Presets
Playback
Pendulum
Run
Camera
Layers
Quality
Model check
The science
What is modelled
Two rigid, massless arms of lengths L1, L2 carry point masses m1 (at the elbow) and m2 (at the free end) and swing in a vertical plane under gravity g = 9.80665 m/s². The orange trail is the original lower bob; the translucent blue copy is a twin started a tiny angle away in θ1.
Equations
- Kinetic energy
T = ½(m1+m2)L1²ω1² + ½m2L2²ω2² + m2L1L2ω1ω2cos(θ1-θ2), potentialU = -(m1+m2)gL1cosθ1 - m2gL2cosθ2. - The Euler-Lagrange equations give a 2x2 linear system for the angular accelerations,
M(θ) α = F(θ, ω), solved exactly at each step, withM = [[(m1+m2)L1², m2L1L2cosδ],[m2L1L2cosδ, m2L2²]], δ = θ1 - θ2. - Friction is a viscous torque
-cωat each joint (Rayleigh dissipation). With c = 0 the energy T + U is conserved. - Integration is classical fourth-order Runge-Kutta with a fixed step of 1 ms, so the error per unit time scales as dt4.
- Small oscillations:
(a d - b²)ω4 - (a k2 + d k1)ω² + k1k2 = 0with a, b, d the entries of M at rest, k1 = (m1+m2)gL1, k2 = m2gL2. This gives two normal-mode periods. - Sensitive dependence: the gap between twins grows roughly as
δ(t) ≈ δ0 eλtuntil it saturates at the size of the accessible phase space. The page prints the mean rateln(δ/δ0)/t; it is a finite-time estimate, not a rigorous Lyapunov exponent, and depends on the energy. - Poincare section: every time θ1 passes through 0 with ω1 > 0 the point (θ2, ω2) is plotted. Regular motion gives points on closed curves, chaotic motion a scattered cloud.
Checks run in the page
- Energy drift of a 20 s chaotic run, at the current lengths and masses, with no friction.
- Slow-mode period from the nonlinear integration (tiny amplitude) against the closed-form eigenvalue.
- Time reversal: forward 6 s, flip velocities, back 6 s returns to the start.
Honest simplifications
- Arms are massless and rigid and the bobs are points; the drawn rods and spheres are only for looks. A real pendulum with heavy arms has different moments of inertia.
- No air drag and no bearing stiction; friction is a linear viscous torque, real bearings add Coulomb friction.
- The twin uses the same equations, so the divergence you see is a property of the equations, amplified from the chosen offset. In a real experiment the offset can never be smaller than the noise of the apparatus.
- The twin is drawn 4.5 cm behind the original only so both can be seen.
What to try
- "Small swings" gives two clean modes and a Poincare plot on closed curves; "Chaotic release" scatters.
- Change the twin offset between 10-9 and 10-3 and watch how long the gap takes to reach order one: roughly ln(offset ratio)/λ longer for the smaller offset.
- Turn friction on: the energy line falls and the chaos dies away.
Loading model...
Loading Blender model...