The end of Laplace's dream
In 1814 Laplace imagined an intellect that, knowing the exact position and velocity of every particle, could compute the entire future and past. Newtonian mechanics is deterministic: the present state fixes all later states uniquely. For two centuries physicists quietly assumed determinism meant predictability. It does not.
Chaos is bounded, aperiodic, long-term behaviour in a deterministic system that shows extreme sensitivity to initial conditions. No dice are rolled; every step follows exactly from the equations. Yet the outcome is, in practice, unforecastable. This track is about how that happens and how we quantify it.
Why nonlinear?
A linear system obeys superposition: double the input, double the response; add two solutions and you get a third. Linear equations can always be solved and never surprise you. The whole first-course toolkit — normal modes, Fourier analysis, RLC circuits — lives in this comfortable world.
A dynamical system: the state x evolves by a rule f. When f is nonlinear (terms like x², sin θ, or products of variables), superposition fails and everything changes.
Real physics is mostly nonlinear: the \sin\theta of a large-amplitude pendulum, the v^2 of air resistance, the self-coupling of fluids, gravity's inverse square. Nonlinearity means feedback — the system acts back on itself — and feedback is what makes rich behaviour, including chaos, possible.
State space: the arena of dynamics
Instead of tracking position versus time, we plot the full instantaneous state as a single point in phase space — for a particle in 1D, the plane of position x and velocity v. As time runs, the point traces a trajectory. Determinism has a beautiful geometric consequence: through each point passes exactly one trajectory, so trajectories can never cross.
A simple harmonic oscillator traces a closed ellipse — it returns to the same state forever. A damped one spirals inward to rest. These orderly pictures are what chaos will break: a chaotic trajectory wanders a bounded region forever without ever closing or repeating.
Feel it: the double pendulum
The cleanest demonstration is the double pendulum: one pendulum hung from the end of another. It is a textbook mechanics problem — two rigid rods, gravity, no approximations needed — yet its motion is chaotic. Watch two copies started from almost the same angle.
For a few swings they move as one. Then the tiny difference between them — invisible at the start — grows until the two pendulums are doing utterly different things. This is sensitive dependence on initial conditions, the defining fingerprint of chaos.
The butterfly and the horizon of prediction
In 1963 meteorologist Edward Lorenz found the same effect in a toy weather model: rounding a number from six digits to three changed the forecast entirely. He coined the butterfly effect — a butterfly's wingbeat could, in principle, alter whether a tornado forms weeks later. The seed of the difference grows exponentially.
A separation δx between two nearby states grows exponentially at rate λ (the Lyapunov exponent, Guide 4). Positive λ is the signature of chaos.
Exponential growth sets a predictability horizon. Because errors grow like e^{\lambda t}, improving your initial data buys only logarithmic extra time: measure a thousand times more precisely and you extend the forecast by a fixed \ln(1000)/\lambda, not a thousandfold. This is why detailed weather forecasts fade after about ten days no matter how good the satellites.
The roadmap
Here is the arc of this track. Guide 2 builds the qualitative toolkit — fixed points, linear stability, phase portraits — that lets us read a flow without solving it. Guide 3 turns a control knob and watches structure change abruptly at bifurcations, following the logistic map's period-doubling road into chaos. Guide 4 puts numbers on chaos with Lyapunov exponents and slices flows into Poincaré sections. Guide 5 reaches the strange, fractal geometry of attractors and where chaos leads across physics.