The Butterfly Effect: lesson plan

Students release a thousand double pendulums a billionth of a degree apart, watch them split, drop particles onto the Lorenz attractor and try to forecast chaos, then use exponential growth to explain why perfect prediction is impossible even when the laws are exact.

Ages
14–18 (grades 9–12)
Time
50–60 minutes
Subject
Maths and physics: chaos, exponential growth and prediction
You need
Devices with a browser (one per pair), calculators, squared paper or a spreadsheet for a doubling table

Open The Butterfly Effect Jump to the quiz

Learning objectives

By the end of the lesson, students can:

Key vocabulary

Deterministic
Fully decided by its starting state and the rules. Run it again from exactly the same start and you get exactly the same result.
Chaos
Deterministic behaviour that is so sensitive to its starting state that long-term prediction is impossible in practice.
Sensitive dependence on initial conditions
Tiny differences at the start grow, roughly exponentially, until they are as large as the motion itself.
Exponential growth
Growth by the same factor in each equal time step, such as doubling every second.
Lyapunov exponent
The average rate at which nearby paths separate. For the Lorenz system it is about 0.91 per time unit.
Attractor
The set of states a system settles into over time. The Lorenz attractor is the famous butterfly-shaped one.
Ensemble forecast
Many forecasts started from slightly different states, used to show how confident a prediction is.

Background for the teacher

The Butterfly Effect is a guided walk in five chapters (Swing, 1,000, Butterfly, Forecast, Wings). Each pendulum is two equal masses on two equal 1 m rods under 9.81 m/s², integrated with the fourth-order Runge–Kutta method 240 times a second, without friction. In chapter 2, 1,000 pendulums start a billionth of a degree apart. Because the simulation is deterministic, the game can compute in advance exactly when the outermost two will disagree by a tenth of a rod length, and lets students bet on whether it will be under 11 s, 11–14 s, 14–18 s or after 18 s.

The key maths is exponential growth. If a gap doubles every second, after n seconds it has grown by 2n. Since 230 ≈ 1.07 billion, a gap of a billionth of a degree reaches about 1° in only 30 seconds, and making the start a billion times more precise buys only about 30 extra seconds. In 1961 Edward Lorenz noticed this when he re-entered 0.506 instead of 0.506127 into his weather model; his 1963 paper described the Lorenz attractor, where nearby paths separate by a factor of e every 1.1 time units (Lyapunov exponent about 0.91). Chapter 4 runs forecasts of 0.5, 1, 3, 6 and 10 s alongside an ensemble of 80 runs with tiny measurement errors (a thousandth of a degree), which shows why real forecasters such as ECMWF give probabilities, and why detailed weather forecasts lose skill after roughly one to two weeks.

Warm-up (5 minutes)

Activity with the game

  1. Pairs open The Butterfly Effect. In chapter 1 (Swing), drag the red bob and let go several times. Pairs describe the motion and note that starting in nearly the same place does not give nearly the same swing for long.
  2. Chapter 2 (1,000): before release, each pair places a bet on when the pendulums will split. Release them, record the actual time, and try two or three different starting poses by dragging. Do small swings split later than big ones?
  3. Build a doubling table: starting gap 10−9 degree, doubling every second. Find how many seconds until the gap reaches 1° and then 100°. Compare with the split times seen in the game.
  4. Chapter 3 (Butterfly): tap the Lorenz attractor to drop two particles almost on top of each other. Sketch the graph of the gap between them and describe its shape (it grows steadily on a log scale, then levels off once they are on opposite wings). Then press Lorenz’s rounding to replay his 1961 mistake.
  5. Chapter 4 (Forecast): set a starting pose, tap where you think the red bob will be after each forecast time, and run it. Record whether each guess was a hit for 0.5, 1, 3, 6 and 10 s. Look at how the ensemble spreads out.
  6. Regroup: plot the class hit rate against forecast time. Discuss where the “predictability horizon” is and link it to real weather forecasts.

Discussion questions

The game told you when the pendulums would split before they did. How is that possible if they are chaotic?
Show a model answerThe simulation is deterministic: the same numbers in give exactly the same numbers out. The game simply ran the calculation ahead. Chaos does not mean the laws are random; it means any uncertainty in the start, however tiny, grows until prediction fails. Here the computer knows the start exactly, so it can predict.
If we measure the start a million times more precisely, how much longer can we predict, if errors double every second?
Show a model answerA million is about 220, so only about 20 extra seconds. Each doubling of precision buys one extra doubling time, which is why better instruments improve forecasts only slowly.
Why do weather centres run dozens of forecasts instead of one?
Show a model answerBecause the atmosphere is chaotic and the starting measurements are imperfect, they start many runs from slightly different states. If the runs agree, the forecast is confident. If they spread out like the pendulums, the honest answer is a probability, such as a 60% chance of rain.
Is a coin toss chaotic or random?
Show a model answerA coin follows Newton’s laws, so it is deterministic, but its result is so sensitive to the exact launch speed and spin that we treat it as random. Chaos is one way that apparently random behaviour can come from exact laws.
Can a butterfly really cause a tornado?
Show a model answerLorenz’s title was a provocation. The point is that in a chaotic system a tiny disturbance can change which weather happens weeks later, so the details cannot be traced or forecast that far ahead. It does not mean butterflies create the energy of a storm.

Common misconceptions

Chaotic means random.
Chaotic systems follow exact, deterministic rules. They are unpredictable in practice only because we can never know the starting state perfectly.
With a big enough computer, we could forecast the weather months ahead.
The limit comes from how fast measurement errors grow, not from computing power. Detailed forecasts lose skill after about one to two weeks whatever the computer.
Small causes always have small effects.
In linear systems, roughly yes. In chaotic systems, small differences are amplified exponentially, so a tiny change can lead to a completely different outcome.

Quick quiz

  1. A double pendulum is…

    1. random
    2. deterministic but chaotic
    3. neither deterministic nor chaotic
    4. only chaotic with friction
    Answer

    B. Its future is fixed by its start, yet tiny differences grow exponentially.

  2. If a gap doubles every second, roughly how many seconds does it take to grow a billion times?

    1. 9
    2. 30
    3. 100
    4. 1,000,000,000
    Answer

    B. 2³⁰ ≈ 1.07 × 10⁹.

  3. What did Lorenz type in that caused his rerun to diverge in 1961?

    1. A wrong equation
    2. A rounded number, 0.506 instead of 0.506127
    3. A negative sign
    4. A different computer
    Answer

    B. The printout showed three decimal places, while the computer stored six.

  4. Why do forecasters use ensembles?

    1. To save computer time
    2. To show how uncertain the forecast is
    3. To make every run identical
    4. Because the laws of physics are random
    Answer

    B. Runs from slightly different starts show how much the forecast can spread.

  5. Roughly how far ahead do detailed day-to-day weather forecasts keep useful skill?

    1. A few hours
    2. About 1 to 2 weeks
    3. About 6 months
    4. Forever, with enough data
    Answer

    B. Errors grow and spread to large scales, limiting detailed skill to around one to two weeks.

  6. What is the shape of the Lorenz attractor often compared to?

    1. A spiral galaxy
    2. A butterfly
    3. A snowflake
    4. A circle
    Answer

    B. Its two lobes look like a butterfly’s wings, a coincidence with Lorenz’s famous talk title.

Going further

Calculators for this lesson

Curriculum links

The science in this plan matches the notes and sources under The Butterfly Effect. Free to use and adapt for teaching. Ideas or corrections: hello@notebookfun.com.