MathLabs
TheoremProved

The Lyapunov exponent of the fully chaotic logistic map equals ln 2

Statement

At r=4r=4, the logistic map xn+1=r xn(1−xn)x_{n+1} = r\,x_n(1-x_n) has Lyapunov exponent ln⁡2\ln 2 for Lebesgue-almost every initial condition x0∈(0,1)x_0\in(0,1).

Why is it true?

The map at r = 4 looks nothing like a simple doubling map at first glance, but a clever change of variables (a smooth conjugacy) turns it into exactly the doubling map, whose stretching rate is obviously 2 at every point — the extra distortion introduced by the change of variables averages out to nothing over long times.

Proof sketch

Step 1 (the conjugacy). Substitute x=sin⁡2(πy)x=\sin^2(\pi y). Using the double-angle identity, 4x(1−x)=4sin⁡2(πy)cos⁡2(πy)=sin⁡2(2πy)4x(1-x)=4\sin^2(\pi y)\cos^2(\pi y)=\sin^2(2\pi y). If we also set y′=T(y)y' = T(y) for the doubling map T(y)=2y mod 1T(y)=2y \bmod 1, then sin⁡2(πy′)=sin⁡2(2πy)\sin^2(\pi y') = \sin^2(2\pi y) matches exactly the right-hand side above, so f(h(y))=h(T(y))f(h(y)) = h(T(y)) with h(y)=sin⁡2(πy)h(y)=\sin^2(\pi y): the logistic map at r = 4 is smoothly conjugate to the doubling map.

Step 2 (Lyapunov exponent of the doubling map). The doubling map is piecewise linear with T′(y)=2T'(y)=2 everywhere it is differentiable, so along every orbit 1n∑i=0n−1ln⁡∣T′(yi)∣=ln⁡2\tfrac1n\sum_{i=0}^{n-1}\ln|T'(y_i)| = \ln 2 exactly, for every nn — the limit is trivially ln⁡2\ln 2.

Step 3 (transporting the exponent through the conjugacy). Differentiating f(h(y))=h(T(y))f(h(y))=h(T(y)) by the chain rule gives f′(h(y))h′(y)=h′(T(y))T′(y)f'(h(y))h'(y)=h'(T(y))T'(y), i.e. ln⁡∣f′(x)∣=ln⁡∣h′(T(y))∣+ln⁡∣T′(y)∣−ln⁡∣h′(y)∣\ln|f'(x)| = \ln|h'(T(y))| + \ln|T'(y)| - \ln|h'(y)| at x=h(y)x=h(y). Summing this identity along an orbit y0,y1,…,yn−1y_0,y_1,\dots,y_{n-1} and dividing by nn, the middle terms telescope: 1n∑i=0n−1ln⁡∣f′(xi)∣=ln⁡2+1n[ln⁡∣h′(yn)∣−ln⁡∣h′(y0)∣]\tfrac1n\sum_{i=0}^{n-1}\ln|f'(x_i)| = \ln2+\tfrac1n\big[\ln|h'(y_n)|-\ln|h'(y_0)|\big].

Step 4 (the boundary term vanishes). Since h′(y)=πsin⁡(2πy)h'(y)=\pi\sin(2\pi y) is bounded (its magnitude never exceeds π\pi), the bracketed term ln⁡∣h′(yn)∣−ln⁡∣h′(y0)∣\ln|h'(y_n)|-\ln|h'(y_0)| stays bounded as n→∞n\to\infty for any orbit that avoids the countably many zeros of h′h' — a set of Lebesgue measure zero. Dividing a bounded quantity by nn sends it to 00, so the Lyapunov exponent of the logistic map at r = 4 equals ln⁡2\ln 2 for almost every initial condition.

Topics that use this theorem

Step-by-step proofs

No step-by-step proof yet for this theorem.

References

  1. Steven H. Strogatz (2015). Nonlinear Dynamics and Chaos: With Applications to Physics, Biology, Chemistry, and Engineering
  2. Edward N. Lorenz (1963). Deterministic Nonperiodic Flow
  3. Robert M. May (1976). Simple mathematical models with very complicated dynamics
  4. Warwick Tucker (2002). A Rigorous ODE Solver and Smale's 14th Problem