Skip to content
Adrien Hubert

Double pendulum, with a perturbed twin.

Two double pendulums are released from the same angle. The amber one starts one-thousandth of a radian away from the ink one — a difference too small to draw at this size. The two systems share their equations of motion and almost all of their initial state. Within a few swings their bobs are nowhere near each other.

t = 0.00s · Δ = 0.000 rad

The system

Each pendulum is two equal rods of length one joined at the pivot, with equal unit masses at each joint. The angles θ₁ and θ₂ are measured from straight down. The Lagrangian L = T − U gives a pair of coupled second-order ODEs; the explicit form is on the Wikipedia article "Double pendulum". A fourth-order Runge-Kutta integrator advances the four-component state (θ₁, ω₁, θ₂, ω₂) by 0.005 seconds per substep, four substeps per drawn frame.

Why it diverges

The double pendulum is one of the cleanest examples of deterministic chaos in undergraduate physics. Above a few degrees of amplitude its largest Lyapunov exponent is positive, so two trajectories that start near each other in phase space move apart roughly exponentially in time. The 10⁻³ radian gap between the two pendulums shown here doubles in something like a second of simulated time, and keeps doubling until it saturates near π. After that the two systems are uncorrelated.

A note on energy

The total energy of an ideal double pendulum is conserved exactly. The RK4 integrator at this step size drifts by a fraction of a percent per minute of simulation, which is invisible at the scale of the picture. It does mean the chaos you see is not the pure mathematical thing: the integrator is itself a small perturbation on top of the deliberate one.