Push one pendulum. It falls still — completely.

Two identical pendulums, joined by a weak spring. Only the left one is set swinging; the right one starts at rest. Watch, and something odd happens: the left pendulum's swing shrinks to nothing while the right one, never touched, picks up the exact same motion — then the whole exchange reverses, forever.

What you're seeing

Two pendulums, matched exactly, hang from a shared frame joined by a weak spring. Only the left one is pulled back and released; the right one starts perfectly at rest. If the spring weren't there, that would be the whole story: the left one swings forever at its own amplitude, the right one just hangs still forever.

The spring changes everything. As the left pendulum swings, it very slightly stretches and compresses the spring, which nudges the right pendulum — barely, at first. But that nudge compounds, swing after swing, and eventually all of the motion has crossed over: the left pendulum comes to a near-standstill while the right one swings at the exact amplitude the left one started with. Then, having nowhere else to go, it reverses — the motion crosses back, and the whole cycle repeats, forever, as long as nothing is damping it. The coupling dial only changes how fast this happens, not whether it happens.

Now drag the mismatch dial. Detune the two pendulums — make one very slightly different from the other — and the complete handoff breaks. The motion still sloshes back and forth, but it's capped: the right pendulum never quite reaches the left one's original swing, and the more mismatched they are, the smaller that cap gets. Perfect crossover is a razor's edge that only identical systems sit on — the same reason a wine glass shatters only near its own resonant note — and the classical cousin of the resonance math behind why two coupled quantum systems can swap energy almost completely, while detuning shuts the exchange down.

The rule, exactly. Small-angle pendulums, natural frequencies ω₁, ω₂, spring strength k: θ₁″ = −ω₁²θ₁ − k(θ₁−θ₂), θ₂″ = −ω₂²θ₂ − k(θ₂−θ₁) integrated live with 4th-order Runge–Kutta. When ω₁=ω₂, the sum u=θ₁+θ₂ and difference v=θ₁−θ₂ decouple exactly into two independent oscillators — u at the plain frequency ω₀ (moving together, the spring never stretches) and v at the higher √(ω₀²+2k) (moving oppositely, the spring fights back) — and recombining them gives the exact textbook beat solution. Verified in node (improve/verify/91-coupled-pendulums.js): the live RK4 integration matches that exact closed form to within 0.002 across a range of coupling strengths; at zero mismatch the crossover reaches 99.98% of the full starting amplitude while the source pendulum simultaneously drops to essentially zero; and total mechanical energy is conserved to within 0.001% throughout, confirming the integrator itself is sound. Negative controls: introducing a mismatch caps the peak crossover well below complete and that cap tightens every time the mismatch is increased further (100% at zero mismatch, falling to 42%, 15%, then 7% across the slider's range); and with the coupling spring strength itself set to zero, the resting pendulum measurably never moves — not approximately, but to floating-point precision — because with no spring the two equations are simply independent.

← the cabinet · Thought Toys — a cabinet of explorable explanations. Exhibit 91.