Chaos Lab: Tiny Differences, Different Futures

  • Release two nearly identical double pendulums and explore sensitivity to a tiny starting-angle difference.
  • Compare trails, separation thresholds, sampled energy drift and a finer-step numerical calculation.
  • Share the applied starting recipe or save a labelled result card with the numerical context included.

Start two double pendulums in almost the same position. Release them under the same rules. For a while, they look like one machine. Then the small difference starts to matter.

ContentBASE Chaos Lab turns that idea into an experiment you can replay. Change the starting angles, give one pendulum a tiny extra tilt, and watch both the motion and the distance between their lower bobs.

Chaos Lab

Compare two nearly identical double pendulums and inspect both their separation and numerical sensitivity.

Loading Chaos Lab… Enable JavaScript to explore.

1. Almost the Same Is Still Different

The default pair starts with the upper rod at 120° and the lower rod at −10°, measured from downward vertical. Pendulum B adds just 0.001° to the lower-rod angle. Both begin at rest.

Choose Release this pair. Teal circles identify A, while amber diamonds identify B. Their rods use solid and dashed lines, so you can distinguish them without relying on colour alone.

At first, their paths overlap closely. Later, their motion can become visibly different. No random gust, hidden kick or change in the rules is added after release.

This is an illustration of sensitive beginnings in a deterministic physical model. It is not a claim that every double-pendulum starting position behaves chaotically.

Teal and amber double-pendulum paths from ContentBASE Chaos Lab, showing how a tiny initial-angle difference grows into visibly different motion.

2. The Simple Machine Behind the Complicated Motion

A double pendulum is a pendulum attached to another pendulum. Here each ideal point mass is 1 kilogram, each rod is 1 metre, and gravity is a fixed modelling value of 9.81 m/s².

The rods are rigid and massless, the pivot stays fixed, and the motion occurs in a vertical plane. Friction, driving forces and collisions are absent. The two displayed pendulums do not interact with each other, even when their images overlap.

The sizes of the drawn bobs are symbolic. Their positions follow the point-mass model, rather than a simulation of physical metal arms with distributed weight.

3. Try the Gentle and Identical Controls

The starting-point selector includes useful comparisons:

  • Tiny difference, big swings: the default high-angle release.
  • Gentle swings: smaller initial angles, useful for comparing less dramatic motion.
  • Exactly identical starts: no extra angle for B. With the same initial state and calculation, the primary paths stay identical.
  • Room to flip: a different high-angle release.

Changing the preset installs a new paused pair. Press Play or Release this pair to move it. You can also adjust the two angle sliders and the size of B’s initial difference.

The zero-difference case matters. The simulation does not need a stream of random numbers to look unpredictable. The changed beginning is explicit and measurable.

4. Read the Separation Graph Carefully

Lower-bob separation is the straight-line distance between the lower bobs of A and B. With two 1-metre rods per pendulum, that distance cannot exceed 4 metres.

The amber graph follows this distance over simulated time. A logarithmic scale makes extremely small differences visible; the linear option shows distance on a 0-to-4-metre scale.

Exact zero remains zero in the numeric readout. The logarithmic graph uses a tiny plotting floor because a logarithm of zero is not defined. That display convention does not add a physical separation.

The threshold cards show the first integration step at which the lower bobs are at least 1 centimetre, 10 centimetres and 1 metre apart. Separation can shrink again afterwards. These crossing times are not a calculation of a Lyapunov exponent.

5. The Computer Has a Tiny Difference Too

The display also reports A–B separation relative to the measured finer-step disagreement. A ratio near or below one means the change in numerical resolution affects positions by a comparable amount. This ratio is neither a rigorous error bound nor a Lyapunov exponent. Graph ticks show time and displacement directly at readable sizes.

A simulation calculates an approximation to continuous motion. Chaos Lab uses a fixed 1/480-second integration step for A and B, using the fourth-order Runge–Kutta method.

It also calculates A again with two half steps, each 1/960 second. This third calculation starts from exactly the same state as A. The blue-grey dashed graph shows how far their lower-bob positions disagree.

That comparison is deliberately visible. Numerical approximations can also diverge in a sensitive system. A smooth animation is not evidence that every later frame predicts an exact real-world position.

Finer-step disagreement is a comparison of two calculations. It is not a rigorous bound on the error and does not make either path ground truth. At later times, the tool reports when the discrepancy becomes visibly large.

Tiny floating-point or mathematical-library differences between browsers and devices can also change a much later path. A shared recipe preserves the experiment’s setup; it does not promise bit-for-bit identical long-term coordinates on every platform.

6. Energy Is a Check, Not a Certificate

The ideal model has no friction or external drive, so its total mechanical energy should remain constant. The diagnostic reports the largest sampled energy drift for A or B, in joules, from that pendulum’s own starting energy.

A and B can begin with slightly different energies because their starting angles differ. The diagnostic does not confuse that legitimate difference with numerical drift.

Small energy drift is reassuring, but it does not prove an accurate long-term trajectory. Two paths can conserve energy closely while occupying different positions on the same energy surface. Read the energy and finer-step checks together.

7. Slow It Down Without Changing the Physics Step

Playback speeds range from 0.25× to 10×. Speed changes how much simulated time passes per real second; it does not enlarge the integration step.

Each run stops at 60 simulated seconds. Use +1 simulated second to advance a paused experiment, or Reset applied pair to return to its exact starting conditions.

The elapsed-time clock keeps advancing when you switch tabs. If the browser suspends execution, the engine catches up in bounded work chunks. Explicit pause remains paused when you return. The graph and trails retain samples every 1/30 simulated second while the physics takes smaller steps.

For a clearer view, switch between overlay and side-by-side layouts, change the trail length, or display A’s finer-step path. Reduced-motion devices start paused and use less frequent repaints when you choose to play.

8. Share the Beginning, Then Compare the Ending

Copy experiment link stores the applied angles, B’s extra tilt and playback speed. It opens paused at time zero, ready for someone else to run. The link recreates the recipe, rather than resuming the exact current frame.

A result card shows both pendulums, the simulated time, their separation and the finer-step disagreement. It keeps the numerical context alongside the attractive trails.

Calculations and cards run locally in your browser. This tool does not upload the simulation state. Save a card for the visual, and send its experiment link for the part someone else can investigate.

9. Chaos Lab Questions

9.1 Does Chaos Mean Random?

These equations are deterministic: the starting state and rules determine the mathematical evolution. Sensitivity concerns how the result changes when the beginning changes. It is different from adding random forces at every step.

9.2 Does Every Pair Eventually Fly Apart?

No. Behaviour depends on the initial state and the interval you observe. Try the gentle preset and the exactly identical case, and avoid treating one dramatic run as a theorem about every possible start.

9.3 Can I Use This to Predict a Real Pendulum?

This is an idealized educational experiment. A real apparatus has dimensions, friction, release uncertainty and other properties absent here. The numerical cross-check also shows why long-term agreement requires more than a convincing animation.

9.4 Is This an Old Simulator Brought Back?

Chaos Lab is an original ContentBASE implementation of an established physical model. Existing resources such as myPhysicsLab remain valuable places to explore more parameters and read the derivation.

10. Find Another Surprising Rule

Explore a different gap between local rules and collective outcomes in Neighbourhood Lab, brake a single car in Traffic Jam Lab, or build a tiny deterministic world in Conway’s Game of Life.

Browse ContentBASE Tools for more experiments with a clear rule, a useful surprise and something worth sharing.


Citations

  1. Erik Neumann's first-party derivation of the ideal point-mass double pendulum and its differential equations. ContentBASE independently implements the equal-mass case with a fixed-step RK4 solver and half-step comparison; no third-party simulation code is included. (myPhysicsLab: Double Pendulum Equations)
  2. Shinbrot, T., Grebogi, C., Wisdom, J. and Yorke, J. A. (1992), Chaos in a double pendulum, American Journal of Physics 60, 491–499, DOI 10.1119/1.16860. Experimental research context for sensitivity to starting conditions; this lab does not replicate its apparatus or measured exponents. (Shinbrot et al.: Chaos in a Double Pendulum)
  3. MIT OpenCourseWare, Nonlinear Dynamics I: Chaos, Fall 2012. Course context covering pendulums, sensitivity to initial conditions, numerical methods and transitions to chaos. (MIT: Nonlinear Dynamics I, Chaos)
Cindy, ContentBASE creator assistant

MEET CINDY

Your ContentBASE creator assistant

Cindy helps creators find Canva templates, content ideas, and simple ways to make better social media posts faster.

Want ready-to-use templates? Claim the free Canva bundles or browse the full bundle store.