What is the butterfly effect?

The butterfly effect, formalised by Edward Lorenz in 1963, is sensitive dependence on initial conditions: two trajectories that start infinitesimally close will diverge exponentially, characterised by the largest Lyapunov exponent λ₁ > 0. For the Lorenz system λ₁ ≈ 0.906, meaning errors roughly double every 0.76 time units. Despite long-term unpredictability, the attractor's shape is deterministic — its structure is identical for all initial conditions that land on it.

How does drag-to-rotate work?

Click and drag the canvas horizontally to rotate around the vertical (Y) axis, and vertically to rotate around the horizontal (X) axis. This applies Euler angle rotations to the 3D coordinates before projecting to 2D using a simple perspective transform. Releasing the drag restores auto-rotation. The projection uses a depth field of view of 4 to give a slight perspective effect.

Why are the attractors coloured differently?

Each preset uses a different colour mapping strategy. Lorenz, Halvorsen, and Chen colour by z-coordinate: low z gives blue-green, high z gives red-orange. Aizawa and Dadras colour by angular position arctan(y/x) — the hue rotates as the trajectory orbits, revealing the toroidal / gyrating structure. Thomas colours by instantaneous speed |ẋ|, so fast segments appear warm and slow segments appear cool.

Strange Attractors Gallery – Lorenz, Thomas, Halvorsen, Chen, Dadras

What Is a Strange Attractor?

A strange attractor is the long-term geometric shape that a chaotic trajectory inhabits in phase space. Unlike a fixed point (0-dimensional) or a limit cycle (1-dimensional), strange attractors have non-integer (fractal) dimension — the Lorenz attractor has Hausdorff dimension ≈ 2.06, meaning it is more than a surface but less than a volume.

Three defining properties of a strange attractor:
• Boundedness — trajectories remain in a finite region of phase space.
• Chaos — nearby trajectories diverge exponentially (positive Lyapunov exponent λ₁ > 0).
• Fractal structure — infinite self-similar fine structure; the attractor has fractional Hausdorff dimension.

Lorenz (1963)

σ=10, ρ=28, β=8/3. The original strange attractor, derived from a 3-mode truncation of Rayleigh-Bénard convection. λ₁ ≈ 0.906. Dimension ≈ 2.06.

Thomas (1999)

Cyclically symmetric in x, y, z; dissipation b=0.19 gives chaos. Each coordinate acts as a driving force for the next. Three butterfly wings interlocked.

Halvorsen (2013)

Cyclic structure: f(x,y,z)→f(y,z,x). Parameter a=1.4. Cross-terms −4y and −y² create complex spiralling orbits.

Aizawa (1994)

Toroidal / shrimp-shaped strange attractor. Six parameters; produces wrapped spiral structure around z-axis. Dimension ≈ 2.38.

Chen (1999)

Dual-wing butterfly similar to Lorenz but with different topology. Parameters a=35, b=3, c=28. Lorenz and Chen are topologically non-equivalent.

Dadras (2009)

Pan-shape morphology with five-parameter system. Exhibits rich period-doubling route to chaos as parameter e varies. Dimension ≈ 2.31.

The Butterfly Effect and Lyapunov Exponents

Two trajectories starting at (x₀, y₀, z₀) and (x₀+ε, y₀, z₀) diverge as |δ(t)| ≈ |δ(0)| · e^{λ₁ t} where λ₁ is the largest Lyapunov exponent. For the Lorenz system λ₁ ≈ 0.906 nats/time unit — errors double roughly every ln(2)/0.906 ≈ 0.76 time units. This is not numerical error — it is genuine dynamical instability intrinsic to the equations.

A system has a strange attractor iff: (1) the sum of Lyapunov exponents is negative (dissipation ensures orbits contract in volume) and (2) at least one Lyapunov exponent is positive (stretching in at least one direction). For Lorenz: λ₁≈+0.906, λ₂≈0, λ₃≈−14.57, sum ≈ −13.67 = −(σ+1+β). The Kaplan–Yorke conjecture relates the Lyapunov spectrum to attractor dimension: D_KY = j + (λ₁+…+λⱼ)/|λⱼ₊₁|.

The Six Attractors at a Glance

Attractor	Equations (schematic)	Key parameters	Dimension
Lorenz	σ(y−x) / ρx−y−xz / xy−βz	σ=10, ρ=28, β=8/3	≈ 2.06
Thomas	sin(y)−bx / sin(z)−by / sin(x)−bz	b=0.19	≈ 2.08
Halvorsen	−ax−4y−4z−y² (cyclic)	a=1.4	≈ 2.14
Aizawa	(z−b)x−dy / dx+(z−b)y / c+az−z³/3−…	6 params	≈ 2.38
Chen	a(y−x) / (c−a)x−xz+cy / xy−bz	a=35, b=3, c=28	≈ 2.09
Dadras	y−px+qyz / ry−xz+z / sxy−ez	p=3, q=2.7, r=1.7, s=2, e=9	≈ 2.31

Route to Chaos: Period Doubling

Many of these attractors emerge from a period-doubling cascade as a parameter is varied: a stable fixed point → period-2 cycle → period-4 → period-8 → …→ chaos. The Feigenbaum constant δ ≈ 4.669 governs the rate at which successive bifurcations accumulate. For the Lorenz system, positive ρ (Rayleigh number) first gives stable convection rolls, then periodic solutions, then a strange attractor beyond ρ ≈ 24.74. The Lorenz attractor as classically visualised (ρ=28) is firmly in the chaotic regime.

Dissipation: Strange attractors live in dissipative systems where phase-space volume contracts: div(f) = ∂ẋ/∂x + ∂ẏ/∂y + ∂ż/∂z < 0. For Lorenz: −(σ+1+β) = −13.67. For Thomas: −3b = −0.57. Despite this overall contraction, chaos occurs because the attractor folds and stretches, creating the characteristic fractal layering visible as colour bands in the simulation.

Frequently Asked Questions

Why does the same attractor look slightly different in every run?

Every run starts from the same initial condition, but floating-point rounding introduces ~10⁻¹⁵ differences that grow exponentially at rate λ₁. After a few hundred time units the trajectory has completely decorrelated from any previous run. However, the shape of the attractor — its fractal geometry — remains identical. You are always seeing the same infinite-time geometric object, just traversed in a different order. The aesthetic result varies, but the underlying mathematics is fixed.

Are strange attractors found in real physical systems?

Yes, extensively. The Lorenz equations were originally derived from atmospheric convection — they motivated meteorological chaos forecasting limits. Strange attractors appear in: laser intensity fluctuations (Lorenz-type), chemical oscillations (Belousov-Zhabotinsky reaction, Rössler equations), cardiac arrhythmias (abnormal attractors in the heart's rhythm), dripping faucets (period-doubling to chaotic drip timing), electrical circuits (Chua's circuit), and fluid turbulence (Ruelle-Takens-Newhouse theorem predicts strange attractors after a few Hopf bifurcations). The Thomas attractor was proposed as a model for three cyclically coupled systems in physics.

What is the difference between chaos and randomness?

Chaotic systems are entirely deterministic — given the exact initial condition and exact equations, the future is uniquely determined. The apparent randomness arises from sensitive dependence: since we can only specify initial conditions to finite precision, our predictions degrade exponentially. True randomness (quantum measurement outcomes, thermal noise) is non-deterministic even in principle. You can tell them apart: chaotic systems produce low-dimensional attractors with definite fractal structure; true random processes produce uniform measure in phase space. Chaos also has a positive but finite information production rate (Kolmogorov-Sinai entropy ≈ sum of positive Lyapunov exponents), while truly random processes have infinite entropy per unit time.

Part of MySimulator.uk — 53+ interactive simulations · Browse all · Blog