The Lorenz System

Core Idea

The Lorenz system ẋ = σ(y - x), ẏ = rx - y - xz, ż = xy - bz is a three-dimensional ODE derived from a simplified model of atmospheric convection. For certain parameter values (classically σ = 10, b = 8/3, r = 28), it produces chaotic behavior: trajectories loop around two unstable fixed points in a butterfly-shaped pattern, never repeating and sensitive to initial conditions. It was the first widely-studied example of deterministic chaos and the origin of the "butterfly effect" metaphor.

Explainer

The Lorenz system holds a unique place in the history of science. In 1963, meteorologist Edward Lorenz published a set of three ordinary differential equations derived from a heavily truncated model of atmospheric convection — fluid heated from below, like the atmosphere warmed by the Earth's surface. The equations were simple enough to simulate on a 1960s computer, and what Lorenz discovered changed science: these three deterministic equations, with no randomness whatsoever, produced behavior that never repeated and was exquisitely sensitive to initial conditions. Weather prediction had a fundamental limit, and it wasn't about building better instruments.

The equations describe the evolution of three variables: x measures the rate of convective overturning, y measures the horizontal temperature variation, and z measures the vertical temperature stratification. The parameter r is the Rayleigh number (a dimensionless measure of how strongly the fluid is heated), σ is the Prandtl number (ratio of viscous to thermal diffusion), and b relates to the geometry of the convection cell. For r < 1, the only fixed point (the origin, representing no convection) is globally stable — heating is too weak to drive convection. At r = 1, a pitchfork bifurcation creates two new fixed points C+ and C-, representing steady convective rolls turning in opposite directions.

As r increases further, C+ and C- undergo a subcritical Hopf bifurcation and become unstable. Now all three fixed points are unstable, yet the system is dissipative (volumes in phase space contract at rate -(σ + 1 + b)). Where do trajectories go? They settle onto the Lorenz attractor — the famous butterfly-shaped set in three-dimensional space. Trajectories loop around C+, then switch to C-, then back, in a pattern that appears random but is completely determined by the initial conditions. The number of loops around one wing before switching to the other is exquisitely sensitive to the starting point — this is the butterfly effect.

The Lorenz attractor is a strange attractor: it has zero volume (the system is dissipative, so volumes collapse), yet it has a complicated, fractal internal structure. Its fractal dimension is approximately 2.06 — slightly more than a surface but far less than a volume. It attracts all nearby trajectories (it's an attractor) but nearby trajectories on the attractor diverge exponentially (it's strange). The largest Lyapunov exponent is about 0.9, meaning perturbations multiply by a factor of roughly e per unit time. This is the quantitative backbone of the butterfly effect: an initial uncertainty of 10⁻⁶ becomes order 1 in about 14 time units, setting a finite prediction horizon no matter how precisely you measure the initial state.