, for example, is in its infancy-we can't make rain with any great success, even
when there are rainclouds about. Examples of control systems range from the thermostat on a boiler, which keeps it
at a fixed temperature, to the medieval practice of coppicing woodland. Without a sophisticated mathematical
control system, the space shuttle would fly like the brick it is, for no human pilot can respond quickly enough to
correct its inherent instabilities. The use of electronic pacemakers to help people with heart disease is another
example of control.
These examples bring us to the most down-to-earth aspect of mathematics: its practical applications-how
mathematics earns its keep. Our world rests on mathematical foundations, and mathematics is unavoidably
embedded in our global culture. The only reason we don't always realize just how strongly our lives are affected by
mathematics is that, for sensible reasons, it is kept as far as possible behind the scenes. When you go to the travel
agent and book a vacation, you don't need to understand the intricate mathematical and physical theories that make it
possible to design computers and telephone lines, the optimization routines that schedule as many flights as possible
around any particular airport, or the signal-processing methods used to provide accurate radar images for the pilots.
When you watch a television program, you don't need to understand the three-dimensional geometry used to
produce special effects on the screen, the coding methods used to transmit TV signals by satellite, the mathematical
methods used to solve the equations for the orbital motion of the satellite, the thousands of different applications of
mathematics during every step of the manufacture of every component of the spacecraft that launched the satellite
into position. When a farmer plants a new strain of potatoes, he does not need to know the statistical theories of
genetics that identified which genes made that particular type of plant resistant to disease.
But somebody had to understand all these things in the past, otherwise airliners, television, spacecraft, and
disease-resistant potatoes wouldn't have been invented. And some- body has to understand all these things now, too,
otherwise they won't continue to function. And somebody has to be inventing new mathematics in the future, able to
solve problems that either have not arisen before or have hitherto proved intractable, otherwise our society will fall
apart when change requires solutions to new problems or new solutions to old problems. If mathematics, including
everything that rests on it, were somehow suddenly to be withdrawn from our world, human society would collapse
in an instant. And if mathematics were to be frozen, so that it never went a single step farther, our civilization would
start to go backward.
We should not expect new mathematics to give an immediate dollars-and-cents payoff. The transfer of a
mathematical idea into something that can be made in a factory or used in a home generally takes time. Lots of time:
a century is not unusual. In chapter 5, we will see how seventeenth-century interest in the vibrations of a violin
string led, three hundred years later, to the discovery of radio waves and the invention of radio, radar, and television.
It might have been done quicker, but not that much quicker. If you think-as many people in our increasingly
managerial culture do-that the process of scientific discovery can be speeded up by focusing on the application as a
goal and ignoring "curiosity-driven" research, then you are wrong. In fact that very phrase, "curiosity-driven

9
research," was introduced fairly recently by unimaginative bureaucrats as a deliberate put-down. Their desire for
tidy projects offering guaranteed short-term profit is much too simple- minded, because goal-oriented research can
deliver only predictable results. You have to be able to see the goal in order to aim at it. But anything you can see,
your competitors can see, too. The pursuance of safe research will impoverish us all. The really important
breakthroughs are always unpredictable. It is their very unpredictability that makes them important: they change our
world in ways we didn't see coming.
Moreover, goal-oriented research often runs up against a brick wall, and not only in mathematics. For
example, it took approximately eighty years of intense engineering effort to develop the photocopying machine after
the basic principle of xerography had been discovered by scientists. The first fax machine was invented over a
century ago, but it didn't work fast enough or reliably enough. The principle of holography (three-dimensional
pictures, see your credit card) was discovered over a century ago, but nobody then knew how to pro- duce the
necessary beam of coherent light-light with all its waves in step. This kind of delay is not at all unusual in industry,
let alone in more intellectual areas of research, and the impasse is usually broken only when an unexpected new idea
arrives on the scene.
There is nothing wrong with goal-oriented research as a way of achieving specific feasible goals. But the
dreamers and the mavericks must be allowed some free rein, too. Our world is not static: new problems constantly
arise, and old answers often stop working. Like Lewis Carroll's Red Queen, we must run very fast in order to stand
still.

CHAPTER 3
WHAT MATHEMATICS IS ABOUT

When we hear the word "mathematics," the first thing that springs to mind is numbers. Numbers are the heart of
mathematics, an all-pervading influence, the raw materials out of which a great deal of mathematics is forged. But
numbers on their own form only a tiny part of mathematics. I said earlier that we live in an intensely mathematical
world, but that whenever possible the mathematics is sensibly tucked under the rug to make our world "user-
friendly." However, some mathematical ideas are so basic to our world that they cannot stay hidden, and numbers
are an especially prominent example. Without the ability to count eggs and subtract change, for instance, we could
not even buy food. And so we teach arithmetic. To everybody. Like reading and writing, its absence is a major
handicap. And that creates the overwhelming impression that mathematics is mostly a matter of numbers-which isn't
really true. The numerical tricks we learn in arithmetic are only the tip of an iceberg. We can run our everyday lives
without much more, but our culture cannot run our society by using such limited ingredients. Numbers are just one
type of object that mathematicians think about. In this chapter, I will try to show you some of the others and explain
why they, too, are important.

Inevitably my starting point has to be numbers. A large part of the early prehistory of mathematics can be
summed up as the discovery, by various civilizations, of a wider and wider range of things that deserved to be called
numbers. The simplest are the numbers we use for counting. In fact, counting began long before there were symbols
like 1, 2, 3, because it is possible to count without using numbers at all-say, by counting on your fingers. You can
work out that "I have two hands and a thumb of camels" by folding down fingers as your eye glances over the
camels. You don't actually have to have the concept of the number "eleven" to keep track of whether anybody is
stealing your camels. You just have to notice that next time you seem to have only two hands of camels-so a thumb
of camels is missing.
You can also record the count as scratches on pieces of wood or bone. Or you can make tokens to use as
counters clay disks with pictures of sheep on them for counting sheep, or disks with pictures of camels on them for
counting camels. As the animal’s parade past you, you drop tokens into a bag- one token for each animal. The use of
symbols for numbers probably developed about five thousand years ago, when such counters were wrapped in a clay
envelope. It was a nuisance to break open the clay covering every time the accountants wanted to check the contents,
and to make another one when they had finished. So people put special marks on the outside of the envelope
summarizing what was inside. Then they realized that they didn't actually need any counters inside at all: they could
just make the same marks on clay tablets.
It's amazing how long it can take to see the obvious. But of course it's only obvious now.

The next invention beyond counting numbers was fractions-the kind of number we now symbolize as 2/3
(two thirds) or 22/7 (twenty-two sevenths-or, equivalently, three and one-seventh). You can't count with fractions-
although two-thirds of a camel might be edible, it's not countable-but you can do much more interesting things
instead. In particular, if three brothers inherit two camels between them, you can think of each as owning two-thirds

10
of a camel-a convenient legal fiction, one with which we are so comfortable that we forget how curious it is if taken
literally.
Much later, between 400 and 1200 AD, the concept of zero was invented and accepted as denoting a
number. If you think that the late acceptance of zero as a number is strange, bear in mind that for a long time "one"
was not considered a number because it was thought that a number of things ought to be several of them. Many
history books say that the key idea here was the invention of a symbol for "nothing." That may have been the key to
making arithmetic practical; but for mathematics the important idea was the concept of a new kind of number, one
that represented the concrete idea "nothing." Mathematics uses symbols, but it no more is those symbols than music
is musical notation or language is strings of letters from an alphabet. Carl Friedrich Gauss, thought by many to be
the greatest mathematician ever to have lived, once said (in Latin) that what matters in mathematics is "not
notations, but notions." The pun "non notationes, sed notiones" worked in Latin, too.

The next extension of the number concept was the invention of negative numbers. Again, it makes little
sense to think of minus two camels in a literal sense; but if you owe some- body two camels, the number you own is
effectively diminished by two. So a negative number can be thought of as representing a debt. There are many
different ways to interpret these more esoteric kinds of number; for instance, a negative temperature (in degrees
Celsius) is one that is colder than freezing, and an object with negative velocity is one that is moving backward, So
the same abstract mathematical object may represent more than one aspect of nature.
Fractions are all you need for most commercial transactions, but they're not enough for mathematics. For
example, as the ancient Greeks discovered to their chagrin, the square root of two is not exactly representable as a
fraction. That is, if you multiply any fraction by itself, you won't get two exactly. You can get very close-for
example, the square of 17/12 is 289/144, and if only it were 288/144 you would get two. But it isn't, and you don't-
and whatever fraction you try, you never will. The square root of two, usually denoted .,,)2, is therefore said to be
"irrational." The simplest way to enlarge the number system to include the irrationals is to use the so- called real
numbers-a breathtakingly inappropriate name, inasmuch as they are represented by decimals that go on for- ever,
like 3.14159 ... , where the dots indicate an infinite number of digits. How can things be real if you can't even write
them down fully? But the name stuck, probably because real numbers formalize many of our natural visual
intuitions about lengths and distances.

The real numbers are one of the most audacious idealizations made by the human mind, but they were used
happily for centuries before anybody worried about the logic behind them. Paradoxically, people worried a great
deal about the next enlargement of the number system, even though it was entirely harmless. That was the
introduction of square roots for negative numbers, and it led to the "imaginary" and "complex" numbers. A
professional mathematician should never leave home without them, but fortunately nothing in this book will require
a knowledge of complex numbers, so I'm going to tuck them under the mathematical carpet and hope you don't
notice. However, I should point out that it is easy to interpret an infinite decimal as a sequence of ever-finer
approximations to some measurement-say, of a length or a weight-whereas a comfortable interpretation of the
square root of minus one is more elusive.
In current terminology, the whole numbers 0, 1, 2, 3, ... are known as the natural numbers. If negative
whole numbers are included, we have the integers. Positive and negative fractions are called rational numbers. Real
numbers are more general; complex numbers more general still. So here we have five number systems, each more
inclusive than the previous: natural numbers, integers, rationales, real numbers, and complex numbers. In this book,
the important number systems will be the integers and the reals. We'll need to talk about rational numbers every so
often; and as I've just said, we can ignore the complex numbers altogether. But I hope you under- stand by now that
the word "number" does not have any immutable God-given meaning. More than once the scope of that word was
extended, a process that in principle might occur again at any time.
However, mathematics is not just about numbers. We've already had a passing encounter with a different
kind of object of mathematical thought, an operation; examples are addition, subtraction, multiplication, and
division. In general, an operation is something you apply to two (sometimes more) mathematical objects to get a
third object. I also alluded to a third type of mathematical object when I mentioned square roots. If you start with a
number and form its square root, you get another number. The term for such an "object" is function. You can think
of a function as a mathematical rule that starts with a mathematical object-usually a number-and associates to it
another object in a specific manner. Functions are often defined using algebraic formulas, which are just short- hand
ways to explain what the rule is, but they can be defined by any convenient method. Another term with the same
meaning as "function" is transformation: the rule trans- forms the first object into the second. This term tends to be
used when the rules are geometric, and in chapter 6 we will use transformations to capture the mathematical essence
of symmetry.
Operations and functions are very similar concepts. Indeed, on a suitable level of generality there is not
much to distinguish them. Both of them are processes rather than things. And now is a good moment to open up

11
Pandora's box and explain one of the most powerful general weapons in the mathematician's armory, which we
might call the "trinification of processes." (There is a dictionary term, reification, but it sounds pretentious.)
Mathematical "things" have no existence in the real world: they are abstractions. But mathematical processes are
also abstractions, so processes are no less "things" than the "things" to which they are applied. The trinification of
processes is commonplace. In fact, I can make out a very good case that the number "two" is not actually a thing but
a process-the process you carry out when you associate two camels or two sheep with the symbols "1, 2" chanted in
turn. A number is a process that has long ago been thingified so thoroughly that everybody thinks of it as a thing. It
is just as feasible-though less familiar to most of us-to think of an operation or a function as a thing. For example,
we might talk of "square root" as if it were a thing and I mean here not the square root of any particular number, but
the function itself. In this image, the square-foot function is a kind of sausage machine: you stuff a number in at one
end and its square root pops out at the other.
In chapter 6, we will treat motions of the plane or space as if they are things. I'm warning you now because
you may find it disturbing when it happens. However, mathematicians aren't the only people who play the
trinification game. The legal profession talks of "theft" as if it were a thing; it even knows what kind of thing it is-a
crime. In phrases such as "two major evils in Western society are drugs and theft" we find one genuine thing and
one thingified thing, both treated as if they were on exactly the same level. For theft is a process, one whereby my
property is transferred without my agreement to somebody else, but drugs have a real physical existence.
Computer scientists have a useful term for things that can be built up from numbers by thingifying
processes: they call them data structures. Common examples in computer science are lists (sets of numbers written
in sequence) and arrays (tables of numbers with several rows and columns). I've already said that a picture on a
computer screen can be represented as a list of pairs of numbers; that's a more complicated but entirely sensible data
structure. You can imagine much more complicated possibilities-arrays that are tables of lists, not tables of numbers;
lists of arrays; arrays of arrays; lists of lists of arrays of lists .... Mathematics builds its basic objects of thought in a
similar manner. Back in the days when the logical foundations of mathematics were still being sorted out, Bertrand
Russell and Alfred North Whitehead wrote an enormous three-volume work, Principia Mathematica, which began
with the simplest possible logical ingredient-the idea of a set, a collection of things. They then showed how to build
up the rest of mathematics. Their main objective was to analyze the logical structure of mathematics, but a major
part of their effort went into devising appropriate data structures for the important objects of mathematical thought.
The image of mathematics raised by this description of its basic objects is something like a tree, rooted in
numbers and branching into ever more esoteric data structures as you proceed from trunk to bough, bough to limb,
limb to twig .... But this image lacks an essential ingredient. It fails to describe how mathematical concepts interact.
Mathematics is not just a collection of isolated facts: it is more like a landscape; it has an inherent geography that its
users and creators employ to navigate through what would otherwise be an impenetrable jungle. For instance, there
is a metaphorical feeling of distance. Near any particular mathematical fact we find other, related facts. For example,
the fact that the circumference of a circle is 1t (pi) times its diameter is very close to the fact that the circumference
of a circle is 21t times its radius. The connection between these two facts is immediate: the diameter is twice the
radius. In contrast, unrelated ideas are more distant from each other; for example, the fact that there are exactly six
different ways to arrange three objects in order is a long way away from facts about circles. There is also a
metaphorical feeling of prominence. Soaring peaks pierce the sky- important ideas that can be used widely and seen
from far away, such as Pythagoras's theorem about right triangles, or the basic techniques of calculus. At every turn,
new vistas arise-an unexpected river that must be crossed using step- ping stones, a vast, tranquil lake, an impassable
crevasse. The user of mathematics walks only the well-trod parts of this mathematical territory. The creator of
mathematics explores its unknown mysteries, maps them, and builds roads through them to make them more easily
accessible to everybody else.
The ingredient that knits this landscape together is proof Proof determines the route from one fact to
another. To professional mathematicians, no statement is considered valid unless it is proved beyond any possibility
of logical error. But there are limits to what can be proved, and how it can be proved. A great deal of work in
philosophy and the foundations of mathematics has established that you can't prove everything, because you have to
start somewhere; and even when you've decided where to start, some statements may be neither provable nor
disprovable. I don't want to explore those issues here; instead, I want to take a pragmatic look at what proofs are and
why they are needed.
Textbooks of mathematical logic say that a proof is a sequence of statements, each of which either follows
from previous statements in the sequence or from agreed axioms- unproved but explicitly stated assumptions that in
effect define the area of mathematics being studied. This is about as informative as describing a novel as a sequence
of sentences, each of which either sets up an agreed context or follows credibly from previous sentences. Both
definitions miss the essential point: that both a proof and a novel must tell an interesting story. They do capture a
secondary point, that the story must be convincing, and they also describe the overall format to be used, but a good
story line is the most important feature of all.

12
 Very few textbooks say that.

Most of us are irritated by a movie riddled with holes, however polished its technical production may be. I
saw one recently in which an airport is taken over by guerrillas who shut down the electronic equipment used by the
control tower and substitute their own. The airport authorities and the hero then spend half an hour or more of movie
time-several hours of story time-agonizing about their inability to communicate with approaching aircraft, which are
stacking up in the sky overhead and running out of fuel. It occurs to no one that there is a second, fully functioning
airport no more than thirty miles away, nor do they think to telephone the nearest Air Force base. The story was
brilliantly and expensively filmed-and silly.

That didn't stop a lot of people from enjoying it: their critical standards must have been lower than mine.
But we all have limits to what we are prepared to accept as credible. If in an otherwise realistic film a child saved
the day by picking up a house and carrying it away, most of us would lose interest. Similarly, a mathematical proof
is a story about mathematics that works. It does not have to dot every i and cross every t; readers are expected to fill
in routine steps for themselves-just as movie characters may suddenly appear in new surroundings without it being
necessary to show how they got there. But the story must not have gaps, and it certainly must not have an
unbelievable plot line. The rules are stringent: in mathematics, a single flaw is fatal. Moreover, a subtle flaw can be
just as fatal as an obvious one.

Let's take a look at an example. I have chosen a simple one, to avoid technical background; in consequence,
the proof tells a simple and not very significant story. I stole it from a colleague, who calls it the SHIP/DOCK
Theorem. You probably know the type of puzzle in which you are given one word (SHIP) and asked to turn it into
another word (DOCK) by changing one letter at a time and getting a valid word at every stage. You might like to try
to solve this one before reading on: if you do, you will probably understand the theorem, and its proof, more easily.

Here's one solution:

SHIP
SLIP
SLOP
SLOT
SOOT
LOOT
LOOK
LOCK
DOCK
There are plenty of alternatives, and some involve fewer words. But if you play around with this problem,
you will eventually notice that all solutions have one thing in common: at least one of the intermediate words must
contain two vowels.

O.K., so prove it.

I'm not willing to accept experimental evidence. I don't care if you have a hundred solutions and every
single one of them includes a word with two vowels. You won't be happy with such evidence, either, because you
will have a sneaky feeling that you may just have missed some really clever sequence that doesn't include such a
word. On the other hand, you will probably also have a distinct feeling that somehow "it's obvious." I agree; but why
is it obvious?
You have now entered a phase of existence in which most mathematicians spend most of their time:
frustration. You know what you want to prove, you believe it, but you don't see a convincing story line for a proof.
What this means is that you are lacking some key idea that will blow the whole problem wide open. In a moment I'll
give you a hint. Think about it for a few minutes, and you will probably experience a much more satisfying phase of
the mathematician's existence: illumination.

Here's the hint. Every valid word in English must contain a vowel.

It's a very simple hint. First, convince yourself that it's true. (A dictionary search is acceptable, provided it's
a big dictionary.) Then consider its implications ....

13
 O.K., either you got it or you've given up. Whichever of these you did, all professional mathematicians
have done the same on a lot of their problems. Here's the trick. You have to concentrate on what happens to the
vowels. Vowels are the peaks in the SHIP/DOCK landscape, the landmarks between which the paths of proof wind.
In the initial word SHIP there is only one vowel, in the third position. In the final word DOCK there is also
only one vowel, but in the second position. How does the vowel change position? There are three possibilities. It
may hop from one location to the other; it may disappear altogether and reappear later on; or an extra vowel or
vowels may be created and subsequently eliminated.
The third possibility leads pretty directly to the theorem. Since only one letter at a time changes, at some
stage the word must change from having one vowel to having two. It can't leap from having one vowel to having
three, for example. But what about the other possibilities? The hint that I mentioned earlier tells us that the single
vowel in SHIP cannot disappear altogether. That leaves only the first possibility: that there is always one vowel, but
it hops from position 3 to position 2. However, that can't be done by changing only one letter! You have to move, in
one step, from a vowel at position 3 and a consonant at position 2 to a consonant at position 3 and a vowel at
position 2. That implies that two letters must change, which is illegal. Q.E.D., as Euclid used to say.
A mathematician would write the proof out in a much more formal style, something like the textbook
model, but the important thing is to tell a convincing story. Like any good story, it has a beginning and an end, and a
story line that gets you from one to the other without any logical holes appearing. Even though this is a very simple
example, and it isn't standard mathematics at all, it illustrates the essentials: in particular, the dramatic difference
between an argument that is genuinely convincing and a hand-waving argument that sounds plausible but doesn't
really gel. I hope it also put you through some of the emotional experiences of the creative mathematician:
frustration at the intractability of what ought to be an easy question, elation when light dawned, suspicion as you
checked whether there were any holes in the argument, aesthetic satisfaction when you decided the idea really was
O.K. and realized how neatly it cut through all the apparent complications. Creative mathematics is just like this-but
with more serious subject matter.
Proofs must be convincing to be accepted by mathematicians. There have been many cases where extensive
numerical evidence suggested a completely wrong answer. One notorious example concerns prime numbers-
numbers that have no divisors except themselves and 1. The sequence of primes begins 2, 3, 5, 7, 11, 13, 17, 19 and
goes on forever. Apart from 2, all primes are odd; and the odd primes fall into two classes: those that are one less
than a multiple of four (such as 3, 7, 11, 19) and those that are one more than a multiple of four (such as 5, 13, 17).
If you run along the sequence of primes and count how many of them fall into each class, you will observe that there
always seem to be more primes in the "one less" class than in the "one more" class. For example, in the list of the
seven pertinent primes above, there are four primes in the first class but only three in the second. This pattern
persists for numbers up to at least a trillion, and it seems entirely reasonable to conjecture that it is always true.

However, it isn't.

By indirect methods, number theorists have shown that when the primes get sufficiently big, the pattern
changes and the "one more than a multiple of four" class goes into the lead. The first proof of this fact worked only
when the numbers got bigger than 10'10'10'10'46, where to avoid giving the printer kittens I've used the ' sign to
indicate forming a power. This number is utterly gigantic. Written out in full, it would go 10000 ... 000, with a very
large number of 0s. If all the matter in the universe were turned into paper, and a zero could be inscribed on every
electron, there wouldn't be enough of them to hold even a tiny fraction of the necessary zeros.
No amount of experimental evidence can account for the possibility of exceptions so rare that you need
numbers that big to locate them. Unfortunately, even rare exceptions matter in mathematics. In ordinary life, we
seldom worry about things that might occur on one occasion out of a trillion. Do you worry about being hit by a
meteorite? The odds are about one in a trillion. But mathematics piles logical deductions on top of each other, and if
any step is wrong the whole edifice may tumble. If you have stated as a fact that all numbers behave in some
manner, and there is just one that does not, then you are wrong, and everything you have built on the basis of that
incorrect fact is thrown into doubt.
Even the very best mathematicians have on occasion claimed to have proved something that later turned
out not to be so-their proof had a subtle gap, or there was a simple error in a calculation, or they inadvertently
assumed something that was not as rock-solid as they had imagined. So, over the centuries, mathematicians have
learned to be extremely critical of proofs. Proofs knit the fabric of mathematics together, and if a single thread is
weak, the entire fabric may unravel.

14
CHAPTER 4
THE CONSTANTS OF CHANGES

For a good many centuries, human thought about nature has swung between two opposing points of view. According
to one view, the universe obeys fixed, immutable laws, and everything exists in a well-defined objective reality. The
opposing view is that there is no such thing as objective reality; that all is flux, all is change. As the Greek
philosopher Heraclitus put it, "You can't step into the same river twice." The rise of science has largely been
governed by the first viewpoint. But there are increasing signs that the prevailing
cultural background is starting to switch to the second ways of thinking as diverse as postmodernism, cyberpunk,
and chaos theory all blur the alleged objectiveness of reality and reopen the ageless debate about rigid laws and
flexible change.

What we really need to do is get out of this futile game altogether. We need to find a way to step back from
these opposing worldviews-not so much to seek a synthesis as to see them both as two shadows of some higher
order of reality-shadows that are different only because the higher order is being seen from two different directions.
But does such a higher order exist, and if so, is it accessible? To many-especially scientists-Isaac Newton represents
the triumph of rationality over mysticism. The famous economist John Maynard Keynes, in his essay Newton, the
Man, saw things differently:
In the eighteenth century and since, Newton came to be thought of as the first and greatest of the modern
age of scientists, a rationalist, one who taught us to think on the lines of cold and untinctured reason. I do
not see him in this light. I do not think that anyone who has pored over the contents of that box which he
packed up when he finally left Cambridge in 1696 and which, though partly dispersed, have come down to
us, can see him like that. Newton was not the first of the age of reason. He was the last of the magicians,
the last of the Babylonians and Sumerians, the last great mind which looked out on the visible and
intellectual world with the same eyes as those who began to build our intellectual inheritance rather less
than 10,000 years ago. Isaac Newton, a posthumous child born with no father on Christmas Day, 1642, was
the last wonder-child to whom the Magi could do sincere and appropriate homage.

Keynes was thinking of Newton's personality, and of his interests in alchemy and religion as well as in
mathematics and physics. But in Newton's mathematics we also find the first significant step toward a worldview
that transcends and unites both rigid law and flexible flux. The universe may appear to be a storm-tossed ocean of
change, but Newton- and before him Galileo and Kepler, the giants upon whose shoulders he stood-realized that
change obeys rules. Not only can law and flux coexist, but law generates flux.

Today's emerging sciences of chaos and complexity supply the missing converse: flux generates law. But
that is another story, reserved for the final chapter.

Prior to Newton, mathematics had offered an essentially static model of nature. There are a few exceptions,
the most obvious being Ptolemy's theory of planetary motion, which reproduced the observed changes very
accurately using a system of circles revolving about centers that themselves were attached to revolving circles-
wheels within wheels within wheels. But at that time the perceived task of mathematics was to discover the
catalogue of "ideal forms" employed by nature. The circle was held to be the most perfect shape possible, on the
basis of the democratic observation that every point on the circumference of a circle lies at the same distance from
its center. Nature, the creation of higher beings, is by definition perfect, and ideal forms are mathematical perfection,
so of course the two go together. And perfection was thought to be unblemished by change.

Kepler challenged that view by finding ellipses in place of complex systems of circles. Newton threw it out
altogether, replacing forms by the laws that produce them.

Although its ramifications are immense, Newton's approach to motion is a simple one. It can be illustrated
using the motion of a projectile, such as a cannonball fired from a gun at an angle. Galileo discovered
experimentally that the path of such a projectile is a parabola, a curve known to the ancient Greeks and related to the
ellipse. In this case, it forms an inverted V-shape. The parabolic path can be most easily understood by decomposing
the projectile's motion into two independent components: motion in a horizontal direction and motion in a vertical
direction. By thinking about these two types of motion separately and putting them back together only when each
has been understood in its own right, we can see why the path should be a parabola.

The cannonball's motion in the horizontal direction, parallel to the ground, is very simple: it takes place at a
constant speed. Its motion in the vertical direction is more interesting. It starts moving upward quite rapidly, then it

15
slows down, until for a split second it appears to hang stationary in the air; then it begins to drop, slowly at first but
with rapidly increasing velocity.

Newton's insight was that although the position of the cannonball changes in quite a complex way, its
velocity changes in a much simpler way, and its acceleration varies in a very simple manner indeed. Figure 2
summarizes the relationship between these three functions, in the following example. Suppose for the sake of
illustration that the initial upward velocity is fifty meters per second (50 m/sec). Then the height of the cannonball
above ground, at one-second intervals, is: 0,45,80,105,120,125,120,105,80,45,0.

You can see from these numbers that the ball goes up, levels off near the top, and then goes down again.
But the general pattern is not entirely obvious. The difficulty was com- pounded in Galileo's time-and, indeed, in
Newton's because it was hard to measure these numbers directly. In actual fact, Galileo rolled a ball up a gentle
slope to slow the whole process down. The biggest problem was to measure time accurately: the historian Stillman
Drake has suggested that perhaps Galileo hummed tunes to himself and subdivided the basic beat in his head, as a
musician does.

The pattern of distances is a puzzle, but the pattern of velocities is much clearer. The ball starts with an
upward velocity of 50 m/sec. One second later, the velocity has decreased to (roughly) 40 m/sec; a second after that,
it is 30m/sec; then 20 m/sec, 10 m/sec, then a m/sec (stationary). A second after that, the velocity is 10 m/sec
downward. Using negative numbers, we can think of this as an upward velocity of -10 m/sec. In successive seconds,
the pattern continues: -20m/sec, -30 m/sec, -40 m/sec, -50 m/sec. At this point, the cannonball hits the ground. So
the sequence of velocities, measured at one-second intervals, is: 0,40, 30, 20, 10, 0, -10, -20, -30, -40, -50.

Figure 2: Calculus in a nutshell. Three mathematical patterns determined by a cannonball: height, velocity,
and acceleration. The pattern of heights, which is what we naturally observe, is complicated. Newton realized that
the pattern of velocities is simpler, while the pattern of acceleration is simpler still. The two basic operations of
calculus, differentiation and integration, let us pass from any if these patterns to any other. So we can work with the
simplest acceleration, and deduce the one we really want – height.

16
 Now there is a pattern that can hardly be missed; but let's go one step further by looking at accelerations.
The corresponding sequence for the acceleration of the cannonball, again using negative numbers to indicate
downward motion, is -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10.

I think you will agree that the pattern here is extremely simple. The ball undergoes a constant downward
acceleration of 10 m/sec2 • (The true figure is about 9.81 m/sec2, depending on whereabouts on the Earth you
perform the experiment. But 10 is easier to think about.)

How can we explain this constant that is hiding among the dynamic variables? When all else is flux, why is
the acceleration fixed? One attractive explanation has two elements. The first is that the Earth must be pulling the
ball downward; that is, there is a gravitational force that acts on the ball. It is reasonable to expect this force to
remain the same at different heights above the ground. Indeed, we feel weight because gravity pulls our bodies
downward, and we still weigh the same if we stand at the top of a tall building. Of course, this appeal to everyday
observation does not tell us what happens if the distance becomes sufficiently large-say the distance that separates
the Moon from the Earth. That's a different story, to which we shall return shortly.

The second element of the explanation is the real break- through. We have a body moving under a constant
downward force, and we observe that it undergoes a constant downward acceleration. Suppose, for the sake of
argument, that the pull of gravity was a lot stronger: then we would expect the downward acceleration to be a lot
stronger, too. Without going to a heavy planet, such as Jupiter, we can't test this idea, but it looks reasonable; and it's
equally reasonable to suppose that on Jupiter the downward acceleration would again be constant-but a different
constant from what it is here. The simplest theory consistent with this mixture of real experiments and thought
experiments is that when a force acts on a body, the body experiences an acceleration that is proportional to that
force. And this is the essence of Newton's law of motion. The only missing ingredients are the assumption that this
is always true, for all bodies and for all forces, whether or not the forces remain constant; and the identification of
the constant of proportionality as being related to the mass of the body. To be precise, Newton's law of motion states
that mass x acceleration = force.
That's it. Its great virtue is that it is valid for any system of masses and forces, including masses and forces that
change over time. We could not have anticipated this universal applicability from the argument that led us to the
law; but it turns out to be so.

Newton stated three laws of motion, but the modern approach views them as three aspects of a single
mathematical equation. So, I will use the phrase "Newton's law of motion" to refer to the whole package.

The mountaineer's natural urge when confronted with a mountain is to climb it; the mathematician's natural
urge when confronted with an equation is to solve it. But how? Given a body's mass and the forces acting on it, we
can easily solve this equation to get the acceleration. But this is the answer to the wrong question. Knowing that the
acceleration of a cannonball is always -10 m/sec2 doesn't tell us anything obvious about the shape of its trajectory.
This is where the branch of mathematics known as calculus comes in; indeed, it is why Newton (and Leibniz)
invented it. Calculus provides a technique, which nowadays is called integration, that allows us to move from
knowledge of acceleration at any instant to knowledge of velocity at any instant. By repeating the same trick, we can
then obtain knowledge of position at any instant. And that is the answer to the right question.

As I said earlier, velocity is rate of change of position, and acceleration is rate of change of velocity.
Calculus is a mathematical scheme invented to handle questions about rates of change. In particular, it provides a
technique for finding rates of change-a technique known as differentiation. Integration "undoes" the effect of
differentiation; and integrating twice undoes the effect of differentiating twice. Like the twin faces of the Roman
god Janus, these twin techniques of calculus point in opposite directions. Between them, they tell you that if you
know anyone of the functions-position, velocity, or acceleration-at every instant, then you can work out the other
two.
Newton's law of motion teaches an important lesson: namely, that the route from nature's laws to nature's
behavior need not be direct and obvious. Between the behavior we observe and the laws that produce it is a crevasse,
which the human mind can bridge only by mathematical calculations. This is not to suggest that nature is
mathematics-that (as the physicist Paul Dirac put it) "God is a mathematician." Maybe nature's patterns and
regularities have other origins; but, at the very least, mathematics is an extremely effective way for human beings to
come to grips with those patterns.

All of the laws of physics that were discovered by pursuing Isaac Newton's basic insight-that change in
nature can be described by mathematical processes, just as form in nature can be described by mathematical things-

17
have a similar character. The laws are formulated as equations that relate not the physical quantities of primary
interest but the rates at which those quantities change with time, or the rates at which those rates change with time.
For example, the "heat equation," which determines how heat flows through a conducting body, is all about the rate
of change of the body's temperature; and the "wave equation," which governs the motion of waves in water, air, or
other materials, is about the rate of change of the rate of change of the height of the wave. The physical laws for
light, sound, electricity, magnetism, the elastic bending of materials, the flow of fluids, and the course of a chemical
reaction, are all equations for various rates of change.

Because a rate of change is about the difference between some quantity now and its value an instant into
the future, equations of this kind are called differential equations. The term "differentiation" has the same origin.
Ever since Newton, the strategy of mathematical physics has been to describe the universe in terms of differential
equations, and then solve them.

However, as we have pursued this strategy into more sophisticated realms, the meaning of the word
"solve" has undergone a series of major changes. Originally it implied finding a precise mathematical formula that
would describe what a system does at any instant of time. Newton's discovery of another important natural pattern,
the law of gravitation, rested upon a solution of this kind. He began with Kepler's discovery that planets move in
ellipses, together with two other mathematical regularities that were also noted by Kepler. Newton asked what kind
of force, acting on a planet, would be needed to produce the pattern that Kepler had found. In effect, Newton was
trying to work backward from behavior to laws, using a process of induction rather than deduction. And he
discovered a very beautiful result. The necessary force should always point in the direction of the Sun; and it should
decrease with the distance from the planet to the Sun. Moreover, this decrease should obey a simple mathematical
law, the inverse-square law. This means that the force acting on a planet at, say, twice the distance is reduced to
one-quarter, the force acting on a planet at three times the distance is reduced to one-ninth, and so on. From this
discovery-which was so beautiful that it surely concealed a deep truth about the world-it was a short step to the
realization that it must be the Sun that causes the force in the first place. The Sun attracts the planet, but the
attraction becomes weaker if the planet is farther away. It was a very appealing idea, and Newton took a giant
intellectual leap: he assumed that the same kind of attractive force must exist between any two bodies whatsoever,
anywhere in the universe.

And now, having "induced" the law for the force, Newton could bring the argument full circle by deducing
the geometry of planetary motion. He solved the equations given by his laws of motion and gravity for a system of
two mutually attracting bodies that obeyed his inverse-square law; in those days, "solved" meant finding a
mathematical formula for their motion. The formula implied that they must move in ellipses about their common
center of mass. As Mars moves around the Sun in a giant ellipse, the Sun moves in an ellipse so tiny that its motion
goes undetected. Indeed, the Sun is so massive compared to Mars that the mutual center of mass lies beneath the
Sun's surface, which explains why Kepler thought that Mars moved in an ellipse around the stationary Sun.

However, when Newton and his successors tried to build on this success by solving the equations for a
system of three or more bodies-such as Moon/Earth/Sun, or the entire Solar System-they ran into technical trouble;
and they could get out of trouble only by changing the meaning of the word "solve." They failed to find any
formulas that would solve the equations exactly, so they gave up looking for them. Instead, they tried to find ways to
calculate approximate numbers. For example, around 1860 the French astronomer Charles-Eugene Delaunay filled
an entire book with a single approximation to the motion of the Moon, as influenced by the gravitational attractions
of the Earth and the Sun. It was an extremely accurate approximation-which is why it filled a book-and it took him
twenty years to work it out. When it was subsequently checked, in 1970, using a symbolic-algebra computer
program, the calculation took a mere twenty hours: only three mistakes were found in Delaunay's work, none
serious.

The motion of the Moon/Earth/Sun system is said to be a three-body problem-for evident reasons. It is so
unlike the nice, tidy two-body problem Newton solved that it might as well have been invented on another planet in
another galaxy, or in another universe. The three-body problem asks for a solution for the equations that describe the
motion of three masses under inverse-square-Iaw gravity. Mathematicians tried to find such a solution for centuries
but met with astonishingly little success beyond approximations, such as Delaunay's, which worked only for
particular cases, like Moon/Earth/Sun. Even the so-called restricted three-body problem, in which one body has a
mass so small that it can be considered to exert no force at all upon the other two, proved utterly intractable. It was
the first serious hint that knowing the laws might not be enough to understand how a system behaves; that the
crevasse between laws and behavior might not always be bridgeable.

18
 Despite intensive effort, more than three centuries after Newton we still do not have a complete answer to
the three-body problem. However, we finally know why the problem has been so hard to crack. The two-body
problem is "integrable"-the laws of conservation of energy and momentum restrict solutions so much that they are
forced to take a simple mathematical form. In 1994, Zhihong Xia, ofthe Georgia Institute of Technology, proved
what mathematicians had long suspected: that a system of three bodies is not integrable. Indeed, he did far more, by
showing that such a system can exhibit a strange phenomenon known as Arnold diffusion, first discovered by
Vladimir Arnold, of Moscow State University, in 1964. Arnold diffusion produces an extremely slow, "random"
drift in the relative orbital positions. This drift is not truly random: it is an example of the type of behavior now
known as chaos-which can be described as apparently random behavior with purely deterministic causes.
Notice that this approach again changes the meaning of "solve." First that word meant "find a formula."
Then its meaning changed to "find approximate numbers." Finally, it has in effect become "tell me what the
solutions look like." In place of quantitative answers, we seek qualitative ones. In a sense, what is happening looks
like a retreat: if it is too hard to find a formula, then try an approximation; if approximations aren't available, try a
qualitative description. But it is wrong to see this development as a retreat, for what this change of meaning has
taught us is that for questions like the three-body problem, no formulas can exist. We can prove that there are
qualitative aspects to the solution that a formula cannot capture. The search for a formula in such questions was a
hunt for a mare's nest.
Why did people want a formula in the first place? Because in the early days of dynamics, that was the only
way to work out what kind of motion would occur. Later, the same information could be deduced from
approximations. Nowadays, it can be obtained from theories that deal directly and precisely with the main
qualitative aspects of the motion. As we will see in the next few chapters, this move toward an explicitly qualitative
theory is not a retreat but a major advance. For the first time, we are starting to understand nature's patterns in their
own terms.

CHAPTER 5
FROM VIOLINS TO VIDEOS

It has become conventional, as I have noted, to separate mathematics into two distinct subdisciplines labeled pure
mathematics and applied mathematics. This is a separation that would have baffled the great mathematicians of
classical times. Carl Friedrich Gauss, for example, was happiest in the ivory tower of number theory, where he
delighted in abstract numerical patterns simply because they were beautiful and challenging. He called number
theory "the queen of mathematics," and the poetic idea that queens are delicate beauties who do not sully their hands
with anything useful was not far from his mind. However, he also calculated the orbit of Ceres, the first asteroid to
be discovered. Soon after its discovery, Ceres passed behind the Sun, as seen from Earth, and could no longer be
observed. Unless its orbit could be calculated accurately, astronomers would not be able to find it when it again
became visible, months later. But the number of observations of the asteroid was so small that the standard methods
for calculating orbits could not provide the required level of accuracy. So, Gauss made several major innovations,
some of which remain in use to this day. It was a virtuoso performance, and it made his public reputation. Nor was
that his only practical application of his subject: among other things, he was also responsible for major
developments in surveying, telegraphy, and the understanding of magnetism.

In Gauss's time, it was possible for one person to have a fairly good grasp of the whole of mathematics. But
because all of the classical branches of science have grown so vast that no single mind can likely encompass even
one of them, we now live in an age of specialists. The organizational aspects of mathematics function more tidily if
people specialize either in the theoretical areas of the subject or its practical ones. Because most people feel happier
working in one or the other of these two styles, individual preferences tend to reinforce this distinction.
Unfortunately, it is then very tempting for the outside world to assume that the only useful part of mathematics is
applied mathematics; after all, that is what the name seems to imply. This assumption is correct when it comes to
established mathematical techniques: anything really useful inevitably ends up being considered "applied," no
matter what its origins may have been. But it gives a very distorted view of the origins of new mathematics of
practical importance. Good ideas are rare, but they come at least as often from imaginative dreams about the internal
structure of mathematics as they do from attempts to solve a specific, practical problem. This chapter deals with a
case history of just such a development, whose most powerful application is television-an invention that arguably
has changed our world more than any other. It is a story in which the pure and applied aspects of mathematics
combine to yield something far more powerful and compelling than either could have produced alone. And it begins
at the start of the sixteenth century, with the problem of the vibrating violin string. Although this may sound like a
practical question, it was studied mainly as an exercise in the solution of differential equations; the work was not
aimed at improving the quality of musical instruments.

19
 Imagine an idealized violin string, stretched in a straight line between two fixed supports. If you pluck the
string, pulling it away from the straight-line position and then letting go, what happens? As you pull it sideways, its
elastic tension increases, which produces a force that pulls the string back toward its original position. When you let
go, it begins to accelerate under the action of this force, obeying Newton's law of motion. However, when it returns
to its initial position it is moving rapidly, because it has been accelerating the whole time-so it overshoots the
straight line and keeps moving. Now the tension pulls in the opposite direction, slowing it down until it comes to a
halt. Then the whole story starts over. If there is no friction, the string will vibrate from side to side forever.

That's a plausible verbal description; one of the tasks for a mathematical theory is to see whether this
scenario really holds good, and if so, to work out the details, such as the shape that the string describes at any
instant. It's a complex problem, because the same string can vibrate in many different ways, depending upon how it
is plucked. The ancient Greeks knew this, because their experiments showed that a vibrating string can produce
many different musical tones. Later generations realized that the pitch of the tone is determined by the frequency of
vibration-the rate at which the string moves to and fro-so the Greek discovery tells us that the same string can
vibrate at many different frequencies. Each frequency corresponds to a different configuration of the moving string,
and the same string can take up many different shapes.

Strings vibrate much too fast for the naked eye to see anyone instantaneous shape, but the Greeks found
important evidence for the idea that a string can vibrate at many different frequencies. They showed that the pitch
depends on the positions of the nodes-places along the length of the string which remain stationary. You can test this
on a violin, banjo, or guitar. When the string is vibrating in its "fundamental" frequency-that is, with the lowest
possible pitch-only the end points are at rest. If you place a finger against the center of the string, creating a node,
and then pluck the string, it produces a note one octave higher. If you place your finger one-third of the way along
the string, you actually create two nodes (the other being two-thirds of the way along, and this produces a yet higher
note. The more nodes, the higher the frequency. In general, the number of nodes is an integer, and the nodes are
equally spaced.

The corresponding vibrations are standing waves, meaning waves that move up and down but do not travel
along the string. The size of the up-and-down movement is known as the amplitude of the wave, and this determines
the tone's loudness. The waves are sinusoidal-shaped like a sine curve, a repetive wavy line of rather elegant shape
that arises in trigonometry.

In 1714, the English mathematician Brook Taylor published the fundamental vibrational frequency of a
violin string in terms of its length, tension, and density. In 1746, the Frenchman Jean Le Rond d'Alembert showed
that many vibrations of a violin string are not sinusoidal standing waves. In fact, he proved that the instantaneous
shape of the wave can be anything you like. In 1748, in response to d'Alembert's work, the prolific Swiss
mathematician Leonhard Euler worked out the "wave equation" for a string. In the spirit of Isaac Newton, this is a
differential equation that governs the rate of change of the shape of the string. In fact it is a "partial differential
equation," meaning that it involves not only rates of change relative to time but also rates of change relative to
space-the direction along the string. It expresses in mathematical language the idea that the acceleration of each tiny
segment of the string is proportional to the tensile forces acting upon that segment; so it is a consequence of
Newton's law of motion.

Not only did Euler formulate the wave equation: he solved it. His solution can be described in words. First,
deform the string into any shape you care to choose-a parabola, say, or a triangle, or a wiggly and irregular curve of
your own devising. Then imagine that shape propagating along the string toward the right. Call this a rightward-
traveling wave. Then turn the chosen shape upside down, and imagine it propagating the other way, to create a
leftward-traveling wave. Finally, superpose these two waveforms. This process leads to all possible solutions of the
wave equation in which the ends of the string remain fixed.

Almost immediately, Euler got into an argument with Daniel Bernoulli, whose family originally hailed
from Antwerp but had moved to Germany and then Switzerland to escape religious persecution. Bernoulli also
solved the wave equation, but by a totally different method. According to Bernoulli, the most general solution can be
represented as a superposition of infinitely many sinusoidal standing waves. This apparent disagreement began a
century-long controversy, eventually resolved by declaring both Euler and Bernoulli right. The reason that they are
both right is that every periodically varying shape can be represented as a superposition of an infinite number of sine
curves. Euler thought that his approached to a greater variety of shapes, because he didn't recognize their periodicity.
However, the mathematical analysis works with an infinitely long curve. Because the only part of the curve that

20
matters is the part between the two endpoints, it can be repeated periodically along a very long string without any
essential change. So Euler's worries were unfounded.
The upshot of all this work, then, is that the sinusoidal waves are the basic vibrational components. The
totality of vibrations that can occur is given by forming all possible sums of finitely or infinitely many sinusoidal
waves of all possible amplitudes. As Daniel Bernoulli had maintained all along, "all new curves given by
d' Alembert and Euler are only combinations of the Taylor vibrations."

With the resolution of this controversy, the vibrations of a violin string ceased to be a mystery, and the
mathematicians went hunting for bigger game. A violin string is a curve-a one-dimensional object-but objects with
more dimensions can also vibrate. The most obvious musical instrument that employs a two-dimensional vibration is
the drum, for a drumskin is a surface, not a straight line. So mathematicians turned their attention to drums, starting
with Euler in 1759. Again, he derived a wave equation, this one describing how the displacement of the drums kin in
the vertical direction varies over time. Its physical interpretation is that the acceleration of a small piece of the drums
kin is proportional to the average tension exerted on it by all nearby parts of the drumskin: symbolically, it looks
much like the one-dimensional wave equation; but now there are spatial (second order) rates of change in two
independent directions, as well as the temporal rate of change.

Violin strings have fixed ends. This "boundary condition" has an important effect: it determines which
solutions to the wave equation are physically meaningful for a violin string. In this whole subject, boundaries are
absolutely crucial. Drums differ from violin strings not only in their dimensionality but in having a much more
interesting boundary: the boundary of a drum is a closed curve, or circle. However, like the boundary of a string, the
boundary of the drum is fixed: the rest of the drumskin can move, but its rim is firmly strapped down. This boundary
condition restricts the possible motions of the drumskin. The isolated endpoints of a violin string are not as
interesting and varied a boundary condition as a closed curve is; the true role of the boundary becomes apparent only
in two or more dimensions.
As their understanding of the wave equation grew, the mathematicians of the eighteenth century learned to
solve the wave equation for the motion of drums of various shapes. But now the wave equation began to move out
of the musical domain to establish itself as an absolutely central feature of mathematical physics. It is probably the
single most important mathematical formula ever devised-Einstein's famous relation between mass and energy
notwithstanding. What happened was a dramatic instance of how mathematics can lay bare the hidden unity of
nature. The same equation began to show up everywhere. It showed up in fluid dynamics, where it described the
formation and motion of water waves. It showed up in the theory of sound, where it described the transmission of
sound waves-vibrations of the air, in which its molecules become alternately compressed and separated. And then it
showed up in the theories of electricity and magnetism, and changed human culture forever.
Electricity and magnetism have a long, complicated history, far more complex than that of the wave
equation, involving accidental discoveries and key experiments as well as mathematical and physical theories, Their
story begins with William Gilbert, physician to Elizabeth I, who described the Earth as a huge magnet and observed
that electrically charged bodies can attract or repel each other. It continues with such people as Benjamin Franklin,
who in 1752 proved that lightning is a form of electricity by flying a kite in a thunderstorm; Luigi Galvani, who
noticed that electrical sparks caused a dead frog's leg muscles to contract; and Alessandro Volta, who invented the
first battery. Throughout much of this early development, electricity and magnetism were seen as two quite distinct
natural phenomena. The person who set their unification in train was the English physicist and chemist Michael
Faraday. Faraday was employed at the Royal Institution in London, and one of his jobs was to devise a weekly
experiment to entertain its scientifically minded members. This constant need for new ideas turned Faraday into one
of the greatest experimental physicists of all time. He was especially fascinated by electricity and magnetism
because he knew that an electric current could create a magnetic force. He spent ten years trying to prove that,
conversely, a magnet could produce an electric current, and in 1831 he succeeded. He had shown that magnetism
and electricity were two different aspects of the same thing-electromagnetism. It is said that King William IV asked
Faraday what use his scientific parlor tricks were, and received the reply "I do not know, Your Majesty, but I do
know that one day you will tax them." In fact, practical uses soon followed, notably the electric motor (electricity
creates magnetism creates motion) and the electrical generator (motion creates magnetism creates electricity). But
Faraday also advanced the theory of electromagnetism. Not being a mathematician, he cast his ideas in physical
imagery, of which the most important was the idea of a line of force. If you place a magnet under a sheet of paper
and sprinkle iron filings on top, they will line up along well-defined curves. Faraday's interpretation of these curves
was that the magnetic force did not act "at a distance" without any intervening medium; instead, it propagated
through space along curved lines. The same went for electrical force.

Faraday was no mathematician, but his intellectual successor James Clerk Maxwell was. Maxwell
expressed Faraday's ideas about lines of force in terms of mathematical equations for magnetic and electric fields-

21
that is, distributions of magnetic and electrical charge throughout space. By 1864, he had refined his theory down to
a system of four differential equations that related changes in the magnetic field to changes in the electric field. The
equations are elegant, and reveal a curious symmetry between electricity and magnetism, each affecting the other in
a similar manner.
It is here, in the elegant symbolism of Maxwell's equations, that humanity made the giant leap from violins
to videos: a series of simple algebraic manipulations extracted the wave equation from Maxwell's equations-which
implied the existence of electromagnetic waves. Moreover, the wave equation implied that these electromagnetic
waves traveled with the speed of light. One immediate deduction was that light itself is an electromagnetic wave-
after all, the most obvious thing that travels at the speed of light is light. But just as the violin string can vibrate at
many frequencies, so-according to the wave equation-can the electromagnetic field. For waves that are visible to the
human eye, it turns out that frequency corresponds to color. Strings with different frequencies produce different
sounds; visible electromagnetic waves with different frequencies produce different colors. When the frequency is
outside the visible range, the waves are not light waves but something else.
What? When Maxwell proposed his equations, nobody knew. In any case, all this was pure surmise, based
on the assumption that Maxwell's equations really do apply to the physical world. His equations needed to be tested
before these waves could be accepted as real. Maxwell's ideas found some favor in Britain, but they were almost
totally ignored abroad until 1886, when the German physicist Heinrich Hertz generated electromagnetic waves-at
the frequency that we now call radio-and detected them experimentally. The final episode of the saga was supplied
by Guglielmo Marconi, who successfully carried out the first wireless telegraphy in 1895 and transmitted and
received the first transatlantic radio signals in 1901.

The rest, as they say, is history. With it came radar, television, and videotape.

Of course, this is just a sketch of a lengthy and intricate interaction between mathematics, physics,
engineering, and finance. No single person can claim credit for the invention of radio, neither can any single subject.
It is conceivable that, had the mathematicians not already known a lot about the wave equation, Maxwell or his
successors would have worked out what it implied anyway. But ideas have to attain a critical mass before they
explode, and no innovator has the time or the imagination to create the tools to make the tools to make the tools
that ... even if they are intellectual tools. The plain fact is that there is a clear historical thread beginning with violins
and ending with videos. Maybe on another planet things would have happened differently; but that's how they
happened on ours.

And maybe on another planet things would not have happened differently-well, not very differently.
Maxwell's wave equation is extremely complicated: it describes variations in both the electrical and magnetic fields
simultaneously, in three-dimensional space. The violin-string equation is far simpler, with variation in just one
quantity-position-along a one-dimensional line. Now, mathematical discovery generally proceeds from the simple to
the complex. In the absence of experience with simple systems such as vibrating strings, a "goal-oriented" attack on
the problem of wireless telegraphy (sending messages without wires, which is where that slightly old-fashioned
name comes from) would have stood no more chance of success than an attack on antigravity or faster-than-light
drives would do today. Nobody would know where to start.

Of course, violins are accidents of human culture-indeed, of European culture. But vibrations of a linear
object are universal-they arise all over the place in one guise or another. Among the arachnid aliens of Betelgeuse II,
it might perhaps have been the vibrations of a thread in a spiderweb, created by a struggling insect, that led to the
discovery of electromagnetic waves. But it takes some clear train of thought to devise the particular sequence of
experiments that led Heinrich Hertz to his epic discovery, and that train of thought has to start with something
simple. And it is mathematics that reveals the simplicities of nature, and permits us to generalize from simple
examples to the complexities of the real world. It took many people from many different areas of human activity to
turn a mathematical insight into a useful product. But the next time you go jogging wearing a Walkman, or switch
on your TV, or watch a videotape, pause for a few seconds to remember that without mathematicians none of these
marvels would ever have been invented.

CHAPTER 6
BROKEN SYMMETRY

Something in the human mind is attracted to symmetry. Symmetry appeals to our visual sense, and thereby plays a
role in our sense of beauty. However, perfect symmetry is repetitive and predictable, and our minds also like
surprises, so we often consider imperfect symmetry to be more beautiful than exact mathematical symmetry. Nature,
too, seems to be attracted to symmetry, for many of the most striking patterns in the natural world are symmetric.

22
And nature also seems to be dissatisfied with too much symmetry, for nearly all the symmetric patterns in nature are
less symmetric than the causes that give rise to them.

This may seem a strange thing to say; you may recall that the great physicist Pierre Curie, who with his
wife, Marie, discovered radioactivity, stated the general principle that "effects are as symmetric as their causes."
However, the world is full of effects that are not as symmetric as their causes, and the reason for this is a
phenomenon known as "spontaneous symmetry breaking."
Symmetry is a mathematical concept as well as an aesthetic one, and it allows us to classify different types
of regular pattern and distinguish between them. Symmetry breaking is a more dynamic idea, describing changes in
pattern. Before we can understand where nature's patterns come from and how they can change, we must find a
language in which to describe what they are.

What is symmetry?

Let's work our way to the general from the particular. One of the most familiar symmetric forms is the one
inside which you spend your life. The human body is "bilaterally symmetric," meaning that its left half is (near
enough) the same as its right half. As noted, the bilateral symmetry of the human form is only approximate: the heart
is not central, nor are the two sides of the face identical. But the overall form is very close to one that has perfect
symmetry, and in order to describe the mathematics of symmetry we can imagine an idealized human figure whose
left side is exactly the same as its right side. But exactly the same? Not entirely. The two sides of the figure occupy
different regions of space; moreover, the left side is a reversal of the right-its mirror image.
As soon as we use words like "image," we are already thinking of how one shape corresponds to the other-
of how you might move one shape to bring it into coincidence with the other. Bilateral symmetry means that if you
reflect the left half in a mirror, then you obtain the right half. Reflection is a mathematical concept, but it is not a
shape, a number, or a formula. It is a transformation-that is, a rule for moving things around.
There are many possible transformations, but most are not symmetries. To relate the halves correctly, the
mirror must be placed on the symmetry axis, which divides the figure into its two related halves. Reflection then
leaves the human form invariant-that is, unchanged in appearance. So we have found a precise mathematical
characterization of bilateral symmetry-a shape is bilaterally symmetric if it is invariant by reflection. More
generally, a symmetry of an object or system is any transformation that leaves it invariant. This description is a
wonderful example of what I earlier called the "thingification of processes": the process "move like this" becomes a
thing-a symmetry. This simple but elegant characterization opens the door to an immense area of mathematics.

There are many different kinds of symmetry. The most important ones are reflections, rotations, and
translations-or, less formally, flips, turns, and slides. If you take an object in the plane, pick it up, and flip it over
onto its back, you get the same effect as if you had reflected it in a suitable mirror. To find where the mirror should
go, choose some point on the original object and look at where that point ends up when the object is flipped. The
mirror must go halfway between the point and its image, at right angles to the line that joins them (see figure 3).
Reflections can also be carried out in three-dimensional space, but now the mirror is of a more familiar kind-namely,
a flat surface.
To rotate an object in the plane, you choose a point, called the center, and turn the object about that center,
as a wheel turns about its hub. The number of degrees through which you turn the object determines the "size" of the
rotation. For example, imagine a flower with four identical equally spaced petals. If you rotate the flower 90°, it
looks exactly the same, so the transformation "rotate through a right angle" is a symmetry of the flower. Rotations
can occur in three-dimensional space too, but now you have to choose a line, the axis, and spin objects on that axis
as the Earth spins on its axis. Again, you can rotate objects through different angles about the same axis.

Figure 3: Where is the mirror? Given an object and a mirror image of that object, choose any point of the object and
the corresponding point of the image, join them by a line, The mirror must be at right angles to the midpoint of that
line.

23
 Translations are transformations that slide objects along without rotating them. Think of a tiled bathroom
wall. If you take a tile and slide it horizontally just the right distance, it will fit on top of a neighboring tile. That
distance is the width of a tile. If you slide it two widths of a tile, or three, or any whole number, it also fits the
pattern. The same is true if you slide it in a vertical direction, or even if you use a combination of horizontal and
vertical slides. In fact, you can do more than just sliding one tile-you can slide the entire pattern of tiles. Again, the
pattern fits neatly on top of its original position only when you use a combination of horizontal and vertical slides
through distances that are whole number multiples of the width of a tile.
Reflections capture symmetries in which the left half of a pattern is the same as the right half, like the
human body. Rotations capture symmetries in which the same units repeat around circles, like the petals of a flower.
Translations capture symmetries in which units are repeated, like a regular array of tiles; the bees' honeycomb, with
its hexagonal "tiles,” is an excellent naturally occurring example.
Where do the symmetries of natural patterns come from? Think of a still pond, so flat that it can be thought
of as a mathematical plane, and large enough that it might as well be a plane for all that the edges matter. Toss a
pebble into the pond. You see patterns, ripples, circular waves seemingly moving outward away from the point of
impact of the pebble. We've all seen this, and nobody is greatly surprised. After all, we saw the cause: it was the
pebble. If you don't throw pebbles in, or anything else that might disturb the sur- face, then you won't get waves. All
you'll get is a still, flat, planar pond.
Ripples on a pond are examples of broken symmetry. An ideal mathematical plane has a huge amount of
symmetry: every part of it is identical to every other part. You can trans- late the plane through any distance in any
direction, rotate it through any angle about any center, reflect it in any mirror line, and it still looks exactly the same.
The pattern of circular ripples, in contrast, has less symmetry. It is symmetric only with respect to rotations about
the point of impact of the pebble, and reflections in mirror lines that run through that point. No translations, no other
rotations, no other reflections. The pebble breaks the symmetry of the plane, in the sense that after the pebble has
disturbed the pond, many of its symmetries are lost. But not all, and that's why we see a pattern.

However, none of this is surprising, because of the pebble, In fact, since the impact of the pebble creates a
special point, different from all the others, the symmetries of the ripples are exactly what you would expect. They
are precisely the symmetries that do not move that special point. So the symmetry of the pond is not spontaneously
broken when the ripples appear, because you can detect the stone that causes the translational symmetries to be lost.

You would be more surprised-a lot more surprised-if a perfectly flat pond suddenly developed a series of
concentric circular ripples without there being any obvious cause, You would imagine that perhaps a fish beneath
the surface had disturbed it, or that something had fallen in and you had not seen it because it was moving too fast.
So strong is the ingrained assumption that patterns must have evident causes that when in 1958 the Russian chemist
B. P. Belousov discovered a chemical reaction that spontaneously formed patterns, apparently out of nothing, his
colleagues refused to believe him. They assumed that he had made a mistake. They didn't bother checking his work:
he was so obviously wrong that checking his work would be a waste of time.

Which was a pity, because he was right.

The particular pattern that Belousov discovered existed not in space but in time: his reaction oscillated
through a periodic sequence of chemical changes. By 1963, another Russian chemist, A. M. Zhabotinskii, had
modified Belousov's reaction so that it formed patterns in space as well. In their honor, any similar chemical reaction
is given the generic name "Belousov-Zhabotinskii [or B-ZJ reaction." The chemicals used nowadays are different
and simpler, thanks to some refinements made by the British reproductive biologist Jack Cohen and the American
mathematical biologist Arthur Winfree, and the experiment is now so simple that it can be done by anyone with
access to the necessary chemicals. These are slightly esoteric, but there are only four of them.*

In the absence of the appropriate apparatus, I'll tell you what happens if you do the experiment. The
chemicals are all liquids: you mix them together in the right order and pour them into a flat dish. The mixture turns
blue, then red: let it stand for a while. For ten or sometimes even twenty minutes, nothing happens; it's just like
gazing at a featureless flat pond-except that it is the color of the liquid that is feature- less, a uniform red. This
uniformity is not surprising; after all, you blended the liquids. Then you notice a few tiny blue spots appearing-and
that is a surprise. They spread, forming circular blue disks. Inside each disk, a red spot appears, turning the disk into
a blue ring with a red center. Both the blue ring and the red disk grow, and when the red disk gets big enough, a blue
spot appears inside it. The process continues, forming an ever-growing series of "target patterns"-concentric rings of
red and blue. These target patterns have exactly the same symmetries as the rings of ripples on a pond; but this time
you can't see any pebble. It is a strange and mysterious process in which pattern-order-appears to arise of its
own accord from the disordered, randomly mixed liquid. No wonder the chemists didn't believe Belousov.

24
 But that's not the end of the B-Z reaction's party tricks. If you tilt the dish slightly and then put it back
where it was, or dip a hot wire into it, you can break the rings and turn them into rotating red and blue spirals, If
Belousov had claimed that you would have seen steam coming out of his colleagues' ears,

* The precise recipe is given in the Notes to The Collapse of Chaos, by Jack Cohen and Ian Stewart.

This kind of behavior is not just a chemical conjuring trick. The regular beating of your heart relies on
exactly the same patterns, but in that case they are patterns in waves of electrical activity, Your heart is not just a
lump of undifferentiated muscle tissue, and it doesn't automatically contract all at once, Instead, it is composed of
millions of tiny muscle fibers, each one of them a single cell. The fibers contract in response to electrical and
chemical signals, and they pass those signals on to their neighbors, The problem is to make sure that they all
contract roughly in synchrony, so that the heart beats as a whole, To achieve the necessary degree of
synchronization, your brain sends electrical signals to your heart. These signals trigger electrical changes in some of
the muscle fibers, which then affect the muscle fibers next to them-so that ripples of activity spread, just like the
ripples on a pond or the blue disks in the B-Z reaction. As long as the waves form complete rings, the heart's muscle
fibers contract in synchrony and the heart beats normally. But if the waves become spirals-as they can do in diseased
hearts-the result is an incoherent set of local contractions, and the heart fibrillates. If fibrillation goes unchecked for
more than a few minutes, it results in death. So every single one of us has a vested interest in circular and spiral
wave patterns.
However, in the heart, as in the pond, we can see a specific cause for the wave patterns: the signals from
the brain. In the B-Z reaction, we cannot: the symmetry breaks spontaneously-"of its own accord"-without any
external stimulus. The term "spontaneous" does not imply that there is no cause, however: it indicates that the cause
can be as tiny and as insignificant as you please. Mathematically, the crucial point is that the uniform distribution of
chemicals-the featureless red liquid-is unstable. If the chemicals cease to be equally mixed, then the delicate balance
that keeps the solution red is upset, and the resulting chemical changes trigger the formation of a blue spot. From
that moment on, the whole process becomes much more comprehensible, because now the blue spot acts like a
chemical "pebble," creating sequential ripples of chemical activity. But-at least, as far as the mathematics goes-the
imperfection in the symmetry of the liquid which triggers the blue spot can be vanishingly small, provided it is not
zero. In a real liquid, there are always tiny bits of dust, or bubbles-or even just molecules undergoing the vibrations
we call "heat"-to disturb the perfect symmetry. That's all it takes. An infinitesimal cause produces a large-scale
effect, and that effect is a symmetric pattern.
Nature's symmetries can be found on every scale, from the structure of subatomic particles to that of the
entire universe. Many chemical molecules are symmetric. The methane molecule is a tetrahedron-a triangular-sided
pyramid-with one carbon atom at its center and four hydrogen atoms at its corners. Benzene has the sixfold
symmetry of a regular hexagon. The fashionable molecule buckminsterfullerene is a truncated icosahedral cage of
sixty carbon atoms. (An icosahedron is a regular solid with twenty triangular faces; "truncated" means that the
corners are cut off.) Its symmetry lends it a remarkable stability, which has opened up new possibilities for organic
chemistry.
On a slightly larger scale than molecules, we find symme- tries in cellular structure; at the heart of cellular
replication lies a tiny piece of mechanical engineering. Deep within each living cell, there is a rather shapeless
structure known as the centrosome, which sprouts long thin microtubules, basic components of the cell's internal
"skeleton," like a diminutive sea urchin. Centro somes were first discovered in 1887 and play an important role in
organizing cell division. However, in one respect the structure of the centrosome is astonishingly symmetric. Inside
it are two structures, known as centrioles, positioned at right angles to each other. Each centriole is cylindrical, made
from twenty-seven microtubules fused together along their lengths in threes, and arranged with perfect ninefold
symmetry. The microtubules themselves also have an astonishingly regular symmetric form. They are hol- low
tubes, made from a perfect regular checkerboard pattern of units that contain two distinct proteins, alpha- and beta-
tubulin. One day, perhaps, we will understand why nature chose these symmetric forms. But it is amazing to see
symmetric structures at the core of a living cell.
Viruses are often symmetric, too, the commonest shapes being helices and icosahedrons. The helix is the
form of the influenza virus, for instance. Nature prefers the icosahedron above all other viral forms: examples
include herpes, chicken- pox, human wart, canine infectious hepatitis, turnip yellow mosaic, adenovirus, and many
others. The adenovirus is another striking example of the artistry of molecular engineering. It is made from 252
virtually identical subunits, with 21 of them, fitted together like billiard balls before the break, making up each
triangular face. (Subunits along the edges lie on more than one face and corner units lie on three, which is why 20 x
21 is not equal to 252.)
Nature exhibits symmetries on larger scales, too. A developing frog embryo begins life as a spherical cell,
then loses symmetry step by step as it divides, until it has become a

25
 blastula, thousands of tiny cells whose overall form is again spherical. Then the blastula begins to engulf
part of itself, in the process known as gastrulation. During the early stages of this collapse, the embryo has rotational
symmetry about an axis, whose position is often determined by the initial distribution of yolk in the egg, or
sometimes by the point of sperm entry. Later this symmetry is broken, and only a single mirror symmetry is
retained, leading to the bilateral symmetry of the adult.
Volcanoes are conical, stars are spherical, galaxies are spiral or elliptical. According to some cosmologists,
the universe itself resembles nothing so much as a gigantic expanding ball. Any understanding of nature must
include an understanding of these prevalent patterns. It must explain why they are so common, and why many
different aspects of nature show the same patterns. Raindrops and stars are spheres, whirlpools and galaxies are
spirals, honeycombs and the Devil's Cause-way are arrays of hexagons. There has to be a general principle
underlying such patterns; it is not enough just to study each example in isolation and explain it in terms of its own
internal mechanisms.

26