How to run a case 01 Pick a preset. Chaotic 120°, strong chaos 180°/90°, quasi-linear 3° and a light lower bob are chips; the sliders cover angles, masses, lengths and g. 02 Set the initial state. θ₁ and θ₂ from −180° to 180°, initial rates fixed at zero in the panel presets. 03 Choose duration and method. The run lasts four characteristic periods by default; RK4 at 400 steps per period is the default method, with adaptive RK45 available. 04 Read the mode card. Two periods and a mode ratio come from the linearised system — they describe the small-swing limit, not the big swing on screen. 05 Read the chaos card. The largest Lyapunov exponent λ, its R², the initial offset, the time the twin trajectory needs to separate past 0.1 rad, and the Lyapunov time 1/λ when λ is finite and positive.
Worked example Rod 1 = rod 2 = 1 m, g = 9.80665 m/s², both bobs equal (1 kg) unless noted; the two normal-mode periods from the engine:
System
Slow mode period
Fast mode period
Mode ratio θ₁/θ₂
m₁ = m₂ = 1 kg
2.6215 s
1.085861 s
±0.707107
m₁ = 2 kg, m₂ = 1 kg
2.519902 s
1.304397 s
±0.57735
m₁ = 1 kg, m₂ = 3 kg
2.740806 s
0.734397 s
±0.866025
For the 120°/0° preset at the panel's four-period window the chaos card reads λ = 0.60934 /s (R² = 0.941 ), 1/λ = 1.6411 s , and the twin separation has not yet crossed 0.1 rad, so the divergence time is not reached. The stronger 180°/90° preset separates sooner: λ = 1.255345 /s (R² = 0.9446 ), 1/λ = 0.7966 s , divergence at 5.9047 s . Quasi-linear 3° returns no exponent at all — the twin trajectories stay together and the card says so.
Two tones, then a single roar Linearise the double pendulum about hanging straight down and the four degrees of freedom reduce to two independent modes: one where the rods swing together (slow) and one where they oppose (fast). Any small motion is a sum of those two tones, and the periods move with the mass distribution — putting more mass on the upper bob slows the fast mode (1.085861 s → 1.304397 s for m₁ = 2 kg) while making the slow mode slightly quicker. Raise the amplitude and the linearisation fails: the potential is no longer quadratic, the two modes exchange energy, and nearby states separate exponentially with a rate λ. The separation grows like e^{λt} once the exponential regime sets in, so the card's phase-space separation — the θ and ω differences combined — reaches its 0.1 threshold, 10⁵ times the initial 10⁻⁶ rad offset, after 5.9047 s at the strong-chaos preset; the naive ln(10⁵)/λ estimate is about 9.2 s, because the ω difference grows linearly before the exponential takes over. Its Lyapunov time 1/λ is 0.7966 s .
Limits and what it does not do
λ is a fit, not a property of the rods. The card names its fit window and R² because the envelope saturates; the same 180°/90° state integrated over a longer 30 s window fits λ = 1.088 /s (R² = 0.950 ), not 1.255. Quote the exponent with the window that produced it.
No dissipation, no drive, no pivot motion. The rods are frictionless and the support is fixed; a driven or damped pendulum is a different model. What it does not do. It does not estimate a fractal dimension, a Poincaré section or a spectrum; for a control-oriented oscillator model use the spring–mass–damper tab , for a two-body orbit the Kepler tab , and for measurement error on your angles Uncertainty .
The trap: 3° is not "small chaos" Small amplitude does not mean "weakly chaotic" — it means the normal-mode description is the right one. At 3° the twin trajectories never diverge within the run: no λ, no divergence time, just the two mode periods. The mode card stays available at every amplitude, which invites reading it at 120°, where it does not apply: the 2.6215 s and 1.085861 s are the linearised periods near the bottom, not the period of the large swing on screen. Use the card as the small-motion limit and the chaos card for what the motion has become.
Where it is used Teaching deterministic chaos The same deterministic equations produce a repeatable 3° oscillation and an unpredictable 120° tumble; run the twin trajectory and watch a 10⁻⁶ rad difference grow to 0.1 rad in 5.9047 s at 180°/90°. Determinism is not predictability.
Damping and structure Real linkage systems add friction and stiffness; the energy-drift panel shows how well an integrator conserves the total energy before any damping is added, and the mode ratios show how mass placement shifts the two resonances a suspension or linkage must avoid.
Privacy The trajectory integration, the mode solver and the chaos comparison run locally in your browser; nothing is uploaded.
References
Wikipedia, Double pendulum — the coupled equations and the onset of chaos. en.wikipedia.org (访问日期:2026-10-07)
Wikipedia, Normal mode — the two eigenfrequencies of the linearised system. en.wikipedia.org (访问日期:2026-10-07)
Wikipedia, Lyapunov exponent — exponential separation and the 1/λ horizon. en.wikipedia.org (访问日期:2026-10-07)
Wikipedia, Chaos theory — sensitive dependence on initial conditions. en.wikipedia.org (访问日期:2026-10-07)
Strogatz, Nonlinear Dynamics and Chaos — the double pendulum as a standard chaotic system. taylorfrancis.com (访问日期:2026-10-07)
Calculators in this hub
Hand-picked tools, one click away. The mini versions compute live and carry your values into the full calculator.
Sources & review
Reviewed by CalcX Editorial Team
Updated 2026-10-07