Lorenz Attractor

In 1963, Edward Lorenz was modeling atmospheric convection — the circular motion of air as it's heated from below. He had reduced the full fluid dynamics equations down to just three coupled differential equations. The system was deterministic: given any starting state, the future was completely fixed by the equations.

Lorenz ran his simulation, stopped midway, and restarted it using a printout of the intermediate values. The printout showed numbers rounded to three decimal places; the computer was storing six. He expected the two runs to diverge slightly and then converge. Instead they diverged exponentially. The tiny rounding error — one part in a thousand — grew until the two trajectories bore no resemblance to each other.

This was the discovery of sensitive dependence on initial conditions: in some dynamical systems, arbitrarily small differences in starting state produce arbitrarily large differences in outcome over time. The system is entirely deterministic — no randomness, no hidden variables — but long-term prediction is impossible because you can never measure initial conditions with infinite precision.

The three equations above (draggable parameters σ, ρ, β) define the system Lorenz studied. Each colored thread is a trajectory through the three-dimensional state space (x, y, z), starting from a slightly different point near the origin. Watch what happens over time: the threads diverge. Two trajectories that start arbitrarily close will eventually be found anywhere in the attractor, with no correlation between their positions. The whole attractor is traversed by any typical trajectory, but the long-term path is effectively unpredictable.

The shape they trace is not random. The trajectories never cross (that would require two points with identical futures to have different pasts — impossible in a deterministic system). They are bounded — the attractor is compact. And they have a fractal dimension of approximately 2.06: between a surface and a volume. The Lorenz attractor is a strange attractor, a set with fractal structure that trajectories approach but never settle on.

There is a philosophical mistake that Lorenz's discovery corrects. We often conflate two different claims: (1) the universe is deterministic, and (2) the future is in principle predictable from the present. Classical physics encouraged this conflation — Laplace's demon, knowing all positions and velocities, could compute all futures. But the Lorenz system shows these claims come apart. A system can be perfectly deterministic, with no randomness at any level, and still be practically unpredictable. The unpredictability isn't ignorance of the laws; it's sensitivity to facts we cannot measure with infinite precision.

The threads above start within a sphere of radius 0.001 of each other. After a few hundred steps, they are on opposite sides of the attractor. This is not metaphor. The universe's initial conditions are not known to arbitrary precision, and likely cannot be. The butterfly effect — does the flap of a butterfly's wings in Brazil set off a tornado in Texas? — is a question about whether the atmosphere is in a Lorenz-like regime where such differences matter. The answer appears to be: it is.

WebGL3D rendering of the Lorenz system with real-time 4th-order Runge-Kutta integration. Each trajectory is a separate thread from a slightly different starting point. Rotate by dragging; adjust parameters to find qualitatively different regimes — below ρ≈24.74, all trajectories fall into one of two fixed points. At ρ=28, the chaotic attractor emerges.