Back

Lorenz attractor

What you are seeing: a simplified model of atmospheric convection, due to Edward Lorenz in 1963. The trajectory in (x,y,z)(x, y, z) never settles down and never repeats, but it stays trapped in a butterfly-shaped region of space. This was the first rigorously studied example of deterministic chaos: noise-free, completely deterministic, yet long-term prediction is impossible.

The equations are x˙=σ(yx)\dot x = \sigma (y - x), y˙=x(ρz)y\dot y = x (\rho - z) - y, z˙=xyβz\dot z = x y - \beta z, here projected onto the (x,z)(x, z) plane. The red dot is the current state. λ1\lambda_1 in the corner is the running estimate of the largest Lyapunov exponent (analytic value about 0.9060.906); it tells you how fast two nearby starting points drift apart. Drop ρ\rho below 24.7424.74 for a stable spiral; push it past 100100 for periodic windows inside chaos.

Figure 1. strange attractor in the (x, z) projection. Method: classical fourth-order Runge-Kutta from shared/js/engine/ode-rk.js, fixed dt = 0.005. Live max-Lyapunov estimator via tangent-vector renormalization .
attractor
sigma 10.0
rho 28.0
beta 2.67
speed 0.2

WHAT TO TRY

  • Vary each control and watch the rail readouts respond.
  • Compare the diagnostic plot against the live scene.