The Logistic Map and the Feigenbaum Route to Chaos

The precise statement: what the cascade claims

Fix the logistic family f_r(x) = r·x·(1 − x) on the invariant interval [0,1] for parameter r ∈ (0,4]. A period-p point satisfies f_r^p(x) = x, and it is stable (attracting) when the multiplier |(f_r^p)′(x)| < 1. As r increases, the attracting cycle repeatedly loses stability by period doubling: at a bifurcation value rₙ the stable 2ⁿ⁻¹-cycle has multiplier exactly −1 and gives birth to a stable 2ⁿ-cycle.

Feigenbaum's quantitative claim is that the sequence (rₙ) converges to a finite accumulation point r∞ ≈ 3.5699457, and the successive gaps shrink geometrically:

• lim_{n→∞} (rₙ − rₙ₋₁)/(rₙ₊₁ − rₙ) = δ = 4.669201609…
• The spatial scale of the cycle nearest the critical point x = ½ shrinks by α = 2.502907875… per doubling.

Crucially, δ and α are universal: identical for every C³ unimodal map with a nondegenerate (quadratic) maximum and negative Schwarzian derivative.

The picture: a self-similar bifurcation tree

Plot the long-term attractor (the set of x values visited after transients die) vertically against r horizontally. For r < 3 you see a single curve — one stable fixed point x* = 1 − 1/r. At r = 3 the curve forks into two branches: the population now oscillates period-2. Each branch forks again at r₂ = 1+√6 ≈ 3.4495, then again, and again — a cascade of pitchforks.

The visual signature is self-similarity. Zoom into any single fork, rescale the parameter axis by δ and the state axis by α, and you recover a picture statistically identical to the whole tree. This is exactly the geometric content of the two constants: δ measures how fast the forks crowd together horizontally, α how fast the branches pinch together vertically.

• Beyond r∞ ≈ 3.5699 lies a chaotic region shot through with periodic windows — most famously a period-3 window opening at r = 1+√8 ≈ 3.8284, itself born by a tangent (saddle-node) bifurcation.

The key idea: renormalization and a hyperbolic fixed point

Why should one number govern every unimodal map? Feigenbaum's insight was a renormalization argument borrowed from statistical physics. Define the doubling operator T acting on maps: (Tf)(x) = α·f(f(x/α)), i.e. compose f with itself, then rescale space by α to restore the normalization f(0)=1. One period-doubling of f corresponds to one application of T.

The claim is that T has a fixed point g — a universal even function solving the Cvitanović–Feigenbaum equation g(x) = α·g(g(x/α)) with g(0) = 1, α = 1/g(1). All quadratic-maximum maps lie on the stable manifold of g and are attracted to it under repeated doubling; this is why microscopic details wash out and only g's geometry (the α) survives.

• The Feigenbaum δ is precisely the single unstable eigenvalue of the derivative DT at g — the one expanding direction transverse to the stable manifold. Its being > 1 is what makes the bifurcation values converge geometrically.

Oscar Lanford (1982) turned this heuristic into a rigorous computer-assisted proof, verifying existence and hyperbolicity of g with controlled interval arithmetic.

Worked example: the first few bifurcations

Start with the fixed point x* = 1 − 1/r, valid for r > 1. Its multiplier is f_r′(x*) = r(1 − 2x*) = 2 − r. Stability fails when |2 − r| = 1, i.e. at r = 3 (multiplier −1) — the first period-doubling.

For the period-2 cycle, solve f_r(f_r(x)) = x. Dividing out the fixed-point factors leaves the quadratic r²x² − r(r+1)x + (r+1) = 0, whose roots form the 2-cycle. Its multiplier is the product (f_r²)′ = f_r′(p₁)·f_r′(p₂) = 4 + 2r − r². Setting this to −1 gives r² − 2r − 5 = 0, so

• r₂ = 1 + √6 ≈ 3.449490 — the period-4 onset, an exact algebraic number.

From r₃ onward the equations are not solvable in radicals, but numerically r₃ ≈ 3.544090, r₄ ≈ 3.564407, r₅ ≈ 3.568759. The gap ratios 0.4495/0.0946 ≈ 4.751, then 4.656, 4.668 — already converging visibly to δ = 4.6692. This tiny hand computation contains the whole miracle in miniature.

Why the hypotheses matter: what breaks the universality

Universality is not automatic; it hinges on the map's local shape at its peak.

• Quadratic maximum is essential. The values δ = 4.6692 and α = 2.5029 belong to maps with a nondegenerate maximum, f″(c) ≠ 0 (order-2 tangency). Change the tangency order — use a map with a quartic maximum, f(x) ~ 1 − |x|⁴ — and you get a different universal pair δ₄, α₄. So there is one Feigenbaum constant per degree of the critical point, not one universal number for all maps.
• Smoothness / negative Schwarzian. The clean structure (a single attractor, no coexisting stable cycles) uses Singer's theorem: a C³ map with negative Schwarzian derivative Sf = f‴/f′ − (3/2)(f″/f′)² < 0 has at most one attracting cycle, tied to the critical orbit. Drop this and stable cycles can multiply, muddying the cascade.
• The tent map is the boundary case. With a non-smooth corner (piecewise-linear peak), there is no period-doubling cascade at all — it jumps straight to chaos. Smoothness of the hump is what creates the cascade in the first place.

This links the subject to Sharkovskii's theorem (period 3 ⇒ all periods) and to Lyapunov-exponent analysis: at r∞ the largest Lyapunov exponent crosses zero.

Significance: chaos with a measurable, predictable fingerprint

The Feigenbaum route matters because it makes chaos quantitatively predictable in its approach. You cannot forecast individual chaotic trajectories, but you can forecast the metric structure of the transition — a rare foothold in a subject defined by unpredictability.

• Experimental confirmation. Because δ is universal, it is measurable in real systems whose equations we don't know. Libchaber's 1979 convection experiment in liquid helium, plus nonlinear electronic circuits, dripping faucets, and cardiac rhythms, all display period-doubling with δ ≈ 4.67 — a striking cross-validation of the renormalization picture.
• Conceptual reach. It established renormalization-group methods, invented for phase transitions in physics, as a rigorous tool in pure dynamics, and it is a canonical example of emergent universality: microscopic details (the exact nonlinearity) are irrelevant to macroscopic scaling.
• Foundational role. The logistic map is the standard entry point to bifurcation theory, symbolic dynamics, and the Lyapunov-exponent formalism — May's 1976 Nature article popularized it as the cleanest demonstration that simple deterministic models can generate complexity.

The period-doubling cascade of the logistic map: parameter windows shrink by a factor converging to δ = 4.6692…

Bifurcation rₙ	Value	Stable period after	Ratio (rₙ − rₙ₋₁)/(rₙ₊₁ − rₙ)
r₁	3.000000	2	—
r₂ = 1+√6	3.449490	4	4.7514
r₃	3.544090	8	4.6562
r₄	3.564407	16	4.6683
r∞ (limit)	3.569945672…	∞ (chaos onset)	→ δ = 4.66920…