Skip to content

Nonlinear Dynamics and Chaos

11.1 The Logistic Map

The logistic map is the simplest model of chaotic dynamics:

xn+1=rxn(1xn),0x1x_{n+1} = r\,x_n(1 - x_n), \quad 0 \leq x \leq 1

  • 0r10 \leq r \leq 1: xn0x_n \to 0 (extinction)
  • 1<r<31 < r < 3: Fixed point x=11/rx^* = 1 - 1/r (stable population)
  • 3<r<3.4493 < r < 3.449: Period-2 cycle
  • 3.449<r<3.5443.449 < r < 3.544: Period-4 cycle
  • rc3.56995r_c \approx 3.56995: Onset of chaos (period doubling cascade)
  • r>rcr > r_c: Chaotic regime (with periodic windows)

Feigenbaum constants: The ratio of successive bifurcation intervals converges to δ=4.669\delta = 4.669\ldotsA universal constant for all period-doubling transitions.

11.2 Lyapunov Exponents

The Lyapunov exponent λ\lambda measures the rate of exponential divergence of nearby trajectories:

δx(t)eλtδx(0)|\delta x(t)| \sim e^{\lambda t}|\delta x(0)|

  • λ<0\lambda < 0: Trajectories converge (stable, periodic)
  • λ=0\lambda = 0: Trajectories neither converge nor diverge (marginal)
  • λ>0\lambda > 0: Trajectories diverge exponentially (chaos)

For the logistic map at r=4r = 4: λ=ln20.693\lambda = \ln 2 \approx 0.693.

11.3 The Driven Damped Pendulum

The equation of motion for a driven damped pendulum:

θ¨+γθ˙+gsinθ=Acos(ωdt)\ddot{\theta} + \gamma\dot{\theta} + \frac{g}{\ell}\sin\theta = A\cos(\omega_d t)

For certain parameter ranges, this system exhibits deterministic chaos despite being described by a simple differential equation. The transition to chaos occurs via the same period-doubling cascade as the logistic map.

Poincaré section: By sampling the phase space (θ,θ˙)(\theta, \dot{\theta}) at the driving frequency, the chaotic attractor reveals its fractal structure.

Worked Example 11.1: Period Doubling in the Logistic Map

At r=3.2r = 3.2The logistic map has a stable period-2 cycle.

Starting from x0=0.2x_0 = 0.2:

x1=3.2×0.2×0.8=0.512x_1 = 3.2 \times 0.2 \times 0.8 = 0.512

x2=3.2×0.512×0.488=0.799x_2 = 3.2 \times 0.512 \times 0.488 = 0.799

x3=3.2×0.799×0.201=0.513x_3 = 3.2 \times 0.799 \times 0.201 = 0.513

x4=3.2×0.513×0.487=0.799x_4 = 3.2 \times 0.513 \times 0.487 = 0.799

The system settles into the cycle {0.513,0.799}\{0.513, 0.799\}. The period-2 orbit has f(x)=f(f(x))=xf(x^*) = f(f(x^*)) = x^* with f(x)=rx(1x)f(x) = rx(1-x).

To find the period-2 points analytically: solve f(f(x))=xf(f(x)) = x while excluding the fixed points of ff:

r2x(1x)[1rx(1x)]=xr^2 x(1-x)[1 - rx(1-x)] = x

r[r(1x)(1rx+rx2)]=1r[r(1-x)(1 - rx + rx^2)] = 1

At r=3.2r = 3.2: the solutions are x=0.5130x^* = 0.5130 and x=0.7995x^* = 0.7995Matching our numerical result.

Common Pitfalls (Additional)

  1. Euler angles have gimbal lock: When θ=0\theta = 0 (or π\pi), the angles ϕ\phi and ψ\psi become degenerate and cannot independently specify the orientation. This is the same gimbal lock problem encountered in navigation. For numerical simulations, use quaternions instead.

  2. Canonical transformations are not coordinate transformations: A canonical transformation changes both coordinates and momenta in a way that preserves Hamilton”s equations. Not every coordinate transformation (q,p)(Q,P)(q, p) \to (Q, P) is canonical. The test is whether the Poisson brackets {Qi,Qj}={Pi,Pj}=0\{Q_i, Q_j\} = \{P_i, P_j\} = 0 and {Qi,Pj}=δij\{Q_i, P_j\} = \delta_{ij} are preserved.

  3. The adiabatic theorem requires slowness: The adiabatic invariant E/ωE/\omega is conserved only when ω˙/ω21\dot{\omega}/\omega^2 \ll 1 (the parameter changes slowly compared to the period). For sudden changes, the adiabatic theorem does not apply, and the energy change must be computed from the explicit time-dependent Hamiltonian.

  4. Chaos requires nonlinearity but not complexity: The driven damped pendulum and the logistic map are among the simplest dynamical systems, yet they exhibit chaos. Chaos is a property of the dynamics, not the complexity of the equations. A three-dimensional autonomous system (or two-dimensional non-autonomous system) is the minimum for chaos.

  5. Liouville’s theorem constrains but does not determine dynamics: Liouville’s theorem says phase space volume is conserved, but it does not prevent phase space density from becoming filamented (stretched and folded). In fact, this filamentation is the mechanism behind mixing and ergodicity in Hamiltonian systems.

Problems (Additional)

Problem 19: Euler's Equations for an Asymmetric Body

A rigid body with principal moments I1=1I_1 = 1, I2=2I_2 = 2, I3=3I_3 = 3 (in kg\cdotM2^2) rotates freely with initial angular velocity ω(0)=(0.1,0.5,1.0)\boldsymbol{\omega}(0) = (0.1, 0.5, 1.0) rad/s.

(a) Verify that TT and L2L^2 are conserved by computing them at t=0t = 0.

(b) Use Euler’s equations to find ω˙1,ω˙2,ω˙3\dot{\omega}_1, \dot{\omega}_2, \dot{\omega}_3 at t=0t = 0.

(c) Is the motion about the intermediate axis (I2I_2) stable? Predict the qualitative behaviour.

Solution:

(a) T=12(1×0.01+2×0.25+3×1.00)=12(0.01+0.50+3.00)=1.755T = \frac{1}{2}(1 \times 0.01 + 2 \times 0.25 + 3 \times 1.00) = \frac{1}{2}(0.01 + 0.50 + 3.00) = 1.755 J.

L2=1×0.01+4×0.25+9×1.00=0.01+1.00+9.00=10.01L^2 = 1 \times 0.01 + 4 \times 0.25 + 9 \times 1.00 = 0.01 + 1.00 + 9.00 = 10.01 (kg\cdotM2^2/s)2)^2.

(b) ω˙1=(I2I3)I1ω2ω3=(23)1(0.5)(1.0)=0.5\dot{\omega}_1 = \frac{(I_2 - I_3)}{I_1}\omega_2\omega_3 = \frac{(2-3)}{1}(0.5)(1.0) = -0.5 rad/s2^2.

ω˙2=(I3I1)I2ω3ω1=(31)2(1.0)(0.1)=0.1\dot{\omega}_2 = \frac{(I_3 - I_1)}{I_2}\omega_3\omega_1 = \frac{(3-1)}{2}(1.0)(0.1) = 0.1 rad/s2^2.

ω˙3=(I1I2)I3ω1ω2=(12)3(0.1)(0.5)=0.0167\dot{\omega}_3 = \frac{(I_1 - I_2)}{I_3}\omega_1\omega_2 = \frac{(1-2)}{3}(0.1)(0.5) = -0.0167 rad/s2^2.

(c) The initial ω2=0.5\omega_2 = 0.5 is the largest component, so the rotation is predominantly about the intermediate axis. Since I1<I2<I3I_1 < I_2 < I_3Rotation about the intermediate axis is unstable (tennis racket theorem). The body will exhibit periodic flipping, with ω1\omega_1 and ω3\omega_3 growing at the expense of ω2\omega_2Then reversing. This is the Dzhanibekov effect.

Problem 20: Hamilton--Jacobi for the Free Particle

(a) Solve the Hamilton—Jacobi equation for a free particle in one dimension: H=p2/(2m)H = p^2/(2m).

(b) Find the generating function S(x,P,t)S(x, P, t) that transforms to constant momentum PP.

(c) Show that the new coordinate X=S/P=xPt/mX = \partial S/\partial P = x - Pt/m (a freely moving coordinate).

Solution:

(a) The Hamilton—Jacobi equation:

12m(Sx)2+St=0\frac{1}{2m}\left(\frac{\partial S}{\partial x}\right)^2 + \frac{\partial S}{\partial t} = 0

Separate variables: S(x,t)=W(x)EtS(x, t) = W(x) - Et where EE is the separation constant (energy).

12m(dWdx)2=E    W(x)=±2mEx\frac{1}{2m}\left(\frac{dW}{dx}\right)^2 = E \implies W(x) = \pm\sqrt{2mE}\,x

S(x,E,t)=±2mExEtS(x, E, t) = \pm\sqrt{2mE}\,x - Et

(b) With P=2mEP = \sqrt{2mE} (identifying the new momentum with 2mE\sqrt{2mE}):

E=P2/(2m),S(x,P,t)=PxP22mtE = P^2/(2m), \quad S(x, P, t) = Px - \frac{P^2}{2m}t

(c) The new coordinate:

X=SP=xPmtX = \frac{\partial S}{\partial P} = x - \frac{P}{m}t

The new Hamiltonian K=H+S/t=P2/(2m)P2/(2m)=0K = H + \partial S/\partial t = P^2/(2m) - P^2/(2m) = 0. All momenta and energies are constant. The new coordinate evolves as X=x0=constX = x_0 = \text{const} (the initial position).

The original coordinate: x=X+Pt/m=x0+v0tx = X + Pt/m = x_0 + v_0 t (uniform motion). \checkmark