The Rössler Attractor: Chaos from a Single Fold

Three equations, one nonlinear term

The Rössler system is the set of three coupled ordinary differential equations

• dx/dt = −y − z
• dy/dt = x + a·y
• dz/dt = b + z·(x − c)

with the classic constants a = b = 0.2 and c = 5.7. Look closely and the whole system is linear except for one place: the product z·x in the third equation. Everything else is a sum of variables and constants. This is the entire point. Lorenz's celebrated system carries two nonlinear products (xy and xz); Rössler asked how much you could throw away and still get genuine chaos, and the answer turned out to be a single quadratic term.

Each equation plays a distinct role. The first two govern a rotation in the x–y plane — a spiral. The third equation is a switch: the factor (x − c) is negative while x stays small, so z is held down near zero, but the instant x climbs past the threshold c, that factor flips sign and z erupts. The coupling of that eruption back into the first equation (through the −z term) is what bends the flat spiral up and over on itself.

The slow outward spiral on a nearly flat disk

Set the derivatives to zero and the system has exactly two fixed points. They satisfy x² − c·x + a·b = 0, giving x ≈ 0.0070 and x ≈ 5.693 (with y = −x/a and z = x/a). The inner one sits essentially at the origin; the outer one sits far above the plane at about (5.69, −28.5, 28.5).

Near the inner fixed point, while z is negligible, the first two equations reduce to dx/dt = −y, dy/dt = x + a·y. That two-variable system has eigenvalues 0.1 ± 0.995 i: a complex pair with a positive real part. A positive real part means the spiral winds outward, its radius growing like e^{0.1t}, while the imaginary part sets the turning rate — roughly one full loop every ~6.3 time units. So the inner fixed point is an unstable focus, and the trajectory patiently spirals away from it.

Throughout this phase z barely moves. With x near zero the third equation settles at z ≈ b/(c − x) ≈ 0.035 — a whisker above the plane. That is why the attractor's main body looks like a nearly flat disk: for most of every orbit the motion is a two-dimensional outward spiral pinned just above z = 0. This slow radial growth is the stretch — nearby trajectories, riding an expanding spiral, are pulled apart.

Lifting and folding: the reinjection

The spiral cannot grow forever, and here the single nonlinear term does its work. Each loop pushes x to larger values, and once x exceeds the threshold c = 5.7, the factor (x − c) in dz/dt = b + z(x − c) becomes positive. Now z grows exponentially instead of decaying: the trajectory is lifted up out of the disk, climbing toward that outer fixed point high above the plane.

A large z immediately feeds back into the first equation, dx/dt = −y − z. The big negative −z slams x back down and inward, toward the center of the spiral. As x falls below c again, (x − c) turns negative, z collapses back to its floor, and the orbit is deposited near the middle of the disk to begin spiraling out anew. The whole excursion is a brief pulse — the sheet of trajectories is peeled up, folded over, and reinjected near the center.

This fold is also where the flow's dissipation shows itself. The divergence of the vector field is ∂ẋ/∂x + ∂ẏ/∂y + ∂ż/∂z = a + (x − c), i.e. x − 5.5. Its long-run average along the attractor equals the sum of the Lyapunov exponents, about −5.32 — strongly negative. Phase-space volumes therefore shrink by roughly e^{−5.32} per unit time, crushing a fat blob of initial conditions onto a set of zero volume. The attractor is thin because the flow relentlessly contracts, even while it stretches within the sheet.

Stretch and fold: a Smale horseshoe

Stretch a strip, fold it over, lay it back where it started, and repeat — that operation is the Smale horseshoe, the canonical mechanism by which a smooth deterministic map generates chaos. Rössler's flow is essentially a suspended horseshoe: the spiraling disk supplies the stretch, and the z-pulse supplies the fold and re-lay. Two points that begin a hair apart get pulled to opposite ends of the stretched strip, so the flow shows sensitive dependence on initial conditions.

That sensitivity is measured by the Lyapunov exponents, which for the classic parameters are approximately (+0.071, 0, −5.394). The positive value certifies chaos: infinitesimal errors grow like e^{0.071t}, doubling roughly every ten time units. The near-zero exponent reflects the flow direction along the orbit, and the large negative one is the fierce transverse contraction from the fold.

Folding also builds the attractor's fractal cross-section. Each fold lays a doubled sheet back over the disk; the next fold doubles it again; the limit is an infinite stack of sheets separated on every scale — a Cantor-set structure transverse to the flow. The Kaplan–Yorke formula turns the exponents into a dimension, D = 2 + λ₁/|λ₃| = 2 + 0.071/5.394 ≈ 2.013. The attractor is a hair thicker than a two-dimensional surface: a nearly flat sheet whose thickness hides a Cantor dust of dimension about 0.013. It is bounded (trapped in a finite region), never self-intersecting (solutions of an ODE are unique), and never repeating — the defining trio of a strange attractor.

The return map is (almost) the logistic map

The clearest window into Rössler chaos is a Poincaré section: pick a half-plane cutting through the disk and record only where successive loops pierce it. Because the transverse structure is so thin, plotting each crossing's coordinate against the previous one collapses onto a single humped curve — a one-dimensional, unimodal return map that looks strikingly like the parabola of the logistic map. The stretch-then-fold of the flow is exactly the rise-then-fall of that hump.

This is why Rössler inherits the logistic map's route to chaos. Hold a = b = 0.2 and slowly raise c: the attractor is first a simple closed loop (period 1), which splits into a period-2 loop, then period-4, period-8, and so on in an accelerating period-doubling cascade. The successive parameter windows shrink by the universal Feigenbaum constant δ ≈ 4.669, and beyond the cascade's accumulation point the motion becomes chaotic, interrupted by narrow periodic windows. The classic value c = 5.7 lands squarely in the chaotic regime, producing the familiar single-band funnel.

Why chaos needs three dimensions — and how Lorenz differs

Could a flatter, two-variable system do this? No — and the reason is a theorem. The Poincaré–Bendixson theorem says a bounded trajectory of a smooth autonomous flow in the plane can only settle onto a fixed point or a closed periodic orbit; there is simply no room for a curve to keep wandering forever without crossing itself. Continuous-time chaos therefore requires at least three dimensions, which is exactly what Rössler used, and no fewer. The third variable z lets the trajectory hop over itself during the fold, escaping the plane's tyranny.

Against that minimum, compare the Lorenz attractor (1963), which Rössler was deliberately simplifying. Lorenz's equations carry two nonlinear terms and a reflection symmetry, and their geometry has two lobes: the trajectory circulates on one wing, then jumps to the mirror wing, weaving the famous butterfly. Rössler stripped that to a single fold with no symmetry — one wing, one horseshoe — which is easier to reason about and to draw. Both attractors are fractal sheets barely above dimension two (≈2.06 for Lorenz, ≈2.01 for Rössler), but Rössler makes the stretch-and-fold recipe visible with the fewest moving parts.

Where it came from, and where it turns up

Otto Rössler, a German biochemist, published the system in a two-page 1976 note titled "An Equation for Continuous Chaos". Unlike Lorenz's equations, which fell out of a truncated model of convecting air, Rössler's were designed backward: he wanted the simplest possible flow whose Poincaré map would be a single fold, and he tuned the equations by hand to produce it. He often described the intuition with a taffy-pulling machine or kneaded dough — stretch the sheet, fold it back, and mix forever.

Because it is so cheap to integrate and so transparent, the Rössler system became a standard benchmark and teaching example: for estimating Lyapunov exponents and fractal dimensions, for testing chaos-detection and control algorithms, and for demonstrating the period-doubling road to chaos. Rössler-type dynamics also appear as caricatures of oscillating chemical reactions and are readily built as chaotic electronic circuits, where the abstract fold becomes a voltage you can watch on an oscilloscope. Open questions in the field are less about this specific attractor than about the general program it exemplifies — rigorously proving that observed strange attractors really contain a horseshoe, and classifying the geometry of folding in higher-dimensional flows.

The Rössler attractor was built as a stripped-down relative of the Lorenz attractor — the same recipe for chaos with the least possible machinery.

Feature	Rössler (1976)	Lorenz (1963)
Nonlinear terms	1 (the product zx)	2 (the products xy and xz)
Geometry	single funnel — one fold	two-lobed butterfly — two scrolls
Symmetry	none	invariant under (x,y) → (−x,−y)
Fractal dimension	≈ 2.013	≈ 2.06
Classic parameters	a = b = 0.2, c = 5.7	σ = 10, ρ = 28, β = 8/3
Origin	designed as a minimal abstract prototype	derived from atmospheric convection