Introduction
You live in a world that often feels unpredictable, random, and unruly. Traditional science promises that knowing the exact laws of nature lets you forecast the future with total certainty. Yet, you find that simple systems can still produce complete disorder and baffling complexity.
Ian Stewart examines how deterministic laws generate random behavior without any external noise. You discover the mathematical shift from clockwork predictability to the geometry of chaos, fractals, and strange attractors. The book maps out how instability shapes everything from weather forecasts and planetary orbits to biological populations and quantum limits.
- How can deterministic equations generate random results?
- What causes the butterfly effect in physical systems?
- How do you measure and control chaos in the real world?
1. Chaos from Order
Deterministic systems are ruled by exact and unbreakable laws, while stochastic behavior is random and irregular. The paradox of chaos is that simple deterministic laws can generate outputs that appear completely random and sensitive to initial conditions.
Deterministic systems can generate stochastic behavior that appears completely random.
Hyperion, an irregularly shaped moon of Saturn, tumbles in a complex and irregular pattern that obeys the laws of gravitation and dynamics.
You can explore this mathematical shift on a pocket calculator or computer. Iterating simple nonlinear functions can produce complex, seemingly random outputs rather than converging to steady states. Starting with 0.54321 and repeatedly iterating the $2x^2 - 1$ formula results in chaos, and starting with 0.54322 yields a completely different sequence after fifty iterations.
- Input a parameter $k$ into a computer program.
- Set an initial value of $x$ equal to 0.54321.
- Run a loop for 50 iterations to allow the sequence to settle down to long-term behavior without printing.
- Run a second loop for 100 iterations, calculate $x = kx^2 - 1$, and print the resulting values.
2. Equations for Everything
Classical mechanics built an era of differential equations and determinism before encountering limitations.
Cosmic Revolution
Ancient Greek technology included sophisticated mechanical computers that used gear systems to model astronomical phenomena. Fishermen discovered the corroded metal lump in a shipwreck off the Greek island of Antikythera, which Derek de Solla Price later analyzed using X-rays.
- Found in the year 1900
- Sunk in a storm in 70 BC
- X-rayed in 1972
- Thirty-two gear wheels reconstructed
Galileo shifted scientific inquiry from asking abstract purposes of natural events to obtaining accurate mathematical descriptions of how they occur. Galileo watched a swinging lamp in the cathedral of Pisa, timing its swings with his pulse, and studied balls rolling on inclined planes. Isaac Newton unified celestial and terrestrial mechanics through his universal law of gravitation and the invention of calculus.
3. The Laws of Error
When systems become too complex to track individually, probability and statistics reveal hidden regularities in coarse average behavior. Pierre-Simon Laplace published his Analytic Theory of Probabilities in 1812. Astronomy error analysis led directly to the normal distribution or error law.
Statistical methods and probability reveal hidden regularities when complex systems become unobservable.
Adolphe Quetelet imported this error law into social sciences to describe populations. He applied error laws to measurements of the human body, crime, marriage, and suicide, inventing the concept of the average man. To demonstrate the normal distribution of height, you can draw a sample of 1,000 men at random from Outer Mongolia.
Francis Galton, living from 1822 to 1911, applied these statistical concepts to heredity and regression analysis. In 1855, Galton plotted the heights of 928 adult children against their parents.
James Clerk Maxwell successfully applied statistical methods to gases, bridging microscopic molecular chaos with macroscopic averages. Because individual molecules cannot be tracked, Maxwell used statistical methods and symmetry assumptions to prove molecular velocities follow a three-dimensional analogue of Quetelet's error law. In 1873, James Clerk Maxwell proposed using statistical methods to the British Association for the Advancement of Science. J. B. Van Helmont invented the word gas in 1632 with deliberate similarity to the Greek word chaos.
4. The Last Universalist
Henri Poincaré was born on 29 April 1854 and founded topology to analyze dynamical systems. Instead of seeking exact formulas, you study qualitative properties that survive continuous deformations.
Poincaré introduced topology and the qualitative theory of differential equations to analyze the stability and global behavior of dynamical systems.
King Oscar II of Sweden offered a prize of 2,500 crowns in 1887 to determine whether the Solar System is stable. Poincaré attempted to solve this using topological methods in a 270-page memoir published in 1890.
A Question for Topology
You can use a Poincaré section to observe where trajectories repeatedly intersect a lower-dimensional cross-section. Poincaré discovered that deterministic systems can exhibit extreme, tangled complexity akin to randomness.
In Hill's reduced model of the three-body problem involving Neptune, Pluto, and a dust particle, Poincaré found homoclinic tangles. He refused to draw these intersecting sheets because their stretching and folding created extreme complexity.
5. One-way Pendulum
To solve equations for a pendulum, classical physics replaces the true force with a proportional one through approximation. The actual force acting on a pendulum is almost, but not quite, proportional to the angle of displacement.
Phase portraits and multidimensional space reveal the geometric structure of nonlinear dynamics.
You can analyze motion without solving differential equations by plotting position horizontally and velocity vertically. Using the Law of Conservation of Energy, kinetic plus potential energy remains constant.
- Pick a value for the energy constant.
- Work out the square root of the constant plus two times the cosine of $x$ for values of $x$ running from minus 180 degrees to plus 180 degrees.
- Ignore negative values under the square root sign.
- Plot two points on the vertical line through $x$ at the positive and negative values of the square root.
Nature is relentlessly nonlinear. A colleague got married in North Wales, and your family visited a lake about 100 metres across.
6. Strange Attractors
Flows in the plane for systems with two degrees of freedom typically possess four fundamental features: sinks, sources, saddles, and limit cycles. Balthasar van der Pol found an extremely important limit cycle in 1927 in a mathematical model of an electronic valve used in radios.
Flows in phase space rely on sinks, sources, saddles, and limit cycles to form strange attractors.
A sink is a point towards which nearby flow-lines spiral, representing a stable steady state. A source is a point from which flow lines spiral away, representing an unstable steady state. A saddle is a steady state stable in some directions and unstable in others. A limit cycle is a closed loop representing periodic motion towards which nearby trajectories converge.
Quasiperiodic motion combines several independent periodic frequencies and takes place topologically on a torus. When multiple periodic motions with independent frequencies are combined, the motion traces a spiral movement on a torus or doughnut. If the ratio of the periods is a rational number, the combination is periodic and repeats, whereas irrational ratios never repeat exactly.
A system is structurally stable if its overall topology does not change when the equations describing it are altered by a small amount. Aleksandr Andronov and Lev Pontryagin introduced coarse systems in the 1930s to capture this qualitative preservation under small perturbations. Stephen Smale generalized this concept to systems with many degrees of freedom.
Poincaré sections reduce continuous flows to mappings on a lower-dimensional cross-section, where topological fixed-point theorems guarantee the existence of periodic trajectories. John Wiley & Sons Ltd. reproduced the Ueda attractor where points swirl like a stirred cup of coffee through a two-dimensional Poincaré section.
Smale constructed an unorthodox mapping inside a solid torus that produces a structurally stable strange attractor when suspended in higher dimensions. Floris Takens and David Ruelle coined the term strange attractor for structurally stable attractors that are neither points nor circles.
The Cantor set and Cantor cheese provide the topological cross-section for solenoids, consisting of an interval repeatedly cleared of middle segments until its total length is zero. Henry Smith discovered the set in 1875, and Georg Cantor used it in set theory in 1883.
- Start with an interval of length 1.
- Remove its middle third while leaving the end points.
- Repeat the middle-third removal indefinitely on the remaining smaller intervals.
7. The Weather Factory
Numerical weather prediction and Lorenz's equations exposed the butterfly effect and sensitive dependence.
Lewis Fry Richardson envisioned a vast Weather Factory operated by 64,000 humans calculating in real time. Supercomputers like ENIAC, MANIAC, and megaflop systems later realized this vision. Sound wave solutions are intentionally filtered out by coarsening equations to prevent numerical instability.
On 15 October 1987, Britain experienced severe storms not predicted by the Meteorological Office's Control Data Cyber 205 computer. The Cray 1 supercomputer predicted ferocious winds instead.
Sensitive dependence on initial conditions limits long-term weather forecasting because tiny observational errors grow exponentially.
Edward Lorenz discovered by accident that rounding a computer input from 0.506127 to 0.506 caused a weather simulation to diverge completely. Tim Palmer distinguished initial value problems from climate predictions.
8. Recipe for Chaos
Stretching and folding combine with period-doubling cascades to produce robust chaotic mixing.
The fundamental geometric mechanism of chaos is the repeated process of stretching and folding phase space, transforming regular movement into random mixing. While traditional dynamics focuses on sitting still or going round and round, chaos relies on stretching an area or volume and folding it back upon itself. This operation mixes components uniformly while expanding internal structures, as seen when pulling seaside rock or operating a kitchen mixer.
The author recalls a seaside sweet shop where a mechanical machine with two rotating steel arms repeatedly stretched and folded a heavy strand of sugary mixture to produce seaside rock.
- Gather the mass and pull it with fingertips allowing a spread of about 18 inches.
- Fold the mass back on itself.
- Repeat the motion rhythmically.
- Start twisting while folding and pulling as the mass changes into a glistening crystal ribbon.
Stephen Smale's horseshoe mapping provides a simplified topological model demonstrating how geometric stretching and folding creates infinite layers and random motion. By taking a square, stretching it into a long thin rectangle, folding it into a horseshoe shape, and placing it back over its original outline, iterations produce an infinitely wiggly curve. Initial points on this curve hop around randomly, mirroring the complicated aperiodic motion found in forced oscillators.
Mary Lucy Cartwright and John Edensor Littlewood studied forced oscillators like the van der Pol equation during World War II, discovering complex aperiodic motion that stemmed from nonlinear electronics and radar research in 1945. In the 1960s, Stephen Smale studied the forced van der Pol oscillator and invented the horseshoe model.
Computer experiments on galactic stellar motion reveal that regular trajectories break down into a stochastic sea of chaos as system energy increases. Astronomers Michel Hénon and Carl Heiles used computers in 1962 to model stellar motion, discovering that regular orbits disintegrate into a mixture of regular islands and random trajectories. While classical perturbation theory predicted only quasiperiodic paths, computer simulations showed that at higher energies, closed curves fell apart into a random smudge of dots with islands of regularity sitting inside chaos.
The breakdown of KAM tori in area-preserving maps explains how chaos arises in plasma magnetic traps and Hamiltonian systems. B. V. Chirikov worked on plasmas and magnetic traps in 1960, creating models for plasma confinement in magnetic traps that showed how chaos lets plasma escape. This area-preserving mapping reflects energy conservation in Hamiltonian systems. As a numerical parameter crosses a critical value, stable islands break up at the chaos border, causing instabilities.
The logistic mapping demonstrates how simple nonlinear equations transition from steady states to chaos through a period-doubling cascade. Using a pocket calculator or home computer to iterate the logistic mapping for varying $k$ values reveals the transition from steady states to periodic loops and chaotic bands. Iterating the discrete logistic mapping produces behaviors that depend on the control parameter $k$.
- Draw a graph of the formula $y = kx(1 - x)$ to obtain an inverted parabola.
- Draw the diagonal line $y = x$ on the same diagram.
- Draw a vertical cobweb from the starting value $x_0$ to hit the parabola.
- Draw a horizontal cobweb to hit the diagonal line to find the next coordinate.
- Repeat to form a staircase between the parabola and the diagonal line.
Between 0 and 3, the system settles into a stable fixed point. As $k$ increases past 3, the system undergoes a period-doubling cascade where periodic cycles double infinitely often before turning into chaos. Around 3.58, the period-doubling cascade ends and the logistic mapping becomes chaotic, while 3.835 is a value of $k$ inside the chaotic regime where a period-3 cycle appears.
Sharkovskii's theorem establishes a universal ordering of periods in one-hump mappings, and bifurcation diagrams exhibit intricate self-similarity. Sharkovskii proved that if a mapping has a period $p$, it must have cycles for all $q$ following a specific integer ordering ending with powers of 2. Furthermore, bifurcation diagrams contain periodic windows that house tiny, perfect copies of the entire global structure, demonstrating self-similarity.
Visitors to Bourton-on-the-Water can see a model village containing a model of itself, mirroring the endless microscopic replication found in the bifurcation diagram of the logistic map. A value of 3.739 in the logistic mapping generates a cycle of period five.
9. Sensitive Chaos
Experiments with lasers, chemical reactions, and dripping taps confirm strange attractors in physical systems.
Lev Landau and Eberhard Hopf proposed that turbulence arises from accumulating multiple independent periodic frequencies through Hopf bifurcations. David Ruelle and Floris Takens demonstrated through topology that multi-frequency quasiperiodic motions are structurally unstable in dissipative systems with friction, breaking down into strange attractors. In 1934, Leray suggested turbulence stems from atomic scale inaccuracies, while Lev Landau proposed in 1944 that turbulence is an accumulation of nested wobbles.
Illumination from the Laser
Harry Swinney and Jerry Gollub experimented on the classic Taylor-Couette apparatus consisting of two rotating cylinders, using lasers and aluminum powder to record velocity power spectra.
- Suspend tiny flakes of aluminium powder in a fluid.
- Shine a laser beam at the flowing fluid.
- Compare the frequency of the returning scattered light with the original light to measure speed.
- Apply Fourier analysis to extract individual component frequencies and build a power spectrum.
10. Fig-trees and Feigenvalues
Renormalization and universal scaling constants explain the transition to chaos across disparate systems.
Mitchell Feigenbaum discovered that period-doubling cascades to chaos exhibit universal geometric scaling ratios. As parameters change and period-doublings accumulate, the ratio of differences between successive bifurcation values converges to a constant number. Using both logistic and trigonometric mappings, he found the exact same scaling ratio. Completely different formulas share identical quantitative properties at the onset of chaos.
Mitchell Feigenbaum used a Hewlett-Packard HP 65 programmable calculator at Los Alamos Laboratory to compute bifurcation ratios.
- 4.6692016090: the universal Feigenvalue scaling ratio for one-humped mappings
- 2.5029078750957: the universal scaling constant for the rate at which branches open
11. The Texture of Reality
Benoît Mandelbrot developed fractal geometry to describe natural phenomena that exhibit detailed structure across multiple scales without characteristic length scales. While traditional smooth shapes represent classical attractors, strange attractors in chaos theory are defined as fractals, and their fractional dimensions quantify their roughness and complexity.
Fractals provide a new geometric language to describe the structured irregularity of nature and serve as the geometric shape of strange attractors.
Benoît Mandelbrot worked as an IBM Research Fellow at Yorktown Heights, developing fractals that successfully forged coastlines, mountains, and clouds.
- 1.15 to 1.25: typical fractal dimension range of a coastline
- 1.2619: Hausdorff-Besicovitch dimension of the Koch snowflake curve
- 0.6309: Hausdorff-Besicovitch dimension of the Cantor set
12. Return to Hyperion
Saturn's irregularly shaped moon Hyperion tumbles unpredictably in attitude while maintaining a regular orbit. Jack Wisdom, Stanton Peale, and Francois Mignard predicted this chaotic tumbling using spin-orbit geometry and Poincaré sections.
Resonances and digital orreries demonstrate chaos in planetary motions, asteroid gaps, and the stability of Earth.
Voyager spacecraft pictures initially seemed periodic, but further images confirmed that Hyperion was tumbling in full-blooded chaos. The principal axes of Hyperion measure 190 km, 145 km, and 114 km, and its orbital eccentricity is 10 per cent.
Daniel Kirkwood noted that asteroids avoid certain orbital radii that resonate with Jupiter's period. The 3:1 resonance causes sudden spikes in eccentricity, making asteroids Mars-crossing and vulnerable to gravitational ejection.
Wisdom, Sussman, and colleagues built the custom-parallel Digital Orrery to simulate Solar System dynamics over millions of years and demonstrate Pluto's chaos.
13. The Imbalance of Nature
Ecological populations often show cyclic fluctuations rather than steady states. Using discrete iterative models with density-dependent growth and time delays, researchers can simulate phenomena like double peaks and regular population crashes.
Simple deterministic nonlinear models of population growth can spontaneously generate complex periodicity and chaos without external noise.
A. J. Nicholson studied laboratory populations of blowflies restricted by a uniform protein diet, finding that their numbers naturally oscillated with double peaks and irregular fluctuations every thirty-eight days or so. His container reached a peak population of 10,000 blowflies. Certain species of cicada nymphs spend 17 years spent underground.
Stephen Budiansky noted that 90 per cent of the Atlantic coastal forest of Brazil was cleared, yet recorded extinctions remained at zero.
14. Beyond the Butterfly
Distinguishing between a random and a deterministic system depends on whether repeating an experiment from the exact same initial state yields identical results on a short timescale. Dropping a cannonball under gravity is deterministic, whereas shuffling a pack of cards and drawing the top card is random.
The Liapunov exponent quantifies the exponential growth rate of tiny initial errors, serving as a diagnostic test for chaos.
If the Liapunov growth rate is greater than 1, initial errors increase exponentially and the system exhibits the butterfly effect. A multiplying-by-ten map produces a growth rate of 10 where errors grow tenfold at each step.
Short-term Prediction
While long-term prediction is impossible due to the butterfly effect, chaotic systems are predictable in the short term using deterministic models or historical analogues. Tessellation methods tile phase space with small regions to interpolate future trajectories based on local linear dynamics.
You can apply these steps to forecast a system:
- Find a previous point in records that is very close to the current initial state point A.
- Observe what happened previously starting from that point to predict the future of point A.
- Find several previous points surrounding point A to form a bounding region.
- Follow the previous futures of those bounding points to determine the range where the prediction must fall.
- Interpolate the trajectory using local linear dynamics to give an optimal guess for the unrecorded point.
Alister Mees applied tessellation forecasting methods to EEG brain wave traces, monthly measles epidemic differences, and Hudson's Bay lynx pelt numbers.
The F RACMAT Project
Phase space reconstruction techniques derived from chaos theory can solve industrial quality control problems by quantifying sequential variability in raw materials. Standard material tests failed to predict whether wire could be coiled successfully into springs because coilability depends on sequential variability rather than statistical averages.
Len Reynolds and a team from SRAMA and Warwick University built a machine called FRACMAT that uses Ruelle-Takens phase space reconstruction on laser micrometer measurements of test springs. FRACMAT successfully passed blind tests by accurately identifying good and bad wire samples sent by spring and wire manufacturers, winning an award and securing commercial orders.
You can operate the testing procedure through these concrete actions:
- Feed a wire sample at speed into the coiling machine to form a test spring with about 500 individual coils wound on a mandrel.
- Transfer the completed spring to a shelf where a laser micrometer tracks rapidly along it to measure all inter-coil spacings.
- Reconstruct the attractor from the measured time-series using chaos-theoretic algorithms.
- Quantify how long and broad the attractor is, and plot the results on a classification diagram to determine wire coilability.
The total duration of the FRACMAT test takes 3 minutes.
15. Von Neumann's Dream
John von Neumann proposed that once computers arrived, predictable stable phenomena would be forecast while unstable ones would be controlled by introducing small disturbances. Because chaotic systems are exceptionally sensitive to small perturbations, they are also sensitive to errors in those changes.
Edward Ott, Celso Grebogi, and James Yorke discovered that chaotic attractors contain vast numbers of unstable periodic points possessing saddle-like structures that can be stabilized.
You can stabilize chaotic dynamics with high efficiency by monitoring the state and applying tiny corrections. You operate the system through these concrete actions:
- Monitor the current state of the dynamical system.
- Identify the stable manifold associated with an unstable periodic point.
- Apply repeated small perturbations or adjust the driving system so the state falls inside the initially stable zone.
- Avoid trying to push the system directly back to the steady state in a single step to prevent completely unstable perturbations.
NASA engineers used lunar fly-bys and gravitational neutral points to redirect the dying ISEE-3 satellite into an encounter with comet Giacobini-Zinner.
16. Chaos and the Quantum
Quantum mechanics and measurement interpretations connect to determinism through hidden variables and riddled basins.
In 1900, Max Planck announced energy quantum theory, introducing Planck's constant in Joule-seconds as $6.626 \times 10^{-34}$. Einstein rejected fundamental quantum probabilities. You encounter linear wave functions, Schrödinger's equation, and superposition when studying quanta.
Wave Functions
Cat in a Box
In 1935, Schrödinger introduced his cat thought experiment. Terry Pratchett's comic fantasy Witches Abroad features a cat named Greebo locked in a box, yielding three states upon opening: alive, dead, and absolutely bloody furious.
Bohm's Interpretation
David Bohm formulated a deterministic alternative in 1952.
Bell's Inequality
Dice and Determinism
Riddled Basins
In 1995, Tim Palmer discovered the loophole in Bell's inequality.
17. Farewell, Deep Thought
Complexity theory and emergence supersede reductionism by showing how simple rules generate rich macroscopic structures.
Reductionism attempts to explain phenomena by looking down mental funnels to fundamental rules. This approach often leads into an endless hierarchy of branching funnels that fail to explain everyday realities.
Explaining a goat eating a rosebush involves tracking everything from photosynthetic biochemistry in leaves to the complex visual cortex algorithms and chewing muscle mechanics of the goat, trapping science in a reductionist nightmare.
Explanations across hierarchical scientific levels operate through cascades of fungibilities. Detailed low-level mechanics can be substituted without altering the higher-level story.
Cans of baked beans with the same manufacturer and nominal weight are fungible in the eyes of the law. One compared sample contained 1,346 beans, while the other held 1,347.
Ant Country represents the conceptual no-man's-land between top-down observation and bottom-up reductionism. In this gap, emergent phenomena are actually created.
Chris Langton's ant follows simple rules on a grid that initially generate chaos. After 10,000 steps, the system spontaneously organizes into a predictable diagonal band called a highway. That repeating sequence requires exactly 104 steps.
10 Key Ideas
- Simple deterministic laws can generate outputs that appear completely random and sensitive to initial conditions.
- Phase portraits and multidimensional space reveal the geometric structure of nonlinear dynamics without needing exact formulas for differential equations.
- Flows in phase space rely on sinks, sources, saddles, and limit cycles to form strange attractors that are neither points nor circles.
- Sensitive dependence on initial conditions limits long-term weather forecasting because tiny observational errors grow exponentially like the butterfly effect.
- The fundamental geometric mechanism of chaos is the repeated process of stretching and folding phase space, transforming regular movement into random mixing.
- Mitchell Feigenbaum discovered that period-doubling cascades to chaos exhibit universal geometric scaling ratios across disparate systems.
- Fractals provide a new geometric language to describe the structured irregularity of nature and serve as the geometric shape of strange attractors.
- Resonances and digital orreries demonstrate chaos in planetary motions, asteroid gaps, and the stability of Earth.
- Edward Ott, Celso Grebogi, and James Yorke discovered that chaotic attractors contain vast numbers of unstable periodic points possessing saddle-like structures that can be stabilized with high efficiency by monitoring the state and applying tiny corrections.
- Complexity theory and emergence supersede reductionism by showing how simple rules generate rich macroscopic structures in the conceptual no-man's-land between top-down observation and bottom-up reductionism.