Questions: Strange Attractors

Question 1

The Lorenz attractor has a fractal dimension of approximately 2.06. What does it mean for an attractor to have a non-integer dimension?

• A. It means the attractor exists in a non-integer number of spatial dimensions
• B. It means the attractor is slightly thicker than a surface (dimension 2) but much thinner than a volume (dimension 3) — it has a locally layered, Cantor-set-like structure that fills more than a sheet but less than a solid
• C. It means the measurement is imprecise and the true dimension is either 2 or 3
• D. It means the attractor has 2.06 unstable directions

Answer: (B)

Fractal dimension measures how a set scales — how its content grows as you examine it at finer resolution. A smooth surface in 3D has dimension exactly 2 (halving the ruler length quadruples the patches needed). The Lorenz attractor needs slightly more patches than a surface because of its layered structure: if you zoom into any part, you find infinitely many nearly-parallel sheets, separated by gaps at all scales. It's not a surface (too thick) and not a volume (too thin) — it's something in between, and the fractal dimension 2.06 captures this precisely.

Question 2

An attractor is strange but not chaotic if it has fractal geometry but λ₁ ≤ 0. Can this actually occur?

• T. True
• F. False

Answer: True

Yes — strangeness (fractal geometry) and chaos (positive Lyapunov exponent) are logically independent, though they usually occur together. The Feigenbaum attractor at the accumulation point of period-doubling has fractal dimension but λ₁ = 0 — it's strange but not chaotic. Conversely, certain systems (like the Lozi map in special parameter regimes) can be chaotic on a non-fractal set. In practice, dissipative chaos almost always produces strange attractors, but the concepts are formally distinct.

Question 3

Why must a strange attractor have zero volume in phase space (for dissipative systems), yet attract a positive-volume set of initial conditions?

Model answer: The attractor has zero volume because the system is dissipative: the divergence of the flow is negative, meaning phase space volumes contract exponentially over time. Any initial volume shrinks to zero in the limit. But the attractor attracts initial conditions from its entire basin of attraction, which typically has positive volume (often the entire phase space). So a positive-volume set of starting points all converge to a zero-volume set — all the dynamics gets compressed onto this thin fractal structure. The attractor has measure zero but is dynamically dominant.

For the Lorenz system, the volume contraction rate is -(σ + 1 + b) ≈ -13.7. A unit cube of initial conditions shrinks by a factor of e^{-13.7} ≈ 10⁻⁶ per unit time. After a few time units, the initial conditions have been compressed onto a set so thin it has zero volume — yet this set has fractal dimension 2.06, meaning it has infinitely complex internal structure. The fractal geometry arises precisely because the contraction is anisotropic: strong compression in one direction (the strongly negative Lyapunov exponent) combined with expansion in another (the positive exponent) creates the layered, foliated structure.

Question 4

A student asks: 'If the strange attractor has zero volume, how can a computer simulation ever find it?' Explain.

Model answer: Computer simulations find the attractor automatically because it is an attractor — trajectories from almost any initial condition converge to it. The dissipative nature of the system means that after a transient period, the numerical trajectory is effectively on the attractor regardless of where it started. You don't need to find the attractor's exact location; the dynamics take you there. The transient trajectory approaches the attractor exponentially fast (at the rate of the most negative Lyapunov exponent), so after discarding the initial transient, the computed trajectory traces out the attractor's structure.

This is exactly what Lorenz did in 1963. He started his simulation at an arbitrary point, waited for transients to die out, and plotted the trajectory — it traced the butterfly-shaped attractor. The attractor's zero volume is not an obstacle because the simulation produces a one-dimensional curve (the trajectory) on the attractor, not a volume-filling set. The fractal structure becomes visible because the trajectory, evolving for long times, densely covers the attractor.