Discrete Dynamical Systems - Main Theorem

TheoremKAM (Kolmogorov-Arnold-Moser) Theorem

Consider a nearly integrable Hamiltonian system with Hamiltonian $H = H_0(I) + \epsilon H_1(I, \theta)$ where $(I, \theta) \in \mathbb{R}^n \times T^n$ are action-angle variables, $H_0$ is integrable, and $\epsilon \ll 1$ is a small perturbation.

The KAM theorem states that for most initial conditions (in the sense of measure):

Invariant tori persist under small perturbations if the frequency vector $\omega = \partial H_0/\partial I$ satisfies a Diophantine condition (sufficiently irrational)
The surviving tori have measure approaching the full phase space as $\epsilon \to 0$
Motion on these tori remains quasi-periodic

However, resonant tori (with rational frequency ratios) are destroyed, creating chaotic zones. The complement of KAM tori has small but positive measure for small $\epsilon$ .

The KAM theorem is one of the most profound results in dynamical systems, explaining the coexistence of regular and chaotic motion in nearly integrable systems. For the standard map (a discrete version of the kicked rotor), KAM curves are circles in the $(x,y)$ plane that persist for small kick strength $K$ . As $K$ increases, resonant curves break first, creating chaotic seas, while the most irrational curves (golden mean rotation number) survive longest.

TheoremBirkhoff Ergodic Theorem

Let $f: X \to X$ be a measure-preserving transformation on a probability space $(X, \mu)$ . For any integrable function $\phi: X \to \mathbb{R}$ , the time average:

$\lim_{n \to \infty} \frac{1}{n} \sum_{k=0}^{n-1} \phi(f^k(x))$

exists for almost every $x \in X$ . If $f$ is ergodic (the only invariant sets have measure 0 or 1), then this limit equals the space average $\int_X \phi \, d\mu$ for almost every $x$ .

This theorem justifies computing long-term statistical properties from a single trajectory.

The Birkhoff ergodic theorem provides the mathematical foundation for statistical mechanics and numerical simulations. It guarantees that time averages converge and, for ergodic systems, equal ensemble averages. This allows experimental verification of theoretical predictions: a single long trajectory samples the measure $\mu$ correctly.

ExampleRotation Numbers for Circle Maps

For orientation-preserving homeomorphisms $f: S^1 \to S^1$ of the circle, the rotation number is:

$\rho(f) = \lim_{n \to \infty} \frac{\tilde{f}^n(x) - x}{n} \pmod{1}$

where $\tilde{f}$ is a lift to $\mathbb{R}$ . The rotation number is independent of $x$ and characterizes the dynamics:

If $\rho$ is rational $p/q$ , every orbit is periodic or asymptotically periodic
If $\rho$ is irrational and $f$ is a diffeomorphism, either all orbits are dense or there exists an invariant Cantor set

The Arnold tongues in parameter space correspond to regions where $\rho$ locks to rational values.

Remark

The KAM theorem explains why solar system orbits remain stable over billions of years despite mutual perturbations. Planetary orbits lie on KAM tori, protected from chaos by their irrational frequency ratios. However, asteroids in resonant orbits (Kirkwood gaps) exhibit chaotic behavior and are ejected or captured. This demonstrates KAM theory's direct physical relevance beyond mathematical abstraction.

These theorems—KAM and Birkhoff—form cornerstones of ergodic theory and Hamiltonian dynamics. KAM addresses the stability of quasi-periodic motion under perturbations, while Birkhoff justifies statistical approaches to deterministic systems. Together, they explain the delicate balance between order and chaos in conservative dynamics, where regular and chaotic regions coexist in intricate patterns throughout phase space.